Intel Announces New CPUs Integrating AMD Radeon Graphics

Subject: Processors | November 6, 2017 - 02:00 PM |
Tagged: radeon, Polaris, mobile, kaby lake, interposer, Intel, HBM2, gaming, EMIB, apple, amd, 8th generation core

In what is probably considered one of the worst kept secrets in the industry, Intel has announced a new CPU line for the mobile market that integrates AMD’s Radeon graphics.  For the past year or so rumors of such a partnership were freely flowing, but now we finally get confirmation as to how this will be implemented and marketed.

Intel’s record on designing GPUs has been rather pedestrian.  While they have kept up with the competition, a slew of small issues and incompatibilities have plagued each generation.  Performance is also an issue when trying to compete with AMD’s APUs as well as discrete mobile graphics offerings from both AMD and NVIDIA.  Software and driver support is another area where Intel has been unable to compete due largely to economics and the competitions’ decades of experience in this area.

View Full Size

There are many significant issues that have been solved in one fell swoop.  Intel has partnered with AMD’s Semi-Custom Group to develop a modern and competent GPU that can be closely connected to the Intel CPU all the while utilizing HBM2 memory to improve overall performance.  The packaging of this product utilizes Intel’s EMIB (Embedded Multi-die Interconnect Bridge) tech.

EMIB is an interposer-like technology that integrates silicon bridges into the PCB instead of relying upon a large interposer.  This allows a bit more flexibility in layout of the chips as well as lowers the Z height of the package as there is not a large interposer sitting between the chips and the PCB.  Just as interposer technology allows the use of chips from different process technologies to work seamlessly together, EMIB provides that same flexibility.

The GPU looks to be based on the Polaris architecture which is a slight step back from AMD’s cutting edge Vega architecture.  Polaris does not implement the Infinity Fabric component that Vega does.  It is more conventional in terms of data communication.  It is a step beyond what AMD has provided for Sony and Microsoft, who each utilize a semi-custom design for the latest console chips.  AMD is able to integrate the HBM2 controller that is featured in Vega.  Using HBM2 provides a tremendous amount of bandwidth along with power savings as compared to traditional GDDR-5 memory modules.  It also saves dramatically on PCB space allowing for smaller form factors.

View Full Size

EMIB provides nearly all of the advantages of the interposer while keeping the optimal z-height of the standard PCB substrate.

Intel did have to do quite a bit of extra work on the power side of the equation.  AMD utilizes their latest Infinity Fabric for fine grained power control in their upcoming Raven Ridge based Ryzen APUs.  Intel had to modify their current hardware to be able to do much the same work with 3rd party silicon.  This is no easy task as the CPU needs to monitor and continually adjust for GPU usage in a variety of scenarios.  This type of work takes time and a lot of testing to fine tune as well as the inevitable hardware revisions to get thing to work correctly.  This then needs to be balanced by the GPU driver stack which also tends to take control of power usage in mobile scenarios.

This combination of EMIB, Intel Kaby Lake CPU, HBM2, and a current AMD GPU make this a very interesting combination for the mobile and small form factor markets.  The EMIB form factor provides very fast interconnect speeds and a smaller footprint due to the integration of HBM2 memory.  The mature AMD Radeon software stack for both Windows and macOS environments provides Intel with another feature in which to sell their parts in areas where previously they were not considered.  The 8th Gen Kaby Lake CPU provides the very latest CPU design on the new 14nm++ process for greater performance and better power efficiency.

This is one of those rare instances where such cooperation between intense rivals actually improves the situation for both.  AMD gets a financial shot in the arm by signing a large and important customer for their Semi-Custom division.  The royalty income from this partnership should be more consistent as compared to the console manufacturers due to the seasonality of the console product.  This will have a very material effect on AMD’s bottom line for years to come.  Intel gets a solid silicon solution with higher performance than they can offer, as well as aforementioned mature software stack for multiple OS.  Finally throw in the HBM2 memory support for better power efficiency and a smaller form factor, and it is a clear win for all parties involved.

View Full Size

The PCB savings plus faster interconnects will allow these chips to power smaller form factors with better performance and battery life.

One of the unknowns here is what process node the GPU portion will be manufactured on.  We do not know which foundry Intel will use, or if they will stay in-house.  Currently TSMC manufactures the latest console SoCs while GLOBALFOUNDRIES handles the latest GPUS from AMD.  Initially one would expect Intel to build the GPU in house, but the current rumor is that AMD will work to produce the chips with one of their traditional foundry partners.  Once the chip is manufactured then it is sent to Intel to be integrated into their product.

Apple is one of the obvious candidates for this particular form factor and combination of parts.  Apple has a long history with Intel on the CPU side and AMD on the GPU side.  This product provides all of the solutions Apple needs to manufacture high performance products in smaller form factors.  Gaming laptops also get a boost from such a combination that will offer relatively high performance with minimal power increases as well as the smaller form factor.

View Full Size

The potential (leaked) performance of the 8th Gen Intel CPU with Radeon Graphics.

The data above could very well be wrong about the potential performance of this combination.  What we see is pretty compelling though.  The Intel/AMD product performs like a higher end CPU with discrete GPU combo.  It is faster than a NVIDIA GTX 1050 Ti and trails the GTX 1060.  It also is significantly faster than a desktop AMD RX 560 part.  We can also see that it is going to be much faster than the flagship 15 watt TDP AMD Ryzen 7 2700U.  We do not yet know how it compares to the rumored 65 watt TDP Raven Ridge based APUs from AMD that will likely be released next year.  What will be fascinating here is how much power the new Intel combination will draw as compared to the discrete solutions utilizing NVIDIA graphics.

To reiterate, this is Intel as a customer for AMD’s Semi-Custom group rather than a licensing agreement between the two companies.  They are working hand in hand in developing this solution and then both profiting from it.  AMD getting royalties from every Intel package sold that features this technology will have a very positive effect on earnings.  Intel gets a cutting edge and competent graphics solution along with the improved software and driver support such a package includes.

Update: We have been informed that AMD is producing the chips and selling them directly to Intel for integration into these new SKUs. There are no royalties or licensing, but the Semi-Custom division should still receive the revenue for these specialized products made only for Intel.

Source: Intel

November 6, 2017 | 02:15 PM - Posted by Droid126

Wow, this is the clearest sign yet that the apocalypse is upon us.

November 7, 2017 | 07:41 AM - Posted by Power (not verified)

Is it you, Nvidia?

November 7, 2017 | 03:47 PM - Posted by Coppermine (not verified)

Meltdown for some people. Can clearly see in the replies below, especially the one thats blabbering away incoherently with multiple large scale posts.

November 7, 2017 | 04:02 PM - Posted by James

That guy always makes ridiculously long, somewhat off topic and/or babbling post. It isn’t limited to this thread unfortunately.

November 7, 2017 | 07:42 PM - Posted by Coppermine (not verified)

Yeah, noticed that.

Well, expect more meltdowns incoming because Raja Koduri has just resigned from AMD. And saw one web article says he's moving to Intel!

Whoah! What a week!

November 8, 2017 | 06:50 PM - Posted by AnnoyedTownMouse (not verified)

It's just trolling the gaming git peckerwoods time for the posts that are replying to your equally blabbering rants and FUD.

November 9, 2017 | 05:46 AM - Posted by Coppermine (not verified)

No matter how many times you changed your posting name, people will always be able to recognize the blabbering dolt that you have been here.

November 8, 2017 | 06:50 PM - Posted by AnnoyedTownMouse (not verified)

It's just trolling the gaming git peckerwoods time for the posts that are replying to your equally blabbering rants and FUD.

November 9, 2017 | 06:05 AM - Posted by Coppermine (not verified)

And again, no matter how many times you changed your posting name, people will always be able to recognize the blabbering dolt that you have been here.

November 8, 2017 | 06:43 PM - Posted by Coppermine (not verified)

Yes, the sign has arrived https://www.anandtech.com/show/12017/intel-to-develop-discrete-gpus-hire... quote "On Monday, Intel announced that it had penned a deal with AMD to have the latter provide a discrete GPU to be integrated onto a future Intel SoC. On Tuesday, AMD announced that their chief GPU architect, Raja Koduri, was leaving the company. Now today the saga continues, as Intel is announcing that they have hired Raja Koduri to serve as their own GPU chief architect. And Raja's task will not be a small one; with his hire, Intel will be developing their own high-end discrete GPUs."

November 8, 2017 | 08:13 PM - Posted by AnnoyedTownMouse (not verified)

Good for Raja(1), and that means that both Nvidia's and AMD's patents on unified shaders are about to run out and maybe some other AMD and Nvidia patants are about to expire. Now for the patents/copyrights to the x86 32 and x86 64 bit AMD ISA extentions to expire or be forcefully expired so Nvidia/others can make x86 ISA based custom APU/SOC designs.

The x86 32/64 bit ISA is damn near a public utility for PCs/Laptops so let's force Intel and AMD license that x86 32/64 bit ISA to others!

All the processor parts suppliers are just parts supplier filth to the PC/laptop/phone/tablet OEMs anyways so more competition keeps those filthy pocessor parts supplier scum in their proper market place competing for the OEM's business!

(1)

"Intel to Develop Discrete GPUs, Hires Raja Koduri as Chief Architect & Senior VP"

https://www.anandtech.com/show/12017/intel-to-develop-discrete-gpus-hire...

November 9, 2017 | 05:43 AM - Posted by Coppermine (not verified)

Yet I see more lies and incoherent rantings in your replies. Also you definitely did not read the cross license agreement between Intel and AMD. Guess you have also forgotten about the other x86 licensees like VIA (which nowadays found itself a niche in east asian).

November 12, 2017 | 08:34 PM - Posted by IntelGetsRajaAcceloratorsForProMarkets (not verified)

The cross license between Intel and AMD is because AMD owns the x86 64 bit ISA extentions and Intel owns the x86 16/32 bit x86 ISA extentions. Intel's Itanium was a big flop so Intel hasd to go with AMD's x86 64 bit ISA extentions.

Via is there from the time of IBM and IBM needing more than a single source of x86 ISA based parts. IBM was not about to let a single supplier get IBM by the nads.

Anyone still using x86 has to license from both Intel and AMD for the x86 16/32 bit(Intel) and the x86 64 bit(AMD). Now for this MCM Radeon semi-custom Polaris die is sourced from AMD in pretty much the same way that an AIB GPU markers sources their GPU dies. No big licensing required with AMD's AIB partners and the same for Intel and that semi-custom Radeon Polaris and HBM2. With Intel's EMIB there is still a silicon interposer but that interposer is only there to host the traces that connect the Radeon GPU to the HBM2(small bridge interposer) with the rest of the EMIB MCM made of resin/plastic. The only semi-custom work between AMD and Intel mostly concerned the packagng and a custom fitting of thigs on that EMIB MCM. AMD just shipped the semi-custom die that was compatable with Intel's EMIB mounting requirement with no need for any on GPU die licenseing between Intel and AMD required.

Gamers do not matter a whole lot for AMD, Intel and Nvidia as they all know where the real money is and those gaming GPU's are just the runts of the die process that allow the processor makers to unload the under-performing dies to the consumer markets. Intel's got Raja and Raja is making a GV100 competitor for Intel and Intel will source gaming dies from AMD wihile Intel's main focus for GPUs will be the enterprise GPU compute/AI markets.

Dream on with your gamer delusional self, gamers get the runts becaise that's all gamers will pay for! Pro market Money talks gaming Mooks walk!

November 6, 2017 | 02:34 PM - Posted by Mr.Gold (not verified)

"AMD getting royalties from every Intel package sold"

AMD doesn't sell the GPU die ? Is Intel manufacturing the GPU under an AMD license ?

And from AMD history, I wouldnt be surprised if AMD is getting pennies per each of those $600+ chip sold.

November 6, 2017 | 02:37 PM - Posted by Benjamins (not verified)

Well its not going to be on every Intel CPU, just the xxxxG SOC parts.

November 6, 2017 | 04:48 PM - Posted by Purgill (not verified)

Looking at this http://www.amd.com/en-us/press-releases/Pages/press-release-2017oct24.aspx note the sentence...

"In the quarter, AMD closed a patent licensing transaction which positively impacted revenue in the segments."

Also this is in the EESC division part! Is that client Intel or someone else?

November 6, 2017 | 09:00 PM - Posted by OutWithTheDogFoodIntelGraphics (not verified)

That's probably the asian server deal IP licensing with Between AMD/Chinese partner and

EESC(Enterprise Embedded and Semi-Custom) So it depends if the IP licensing branch is under that EESC division. AMD has a Lot of SeaMicro server IP that is licensed to AMD's server partners also. There is still that semi-custom gaming console AtariBox deal and AMD's embedded APUs are used in lots of digital signage, slot machines and medical equipment.

AMD is also working on a workstation grade APU(Interposer Based) with Zen cores/Vega GPU die/HBM2 die stacks but that will be for the professional graphics market. So maybe a portable Workstation grade APU that will not come at a low price point but will be fairly powerful.

November 7, 2017 | 08:42 AM - Posted by Coppermine (not verified)

Unlikely, because AMD already done that much earlier. Talking about this https://www.forbes.com/sites/patrickmoorhead/2016/04/21/amd-forms-china-... Have you forgotten that?

Also AMD's SeaMicro division is buried six foot under, Skybridge cancelled, Seattle is abandoned and left to rot at SoftIron, Hierofalcon gone off the radar, K12 has disappeared into thin air, etc. The list of AMD's projects that didn't come to fruition is numerous. So nobody wants any of those IPs.

Ataribox is unlikely to succeed due to pricing and being crowd funded. Ahem! Crowd funded. Yes, funds required first before any IP licensing agreement can even be negotiated. Its "no money no talk". Ataribox has also been heavily criticised especially considering the type of games its going to run and the asking price.

Sure, those few AMD's embedded APUs design wins highlighted in the news feed are all they have. But AMD's design wins are largely outnumbered by competitors like Intel and fabless ARM chipmakers that were not in the news feeds. Didn't you noticed Intel's IoT division results? Those are not console chips.

For your information, AMD's embedded APUs represents a very tiny fraction of less than 1% of the total emebedded market. After rounding to nearest integer that would be 0% actually. AMD so insignificant that it is not even worth mentioning in this upcoming report https://www.marketwatch.com/story/embedded-systems-market-forecast-2017-... but its not a surprise that Intel is in the list.

AMD working on a workstation grade APU? We've been hearing that for years even before Zen and its very unlikely to become fruition given AMD's limited resources. If AMD would have done it then their first new generation APUs would have used it, but that never happened despite all the surrounding continuous hype. Yes, there will be next year and if not then the year after that, and so on....

November 7, 2017 | 10:42 AM - Posted by OhSuchDaftynessAboundsInSome (not verified)

Deals(Asian/other markets) are announced sometimes before payments are required to be made and deals are signed in advance of payment date/s. And that accounting of the funds sometimes shows up on the balance sheets more than a few business quarters later!

"Also AMD's SeaMicro division is buried six foot under, Skybridge cancelled, Seattle is abandoned and left to rot at SoftIron, Hierofalcon gone off the radar, K12 has disappeared into thin air, etc. The list of AMD's projects that didn't come to fruition is numerous. So nobody wants any of those IPs."

Even though AMD has shuttered its SeaMicro businsee there still remains plenty related SeaMicro Related IP like the FreedomFabric/other SeaMicro IP that AMD will license to its OEM server partners for their server systems. Seattle is the Opteron A1100 ARM server SKUs currently is use by AMD's ARM based server partners Seattle makes use of not the K12 custom ARM design but uses ARM Holdings refrence design A57 CPU cores and AMD has a 5-10+ years contratural agreement to support its ARM server partners and their Opteron(Seattle) A1100 based clients. K12 is not on the roadmap but that Project was completed by Jim Kell/Keller's K12 Team. In 2018 is when AMD's management stated that K12 is scheduled to be announced and by SEC requirement AMD will have to report the K12 project's status to the SEC in AMD's 10K filing so we will know for sure of K12's funal status. Whatever K12's status is that K12 IP is still available to AMD to make use of as that's a done project as far as the design process goes with K12's custom ARMV8A ISA running micro-arch. Project Skybridge is not directly related to the K12 project, or the Seattle(Opteron A1100), and Skybridge was a dual usage motherboard design where socket compatable ARM or x86 CPUs could use the same motherboard.

"Ataribox is unlikely to succeed due to pricing and being crowd funded. Ahem! Crowd funded. Yes, funds required first before any IP licensing agreement can even be negotiated. Its "no money no talk". Ataribox has also been heavily criticised especially considering the type of games its going to run and the asking price. "

That's on the AtariBox investores as AMD is out of the loop once AMD's work is done and I'll bet that the AtariBox is based on a design that's very close to other AMD embedded designs in not exactly the same. So revenuse generated will more than likely be based on the old Atari games that people want to play more than the hardware used. But some hardware sales revenues are still revenues for whatever quarters that AMD cane make from the AtariBox, ditto for mining GPU sales, etc!

"Didn't you noticed Intel's IoT division results?" Intel has shut them down now so Intel really is a little to late for that IOT thingy market and the smartphone SOC market, but maybe the smartphone radio market is there for Intel a little more.

"AMD working on a workstation grade APU? We've been hearing that for years even before Zen and its very unlikely to become fruition given AMD's limited resources"

This Intel/AMD project is proof positive that complete systems on a module are possible unsing HBM2 from AMD as AMD's IP is included for that Radeon/HBM2 part of this Intel projest. And the Workstation market will pay the proper markups for any AMD workstation Grade APU on an Interposer with an even more powerful complement of Vega nCUs/Shaders and Zen cores than these Intel SKUs. So That project for a workstation grade All AMD workstation APU on and interposer project will be easy to fund as AMD will get a proper markup from the Professional Workstation users who can write off the costs as a business expense.

That's smart for AMD to focus on the Professional markets because that's where the big revenues come from.

November 7, 2017 | 10:54 AM - Posted by OhSuchDaftynessAboundsInSome (not verified)

Please Remove this Post too many spelling mistakes, etc.

November 7, 2017 | 10:53 AM - Posted by OhSuchDaftynessAboundsInSome (not verified)

Edited and Re-Posted to fix egregious spelling mistakes!

Deals(Asian/other markets) are announced sometimes before payments are required to be made and deals are signed in advance of payment date/s. And that accounting of the funds sometimes shows up on the balance sheets more than a few business quarters later!

"Also AMD's SeaMicro division is buried six foot under, Skybridge canceled, Seattle is abandoned and left to rot at SoftIron, Hierofalcon gone off the radar, K12 has disappeared into thin air, etc. The list of AMD's projects that didn't come to fruition is numerous. So nobody wants any of those IPs."

Even though AMD has shuttered its SeaMicro business there still remains plenty related SeaMicro Related IP like the Freedom Fabric/other SeaMicro IP that AMD will license to its OEM server partners for their server systems. Seattle is the Opteron A1100 ARM server SKUs currently is use by AMD's ARM based server partners Seattle makes use of not the K12 custom ARM design but uses ARM Holdings reference design A57 CPU cores and AMD has a 5-10+ years contractual agreement to support its ARM server partners and their Opteron(Seattle) A1100 based clients. K12 is not on the road-map but that Project was completed by Jim Keller/Keller's K12 Team. In 2018 is when AMD's management stated that K12 is scheduled to be announced and by SEC requirement AMD will have to report the K12 project's status to the SEC in AMD's 10K filing so we will know for sure of K12's final status. Whatever K12's status is that K12 IP is still available to AMD to make use of as that's a done project as far as the design process goes with K12's custom ARMV8A ISA running micro-arch. Project Skybridge is not directly related to the K12 project, or the Seattle(Opteron A1100), and Skybridge was a dual usage motherboard design where socket compatible ARM or x86 CPUs could use the same motherboard.

"Ataribox is unlikely to succeed due to pricing and being crowd funded. Ahem! Crowd funded. Yes, funds required first before any IP licensing agreement can even be negotiated. Its "no money no talk". Ataribox has also been heavily criticised especially considering the type of games its going to run and the asking price. "

That's on the AtariBox investors as AMD is out of the loop once AMD's work is done and I'll bet that the AtariBox is based on a design that's very close to other AMD embedded designs if not exactly the same. So revenues generated will more than likely be based on the old Atari games that people want to play more than the hardware used. But some hardware sales revenues are still revenues for whatever quarters that AMD can make from the AtariBox, ditto for mining GPU sales, etc!

"Didn't you noticed Intel's IoT division results?" Intel has shut them down now so Intel really is a little to late for that IOT thingy market and the smartphone SOC market, but maybe the smartphone radio market is there for Intel a little more.

"AMD working on a workstation grade APU? We've been hearing that for years even before Zen and its very unlikely to become fruition given AMD's limited resources"

This Intel/AMD project is proof positive that complete systems on a module are possible using HBM2 from AMD as AMD's IP is included for that Radeon/HBM2 part of this Intel project. And the Workstation market will pay the proper markups for any AMD workstation Grade APU on an Interposer with an even more powerful complement of Vega nCUs/Shaders and Zen cores than these Intel SKUs. So That project for a workstation grade All AMD workstation APU on and interposer project will be easy to fund as AMD will get a proper markup from the Professional Workstation users who can write off the costs as a business expense.

That's smart for AMD to focus on the Professional markets because that's where the big revenues come from.

November 7, 2017 | 02:04 PM - Posted by Coppermine (not verified)

Then clearly you missed this http://www.amd.com/en-us/press-releases/Pages/press-release-2017may01.aspx The money from the joint venture was already paid. Next time stop making baseless assumptions.

Have seen any companies talked about using SeaMicro's IPs recently? Hardly, not even AMD themselves because everyone else already have their own established tried and tested cluster interconnect technology. Heck, even AMD used Mellanox's Infiniband in their Project 47 "supercomputer" instead of AMD's own Freedom Fabric. Now that pretty much shows that there's no demand whatsoever for any of SeaMicro's IPs (already considered a full write-off after SeaMicro's demise).

None of the top tier server manufacturers used nor adopted AMD's Seattle or Opteron A1100 series. Surprisingly not even HPE, who had used Applied Micro's ARM chips https://www.pcper.com/news/Systems/Applied-Micros-X-Gene-SoC-Will-Be-Use... in their Moonshot servers. Likewise same goes for GigaByte, which had ARM server solutions like this http://b2b.gigabyte.com/ARM-Server/Cavium-ThunderX but none using AMD's Opteron A1100 series chips.

Have you tried looking for AMD's Seattle or Opteron A1100 series server news after launch? You will find none, except for rather unknown SoftIron https://softiron.com/ Heck even progress at SLS (Silver Lining Solutions) has gone quiet. If you are lucky with the search engines then you may find things like the cancelled 96Boards Huskyboard, and the indefinitely delayed LeMaker Cello http://www.lenovator.com/product/103.html And the reasons for that is down to problems with AMD's Seattle chip itself. Its broken with PCI Express bus problems and also memory compatibility issues. Example http://mrpogson.com/2016/09/01/lemaker-cello-news/ quote "The PCIE power rails, clock, reset signal and tx lane are all work, but only the rx lane can not get the signals". Buyers who did receive the limited number of boards shipped out says its become a paperweight. Additionally Phoronix who also received one https://www.phoronix.com/scan.php?page=news_item&px=AMD-ARM-In-My-Hands until now have yet to be able to run tests on it.

Nopes, K12 never got off the ground nor had started. While at AMD, Jim Keller was also in charge of Seattle project as well. He had several projects under his belt including Zen. He also had to teach AMD's engineers about ARM design also https://www.slashgear.com/amd-chip-guru-jim-keller-arm-left-team-a-littl... and that will delay subsequent ARM related projects. That is why after the Seattle project (was it even fully completed?), all projects related to ARM chips like Project Skybridge and Amur ended up on the chopping block. AMD's HieroFalcon and K12 would have also ended up on the chopping block as well. Thus all resources went into x86 Zen project.

Do you even understand what crowd funding means? If Ataribox does not have much appeal to a wider audience then it may never reach its funding goal (and that subsequently meant cancellation). Ataribox was predicted by many as incoming failure as it doesn't have the GPU grunt of consoles, not geared towards AAA games and does not have exclusive titles as those found on the Playstation4 and Xbox One. Furthermore the pricing is around Playstation4 and Xbox One as well. That did not go un-noticed by most people.

And you are wrong again, Intel's IoT division has not been shutdown and is doing very well, check https://www.anandtech.com/show/11972/intel-announces-q3-2017-results Intel's IoT division is also growing as well. That "IoT" just a new name for the embedded market sector. The division for smartphone chips is under mobile business which has been merged into Client Computing Group since 2015 http://techreport.com/news/28082/intel-client-computing-group-combines-p... Seems that you don't read the news well or simply didn't.

As for that "workstation APU", well you can keep dreaming all you want. Yeah, next year, the year after that and so on. Do note Intel has huge R&D funding and resources (including facilities for fabrication and assembly) to pull it off. However things are very different on AMD's side where they had to rely on (and contracted to) third party facilities most of the time.

November 7, 2017 | 04:03 PM - Posted by DemJokezOnYouzKnuckleheadz (not verified)

No Seamicro's IP is now called AMD IP as AMD has shut that unit down and retired the branding. Jim Keller was only in charge of designing the Zen core x86 micro-arch and the K12 custom ARMv8A ISA runnng micro-arch and that was a full enough plate for any project manager.

Intel IOT has lost some thingys and the bleeding will continue there(1). AMD's K12 micro-arch was finished and AMD will have to give a K12 status come 2018 10K time. AMD has already state that K12 is pushed back to 2018 so that's the time frame to get a final answer. Opteron A1100 is getting that 5-10 year product availability support and AMD is even still supporting some of its SeaMicro X-Division's clients by contract even though AMD has suttered SeaMicro.
SeaMicro's IP and patent portfolio is in AMD's hased since AMD aquired SeaMicro and who Knows how much of the Freedom Fabric IP DNA is IN the Infinity Fabric.

I know what crowd funding is and that Ataribox is most likely using a pre-existing AMD Embedded design so not much work for the AMD side other than a little system integration help for Ataribox for a nominal fee. Just look at the Fitlet PC based AMD embedded options, that's low budget also. It's no big deal if Ataribox fails to AMD, Joe Blow, and his dog!

And that workstation interposer based APU makes perfect sense for the portable workstation market branded under some Pro/WX branding name by AMD and sold at higher markup for peofessionla usage. AMD can get investment loans easily just on the It's Epyc CPU sales revenues projections alone and 2018 is going to be even more intresting for AMD than 2017.

An AMD/Intel for Apple semi-custom design/s by way of Intel may not be able to be sold in the wider market if Apple is the client, and Apple's cash fund/holdings(2) is as big as Intel's market cap(Intel (INTC) as of November 03, 2017 is $216.87B < than Apple's cash stash). Ah Ha Ha Ha!

Apple's Money, It's a gas

grab that dosh with both hands and make an offshore stash!

Intel's Apple's biotch!

And so is AMD!

It's because AMD is not as big as Apple's loose couch cushion cash!

Tim Cook to Intel/AMD, now hold each other's hands and then have a big long Kiss!

(1)

"Intel: Joule's burned, Edison switched off, and Galileo – Galileo is no more"

https://www.theregister.co.uk/2017/06/20/intel_joules_edison_galileo/

(2)

"Apple's cash hoard swells to record $256.8 billion"

https://www.cnbc.com/2017/05/02/apples-cash-hoard-swells-to-record-256-8...

November 7, 2017 | 05:46 PM - Posted by Coppermine (not verified)

So, can you explain why AMD did not use their very own Freedom Fabric in their Project 47 supercomputer? Using Freedom Fabric certainly that would be good pubilicty and advertisement to entice potential customers. But that didn't happen because basically its obselete in today's market. Strangely enough AMD did not use their own Seattle chips in servers sold by SeaMicro. Didn't you know that there are no SeaMicro servers using AMD's own Opteron A1100 CPUs? Heck even AMD don't have faith in their own chips. And these chips were probably finally dumped to SoftIron, free of charge even. You won't find them anywhere else nowadays.

As for K12, not 2018 but 2017 as shown here https://www.anandtech.com/show/9232/amds-k12-arm-cpu-now-in-2017 and that didn't happen. And before that, this https://www.anandtech.com/show/7990/amd-announces-k12-core-custom-64bit-... was the original plan. AMD kept delaying it year after year, until it disappears on future roadmaps. AMD's K12 was never ever going to be finished because it never got started, just exists on old roadmap slides only. Yes, K12 is simply vaporware. AMD has long since stopped talking about ARM related stuff, and often avoid the subject altogether in earnings conference calls.

Then look how long it took before AMD's Seattle finally launch https://www.anandtech.com/show/9956/the-silver-lining-of-the-late-amd-op... That's in 2016 which was late, quote "We were expecting to see Opteron A1100 based servers at the end of 2014, but instead we got more than a year of almost complete silence". And with problematic PCI Express bus and memory compatibility issues, that would be a sign that it was rushed out the door and was not completed properly. Perhaps that's why all the top tier server manufacturers don't want it at all. And you expect K12 to be ready after all those problems in AMD's first ARM based server product?

And again, you are wrong and lying as well. Let me quote from https://www.anandtech.com/show/11972/intel-announces-q3-2017-results "Internet of Things group had revenues go up 23% to $849 million". That Q3 earnings report is the latest. Now let's look at Q2 earnings here https://www.anandtech.com/show/11972/intel-announces-q3-2017-results quote "Intel’s Internet of Things revenue was up 26% year-over-year to $720 million in revenue. This segment has had some product changes over the last quarter, with Intel discontinuing their Galileo, Joule, and Edison lines, but Intel is still moving forward in this space with other products for industrial, video, and automotive. Despite the shakeup, this group increased its operating income 56% to $139 million". They also include the Galileo, Joule and Edison part in there and yet still had profits. Certainly they are not "bleeding" as you had falsely claimed.

Typically crowd funded AMD powered devices don't have a good history. Have you heard of crowd funded SmachZ? It was going to use AMD's Merlin Falcon but was delayed many times until the hardware designer/manufacturer parted ways. Now they are starting from scratch again, supposedly using a newer un-announced AMD APU.

And those Fitlets are actually not cheap because they are industrial mini PCs. But for people longing for such a device using an AMD chip, there is very little choice unfortunately. Incidentally Compulab is an Israeli company http://www.fit-pc.com/web/about/ Do you know that?

Nopes, AMD cannot get any more loans since they have still a large outstanding debt on current loans as well as below than ideal cash level. As for AMD's Epyc revenues, not much since launch as can be seen from its Q3 2017 earnings report https://www.anandtech.com/show/11961/amd-announces-q3-earnings-71m-in-th... under the EESC division. Looking at the income versus revenue percentage (which comes to a paltry 10% margin only), consists mostly of console chips. For your information, datacenter and server CPUs usually command the highest margins (as shown by Intel's earnings). Futhermore a lower Q4 2017 guidance meant that not much meaningful earnings from Epyc CPUs in the next quarter as well.

November 7, 2017 | 06:55 PM - Posted by Coppermine (not verified)

Link error above.

This is Intel's Q2 2017 earnings https://www.anandtech.com/show/11674/intel-announces-q2-fy-2017-earnings...

November 8, 2017 | 10:08 AM - Posted by AMDisBeyondConsumerNOW (not verified)

That project 47 is AMD's and its partner's(Mostly) supercomputer in a cabinet and how do you Know if AMD's Infinity Fabric does not have a little bit of that Freedom Fabric IP included and the Infinity Fabric is in fact just a coherent version of HyperTransport+ like Charlie over at semiaccurate. And for all anyone knows AMD/project 47 partner may not even be using the Freedom Fabric branding for that interconnect IP any longer! And AMD is still on the hook for supporting its SeaMicro clients for at least 5+ more years because of contract obligations.

You can bring up AMD's delivery date issues but that's does not have much to do with AMD's still retaining all the IP that SeaMicro had or that ATI had and that Freedom Fabric branding may no longer be in use as is in part included in the Infinity Fabric IP that's nothing more than a superset of AMD's HyperTracsport with Coherency and other Enhannsements.

And then we have Charlie over at semiaccurate's analysis of this new Intel nano-motherboard on an EMIB MCM(1).

So basically this is an Intel nano-motherboard with limited interposer usage(Just enough to interface the GPU to the HBM2) and that's all purchased from AMD like an AIB partner would(No Licensing Required) because Intel is really not Involved much on the GPU/HBM2 side of the IP other than a similar manner to what an AIB partner would be for a discrete AMD GPU on a PCB based PCIe card. So Intel's only contribution is the EMIB part(That is Innovative) and adding its CPU die on that same EMIB/MCM module. Intel's is no different than a PCIe based AIB partner in this equation save a little custom packaging work necessary from AMD's Semi-Custom division that has all of AMD's RTG resources to utilize also.

So Intel is not really integrating that AMD GPU IP on its CPUs Intel has only created a micro-motherboard and just like a PCIe Card AIB partner sources the GPU/HBM2 part from AMD/HBM2 partner to be integrated this product in onto Intel's mini-motherboard, EMIB/MCM module. The only real innovation is in the packaging(EMIB) and not so much anything else above and beyond what AMD does with its PCIe Card discrete GPU AIB partners.

This is new territory for Intel into this mini-motherboard/EMIB market for laptop mainboards as that's the only really new technology and that's mostly packaging technology that will quickly find its way into Appple products before AMD comes along with some smaller and more tightly integrated CPU cores die to Vega die APUs on a full interposer IP that will be an all AMD affair.

Intel's is only doing this to delay Apple's departure from using Intel's CPUs and Apple surely had a hand in that process. Wait for AMD's workstation Grade APUs on an ACTIVE interposer with actual curcuitry etched onto the interposer in addition to the traces and the various specilized CPU and GPU modular dies used to create that workstation grade APUs on an interposer SKUs. Navi is not going to be much different from Vega micro-arch wise it's just that Navi will use smaller easier to fabricate GPU die chiplets and AMD will move the Infinity Fabric fully onto the interposer's silicon substrate and attatche the GPU/CPU die to that Active Interposer based MCM/Package.

The Vega 10 die basd Vega GPU variants were never intended for the consumer market first and formost as the Vega 10 base die layout with its excess of shader cores was designed first and formost for the Server/Workstation/AI markets. And AMD could very easily drop most of its consumer gaming business and focus on AI/Workstation/Server base GPU designs and still sell its GPUs in those markets, and that mining market. And AMD's Epyc CPU professional markets revenues will be large enough that AMD does not really need that gaming GPU market to survive and prosper.

Realistically AMD needs that Gaming market as a place to get rid of its lowest binned Vega dies in a revenue neutral way, or even a small profit producing way, only as that is a better use(better than throwing away) of any Vega 10 dies that fail to bin high enough to be use in the real revenue producing Radeon Pro WX 9100s and Radeon Instinct MI25 and other Radeon Professional compute/AI GPU SKUs.

The only reason that AMD's Vega 10 die base GPU are not as good for giming and power uasge compared to Nvidia's GP104 based products is that Vega 10 is really only for compute/AI and maybe Vega 11 will be a more gaming tuned Vega micro-arch based variant with a better ROP to shader ratio that's tuned for gaming. And Extra ROPs is the only reason why Nvidia GTX 1080ti(88 ROPs) and the TITAN XP's(96 ROPs) GPU variants outperform Vega 10(64 ROPs). Vega 10 has better compute performance than most of Nvidia's consumer and Quardo SKUs and the Vega 10 Die/Variant makes more revenues for AMD in the Professional markets.

(1) "Intel GPUs officially don’t work, AMD to the rescue"

https://semiaccurate.com/2017/11/06/intel-gpus-officially-dont-work-amd-...

November 8, 2017 | 04:31 PM - Posted by Coppermine (not verified)

And seems that you are a fan of Groo The Wanderer (a.k.a. Charlie Demerjian) and his doodles and often highly biased articles. Here https://hardforum.com/threads/who-is-charlie.1458433/ sums up about his point of view (and fanboyism). He forgot that Intel sold millions of CPUs that comes with that integrated HD Graphics in almost every market (not just in laptops and desktops). Sure, Intel HD Graphics may be slower but its functional and very low power as well. Also there are many other (huge) markets that do not require high gaming performance. But that does not mean its a failure. Seems that your "belief system" is clouding any objective reasoning.

November 8, 2017 | 06:57 PM - Posted by AnnoyedTownMouse (not verified)

No Just the fanboy peckerwood gamers so ready to goosestep when their team's products are not directly ment for gaming only. Ditto for the other team's rabid gaming monobrows that skulk the dark and dank recesses of mom's basement crying about the high prices on GPUs lately.

November 9, 2017 | 05:37 AM - Posted by Coppermine (not verified)

Again from your incoherent replies, you're definitely in full meltdown mode.

November 8, 2017 | 01:27 PM - Posted by AMDisBeyondConsumerNOW (not verified)

Branding names come and go(Freedom Fabric/Whatever) but the IP if rebranded as the Noodle Fabric would still be the same IP. So who knows if some ot that Freedom Fabric is inside the "Infinity Fabric"/Branding(Made up by Marketing Monkeys)

So the Infinity Fabric Is as Charlie at S/A states:

"On the surface it sounds like AMD has a new fabric to replace Hypertransport but that isn’t quite accurate. Infinity Fabric is not a single thing, it is a collection of busses, protocols, controllers, and all the rest of the bits. Infinity Fabric (IF) is based on Coherent Hypertransport “plus enhancements”, at a briefing one engineer referred to it as Hypertransport+ more than once. Think of CHT+ as the protocol that IF talks as a start.

Hypertransport lacked a lot of things that quite frankly were not on the roadmap when it was designed in the DEC Alpha days. It wasn’t really scalable as later Opterons brought into focus, had no real QoS capabilities, wasn’t very flexible, had no inherent security features, and lacked any authentication mechanism. If you throw in higher performance and greater efficiency you have the list of why AMD needed to redo their bus to get Infinity Fabric." (1)

Hell, Intel's HyperThreading is a marketing name for Intel's Version of SMT(Simultaneous multithreading) and Intel did not invent(2) SMT as noted in this wikipedia entry:

" Historical implementations[edit]

While multithreading CPUs have been around since the 1950s, simultaneous multithreading was first researched by IBM in 1968 as part of the ACS-360 project.[1] The first major commercial microprocessor developed with SMT was the Alpha 21464 (EV8). This microprocessor was developed by DEC in coordination with Dean Tullsen of the University of California, San Diego, and Susan Eggers and Henry Levy of the University of Washington. The microprocessor was never released, since the Alpha line of microprocessors was discontinued shortly before HP acquired Compaq which had in turn acquired DEC. Dean Tullsen's work was also used to develop the Hyper-threading (Hyper-threading technology or HTT) versions of the Intel Pentium 4 microprocessors, such as the "Northwood" and "Prescott". " (2)

So using Marketing names to keep track of IP usage is a losing bet and even the IPs code names can not be trusted fully. It's best to use the achedemic/ computing sciences terminology instead and that "Freedom Fabric'(marketing) branding is not the actual proper engineering naming for that IP.

(1)

"AMD Infinity Fabric underpins everything they will make"

https://semiaccurate.com/2017/01/19/amd-infinity-fabric-underpins-everyt...

(2)

"Simultaneous multithreading" [see: Historical implementations sub-heading]

https://en.wikipedia.org/wiki/Simultaneous_multithreading

November 8, 2017 | 05:34 PM - Posted by Coppermine (not verified)

On the subject of AMD's Freedom Fabric, I bet you didn't read this https://www.theinquirer.net/inquirer/news/2221111/amd-will-not-license-s... That certainly shows that AMD isn't going to license this IP to anyone else any time soon. Also by using an outside party cluster interconnect like Mellanox's Infiniband (which is very popular with supercomputer manufacturers nowadays), its quite a telling sign that AMD had no desire to develop nor use Freedom Fabric any further.

That SMT (Simultaneous Multi-threading) concept and technology is not new and existed even before Intel ever incorporated it into their CPUs as HyperThreading. However it was Intel that first introduced SMT technology into the x86 world. The advantages of SMT far outweigh its disadvantages. For one, it doesn't take up much die space https://www.anandtech.com/show/994/2 quote "Including Hyper-Threading only costs about 5% of die space".

However AMD's decision to use an derivative multi-threading technology called CMT (Cluster Multi-threading) turned out to be a disastrous one with AMD's Bulldozer core design. Perhaps AMD should have learned Sun Microsystem's failure in implementing CMT on their CPUs http://www.cpushack.com/2015/01/16/sun-ultrasparc-rock-when-is-a-core-no... "Larry Ellison, never one to mince his words said of the RK: “This processor had two incredible virtues: It was incredibly slow and it consumed vast amounts of energy. It was so hot that they had to put about 12 inches of cooling fans on top of it to cool the processor. It was just madness to continue that project.”". Does that sound familiar?

Eventually AMD had to "follow the leader", by finally ditching CMT and going for SMT like Intel (as well as other technology implemented in Intel's Core series CPUs) in their new CPU designs (like Zen). There's simply no getting away from using tried and proven technologies.

And again, since you're a fan of Groo The Wanderer (a.k.a. Charlie Demerjian), here https://hardforum.com/threads/who-is-charlie.1458433/ sums up about his point of view. Keep listening to him and you'll lose touch of any objective reality.

November 8, 2017 | 07:03 PM - Posted by AnnoyedTownMouse (not verified)

Charlie Demerjian knows a few things and accepts no ad content on his digital rag! But he offends your precious godhead corp that is an abusive monolopolistic market share holder and convicted of that on more than one occation nad forced to pay but a token amount compared to the damage done!

November 9, 2017 | 05:36 AM - Posted by Coppermine (not verified)

Still remember when Charlie "Dancing in the aisle" Demerjian wore a bunny suit? He's quite a character.

November 12, 2017 | 08:42 PM - Posted by IntelGetsRajaAcceloratorsForProMarkets (not verified)

Everybody wears a Bunny suit at anybodys chip fab or they can not go inside and Charlie dancing was just /s on his part.

It's just that you damn diry gaming gits have no clue how the real world works and you'll always be clueless in that regard. Gamers get the reject chips that do not make the professional grade! So that's disfunctional bins for disfunctional gamers and marketing is used to convence the hayseed gamers that gamers matter. Oh how marketing spins those yarns and the gaming bumpkins eat it up right out of the feces filled mud.

November 6, 2017 | 02:38 PM - Posted by Benjamins (not verified)

Now I want to see AMD best it with a 8c16t Zen + 24+ CU GPU with HBM2

November 6, 2017 | 02:54 PM - Posted by JohnGR

We aren't going to see it. Probably OEMs explained to AMD that because of that stupid Bulldozer design, AMD hardware doesn't sell. No matter how good Ryzen can be, it will NOT sell, because consumers not just don't trust AMD hardware, they also learned to avoid it.
So, even if AMD creates an APU monster, OEMs will take a pretty small risk by creating probably one laptop with that monster APU, not a series of models.

November 6, 2017 | 04:06 PM - Posted by TomS. (not verified)

The average consumer doesn't know and doesn't care about Intel vs AMD.

The response I get from family and friends when mentioning any of these companies is either "they make computer stuff, right?" or a blank stare. Even those who have some clue as to who these companies are they certainly don't follow any of them enough to know if any particular company currently has the better product or even what their current product line looks like.

These people buy computers based on price vs numbers vs aesthetic.

November 6, 2017 | 04:46 PM - Posted by Droid126

Yes, this. My friends are always asking me which laptop they should get. I ask how they use their computer, and what their budget is(usually less than $500). I hit up Amazon and find a few choices that would work for them. Then they ignore all of that effort and go to Best Buy and buy the cheapest POS they have. Then complain endlessly too me that its slow. Or I just print out an Intel i(X) logo and say buy one with this sticker that doesn't cost less than X.Even armed with an Intel sticker they will still come back with the cheapest AMD laptop they can find.

November 7, 2017 | 03:04 AM - Posted by JohnGR

The average consumer will get what the OEM or the salesperson will promote to them. And both OEMs or the salesperson will promote an Intel system, 99 times out of 100.

Those who might know a company name or two, that name will be Intel's. And I know plenty of people who know little to nothing about computers, but they have heard or seen that Intel brand, from a salesperson, or seen it on an advertisement. Best case scenario they never have heard of AMD. Worst case scenario, they have heard/read negative things about AMD in those last years.

I wish things where as simple as you say. They are not.

November 6, 2017 | 08:37 PM - Posted by wownwow (not verified)

Ask you wife and kids, which one they prefer or do they even know what is what? Their likely answer: Intel starts with i but AMD with a : )

November 6, 2017 | 08:52 PM - Posted by Droid126

No, most people know their computer "is windows", some know what brand it is(or can find it with direction), few people know their processor.

November 7, 2017 | 03:13 AM - Posted by JohnGR

Many kids will know at the age of 15 more than what we know now. They where born in an era where the internet was not something new and amazing. Just something that was always there. And wives now know more than just brands for clothes.

And as I replied above, even if they don't know about Intel or AMD, OEMs or the salesperson will promote an Intel system anyway.

November 6, 2017 | 09:12 PM - Posted by OutWithTheDogFoodIntelGraphics (not verified)

Not with Zen erasing that Bull-Dozer from history and AMD has plans for an interposer based Workstation Grade APU with Zen cores/Vega Die(Lots of nCUs)/HBM2 stacks.

Intel's still has too much pull($$$$$) with the laptop OEMs and this Intel part is probably for the customer that commissioned the part from Intel --Cough Apple Cough--. So this SKU may not be able to be sold on the open market and is an exclusive use only design that Intel was commissioned --Cough Ordered Cough-- by Apple to create.

November 7, 2017 | 03:16 AM - Posted by JohnGR

It will take years to erase Bull-S-dozer. That's why AMD was so aggressive with Ryzen pricing. They want to erase that disaster as fast as possible.

Yes Apple, is the first thing it comes to mind. But I believe OEMs like Dell will also get it.

November 7, 2017 | 09:10 AM - Posted by BustEmTrustsTeddy (not verified)

AMD still sold pleny of those Bull-S-Dozer CPUs in embedded devices and for PCs/Laptops and AMD's APU graphics was always better all around than Intel's dog food.

That said the Feds need to step in if AMD and Intel get too cozy with a larger market arrangement because AMD needs to be forced to compete with Intel and not collude with Intel.

I'd rather see the x86 ISA declared open for anyone to use if AMD and Intel get too much market penetration in bed together and both AMD and Intel forced to license the x86 ISA to Nvidia/others. AMD had better be designing a Zen based version of this SKU if the Intel SKUs are not for say Apple's/Other semi-custom client's exclusive use.

I'd rather the US justice Department's Anti-Trust division step in and force AMD and Intel to compete in the wider market than allow an AMDTel sort of Trust to dominate the PC/Laptop market. So Apple better be the semi-custom client that comsissioned/financed these Intel parts to be made for Apple's exclusive use only.

Anti-Trusts is a big concern with AMD and Intel coming together for too much x86/Graphics market share and no other x86 competition out of AMD because of such a cozy arrangement. AMD needs to be told that AMD and Intel are there to compete and not collude or both AMD/Intel need to be forced to license the x86 16/23 bit ISA(Intel) and the x86 64 bit ISA extentions(AND) to Nvidia/others to preserve competition in the x86/graphics market place.

Bull-S-Dozer CPUs may have not fully competed with Intel's CPUs in gaming but AMD's CPUs where good enough for some and the embedded market used pleny of Bull-S-Dozer CPUs.
So now AMD has its Zen CPUs and Vega Graphics IP to bring more competition to the market and that's somthing that needs to be required of AMD, ditto for Intel, and this arrangement needs to be looked at in the Legal Framework of the Anti-Trusts laws already on the books.

Gaming is really not that important in the overall market scheme of things that need to be competitive or the consumer will be the loser and not the winner.

November 7, 2017 | 10:01 AM - Posted by JohnGR

Bull-S-Dozer was selling to costumers, but how many of those end up happy? APUs do have nice GPUs, but consider how much lower that GPU performs because of those S-Dozer cores next to it.

Feds will do nothing because Intel and AMD will explain them that it is in USA's best interest ARM NOT win the fight against x86.

Anti-trusts didn't do much with contra revenue programs from Intel or Nvidia trying to create a closed ecosystem in the GPU market, making all other designs incompatible establishing a de facto monopoly, with PhysX, CUDA and GameWorks. You seem to forget those cases but you are getting upset with the collaboration between Intel and AMD. Double standards?

Gaming was not important in the past. It is what drives whatever growth we see in PCs today. With every quad core system from 2008 until today being able to run 90% of most common tasks on a PC, triple A gaming and enthusiasts is what drives grown in the PC market.

November 7, 2017 | 11:24 AM - Posted by OhSuchDaftynessAboundsInSome (not verified)

'Feds will do nothing because Intel and AMD will explain them that it is in USA's best interest ARM NOT win the fight against x86."

The Anti-trust Laws on the books can not be ignored by the Feds too long as Nvidia will bring up a suit in federal court if Nvidia's busines is adversely affected by unlawful collusion between AMD and Intel. The free competative markets will be held to the letter of the law, and Nvidia has deep enough pockets to sue for redress of grievances.

The US Justice Department can find itself in front of a judge explaning why it failed to enforce any laws that the US Justice Department has sworn to uphold. That's some splaning to do for Lucy The JD/Justice Department Official in front of the US courts and the appropriate congressional committees. You can sometimes break the law by not enforcing the laws on the books if you are a Fed, and are found to be in dereliction of duty.

This "USA's best interest" thing is what the Antitrust laws where passed to protect and that would reguire an act of congress and not AMD and Intel asking favors from any Government Department, especially if Nvidia filed a lawsuit in federal court.

The antitrust laws broke up the Standard Oil of Ohio Trust and they can be used to break apart any collusive agreement between AMD and Intel/Others.

The US is a nation of Laws and not a nation of men(women/etc.)!

You get the gist of that Elmer!

November 7, 2017 | 05:09 PM - Posted by JohnGR

AMD is selling GPUs to Intel. I really don't understand what is wrong here and why Feds need to do something. I bet if Nvidia was selling GPUs to Intel you would have no problem.

I like how you totally ignore the whole PhysX and stuff story.

If you just want to start shouting like a 5 years old kid because your favorite team lost, be my guest.

But if you want to do a mature conversation, I want your opinion on Nvidia's moves to make the GPU market, a close market, full of proprietary software that they where having the absolute control. Also your opinion on Contra Revenue program from Intel.

November 8, 2017 | 06:00 AM - Posted by PCIeLane (not verified)

Because that means less competition with a deal that involve 2 companies that together detain 99% of the X86 CPU market (every PC Mac included, many game consoles and servers), unlike AMD NVIDIA doesn't sell x86 CPU, nor CPU that compete with Intel that's the big difference. This is like AMD partnering with NVIDIA and then Intel, the only competitor, not allowed to produce compatible GPU (like with x86), would you consider that a good thing for competition?
Antitrust authorities will decide if and what to do.

What PhysX has to do with that? It's a piece of software not a deal between two companies witch are supposed to compete one against the other and I'm surprised of your ignorance on the fact that PhysX was born to run primarily on CPU (Novodex and Meqon, the father and mother of PhysX was designed without even considering accelerators) and that it run just fine on AMD CPUs.

November 8, 2017 | 01:22 PM - Posted by JohnGR

They are not collaborating in the CPU market. This collaboration doesn't affect the competition in the x86 CPU market. It's still there.

Nvidia doesn't offer x86 CPUs but it does have CPUs. Just ARM based.
x86 is a big deal because of Windows. With Windows getting ready to run on Qualcomm processors, no one stops Nvidia to make it's Tegra SOCs to be able to run Windows for ARM. That means tablets and laptops and desktops with ARM SOCs that can run Windows. It's just that Nvidia will have to invest to get those products ready. Nothing comes for free.

Also, Toyota doesn't sell x86 CPUs. No one says that EVERYONE should be able to produce ALL products for every market. That's why there are proprietary techs and patents all over the place. It's NOT communism.

What PhysX has to do with this? I like how you are prepared to write a book about other stuff, and here you just play it dumb and call me ignorant. PhysX was used to create a distorted market where Nvidia hardware would be able to show all effects in game, while the competing hardware will suffer or offer lesser quality. And while PhysX can run on the CPU, it was programmed to offer much better performance on the GPU. So, for example, you where getting 60fps on a system with a Geforce GPU and about 15 fps on a multicore CPU. Also Nvidia locked these techs, CUDA and PhysX. Those tech are unavailable if the Geforce GPU is secondary in the system. Your lovely company was checking the loyalty of it's customers and punishing those who where not loyal enough. PhysX unlocking patch proved that there where no incompatibility problems to justify the lock.

I could write more about PhysX, but I know I am losing my time, so I stop here.

November 8, 2017 | 07:13 PM - Posted by AnnoyedTownMouse (not verified)

"They are not collaborating in the CPU market", but thay are in the x86 ISA based CPU/Graphics market using Intel's x86 and AMD's Radeon. And most consumer PCs/Laptops use CPUs that execute that x86 ISA. And as much as I dislike Nvidia, Nvidia needs its chance at getting its GPUs on an Intel MCM/EMIB mini-motherboard SKU if that product is targeted for the wider laptop market.

Fair markets must be maintained and regulated to require fair competition for all the market players.

November 9, 2017 | 04:36 AM - Posted by JohnGR

You don't dislike Nvidia. Please.

November 12, 2017 | 08:56 PM - Posted by IntelGetsRajaAcceloratorsForProMarkets (not verified)

I dislike gaming git bumpkins that eat up all that marketing mindshare nonsense. Get back to ESPN with your football fandom crap and leave the technology to the engineers/scientists. Gamers get the pig slop dies served up in a gravy of feces and mud. Raja engineered Vega and the Vega 10 based die design for compute/AI markets with just enough ROPs to alow the binned slop to go to the gaming piggys!

I love to see gaming git tears when they learn the truth about their Technology "Teams" like its some crappy football game lost when its really about money and not gamers who whine about paying enough to cover R&D. And for Nvidia the professional market pays the R&D and JHH still jacks up the prices on his slop consumer gaming die bins. Look at those AMD fanboy hayseed bumpkins cry over the miners driving up the prices on AMD's pig slop binned dies for the gamers, oh how those gaming Mooks cry and cry!

November 8, 2017 | 10:37 AM - Posted by AMDisBeyondConsumerNOW (not verified)

You do not even understand what an abusive Trust is and that's because of JohnGR failure to know the history of abusive Trusts wher one or more companies bind together to dominate a market illegially in violation of the anti-trusts lows that have been around for over a century.

You need to read up on the history of the Standard Oil of Trust of Ohio and see whay the anti-trusts laws wher created in the forst place to combat Crony Capitalism's negative affect on free market Capitalism.

Fair market competition must be enforced by regulation/laws the anti-trust laws that where created to stop Crony Capitalism from destroying the free market system.

The High Technology markets today are even more corrupt that the Oil markets where before the anti-trusts laws where enacted by congress to save the free markets from Crony Capitalism!

AMD and Intel getting too close together to dominate any Laptop market for CPUs/Graphics is the very dictionary definition of a Trust that can not be allowed to happen. This current arrangement is not a trust it's a similar arrangement to what AMD has with its AIB PCIe card based partners. But because AMD and Intel represent 99% of the x86 based CPU market that Nvidia can not access without an x86 license if AMD and Intel go to far with their partnershup into abusive Trust territory the the US justice Department will by law be required to step in.

AMD and Intel can not collude togather to doninatle the laptop CPU/Graphics market to Nvidia's total and unfair disatvantage without Nvidia having access to an x86 ISA license to offer competition to this new MCM/EMIB based market. If Nvidia can get a Power9 license and creats competition that way fine but that's going to be too costly even for Nvidia owing to the software ecosystem not being there to support Power9 processots on any laptops/PCs.

So IF any AMD/Intel joint venture(TRUST) gets to mush market domination because if their x86 licenses possesion advantage and they both use AMD's GPUs in concert with their x86 license domination to corner the PC/Laptop CPU/Grapohics market that's an abusive TRUST, and one solution could be that Intel will be forced to also install Nvidia GPUs/HBM2 on those EIMB/MCM type arrangements. And Nvidia has the rights to sue and force Intel to do so now and not later to begin offering the same arrangement with Nvidia's GPUs/HBM2!

November 8, 2017 | 10:46 AM - Posted by AMDisBeyondConsumerNOW (not verified)

Please remove this Post, damn my Brain is getting worse at spelling! My Brain's wiring is borked from day one!

November 8, 2017 | 01:23 PM - Posted by JohnGR

It's not just the spelling.

November 8, 2017 | 07:16 PM - Posted by AnnoyedTownMouse (not verified)

It's not just your monobrow and road dragging knuckles, Elmer Hayseed fanboy!

November 8, 2017 | 10:44 AM - Posted by AMDisBeyondConsumerNOW (not verified)

******Reposted to fix too many spelling errors!**********
******Reposted to fix too many spelling errors!**********
******Reposted to fix too many spelling errors!**********

You do not even understand what an abusive Trust is and that's because of JohnGR's failure to know the history of abusive Trusts where one or more companies bind together to dominate a market illegally in violation of the anti-trusts lows that have been around for over a century.

You need to read up on the history of the Standard Oil of Trust of Ohio and see why the anti-trusts laws were created in the first place to combat Crony Capitalism's negative affect on free market Capitalism.

Fair market competition must be enforced by regulation/laws the anti-trust laws that where created to stop Crony Capitalism from destroying the free market system.

The High Technology markets today are even more corrupt that the Oil markets where before the anti-trusts laws where enacted by congress to save the free markets from Crony Capitalism!

AMD and Intel getting too close together to dominate any Laptop market for CPUs/Graphics is the very dictionary definition of a Trust that can not be allowed to happen. This current arrangement is not a trust it's a similar arrangement to what AMD has with its AIB PCIe card based partners. But because AMD and Intel represent 99% of the x86 based CPU market that Nvidia can not access without an x86 license if AMD and Intel go to far with their partnership into abusive Trust territory the the US justice Department will by law be required to step in.

AMD and Intel can not collude together to dominate the laptop CPU/Graphics market to Nvidia's total and unfair disadvantage without Nvidia having access to an x86 ISA license to offer competition to this new MCM/EMIB based market. If Nvidia can get a Power9 license and create competition that way fine but that's going to be too costly even for Nvidia owing to the software ecosystem not being there to support Power9 processors on any laptops/PCs.

So IF any AMD/Intel joint venture(TRUST) gets to mush market domination because if their x86 licenses possession advantage and they both use AMD's GPUs in concert with their x86 license domination to corner the PC/Laptop CPU/Graphics market that's an abusive TRUST, and one solution could be that Intel will be forced to also install Nvidia GPUs/HBM2 on those EIMB/MCM type arrangements. And Nvidia has the rights to sue and force Intel to do so now and not later to begin offering the same arrangement with Nvidia's GPUs/HBM2!

November 8, 2017 | 01:24 PM - Posted by JohnGR

The failure here is that you write too much and say too little because you are not objective.

Nothing more.

November 8, 2017 | 07:21 PM - Posted by AnnoyedTownMouse (not verified)

Monopoly apologist ye are, and as dangerous as ISIS in your middle ages mindset where cronys are held above the law!

Troll you later Tater Joe!

November 9, 2017 | 04:39 AM - Posted by JohnGR

LOL. Brain meltdown.

November 8, 2017 | 01:31 AM - Posted by James

I know people who bought the last generation excavator 6 and 8 core parts, and they have been perfectly happy with them. They mostly play WoW though, which doesn’t require a huge amount of processing power. Enthusiast often make a big deal about what are in reality small differences or just plain not important. CPUs have been fast enough for every day usage since about 2006. I was still using an old core 2 duo laptop up until recently. With an SSD it was perfectly usable. As long as you have hardware video acceleration, almost any modern processor will perform acceptably for general pc usage. The recent mobile parts review was a bit ridiculous with enthusiast complaining about an ultra thin device not being able to do video encoding very well. Is that what you buy a thin and light notebook for?

Enthusiast often swing these things in favor of performance over cost, which is often not the best for the person actually purchasing the device. Going intel over AMD excavator parts would have been significantly more expensive and completely unnecessary. With Ryzen, there is little to no performance penalty going with AMD; in many cases the cheaper AMD part will perform better due to more cores. I would have trouble recommending Intel to anyone right now. For the Ryzen mobile parts, they may outperform the Intel parts. It is looking like they can run at higher base clocks than the Intel parts. Then there is still a bunch of fan boys tying to paint AMD as the low end part, even though it is a 4 core/8 thread part. How is a 4 core/8 thread mobile part low end?

November 6, 2017 | 02:46 PM - Posted by DIMITRI JOSEPH

This is exciting news.

Thought all the rumors were false, but guess there were some truth to them.

But is this more of an Intel/AMD deal, or an Intel/Radeon Technologies Group deal? (I understand that at the end of the day its still AMD)

November 6, 2017 | 11:15 PM - Posted by SuchConfusionOutThere (not verified)

Apple's probably behind this and this will be only an option for Apple's customers in their iLappy hardware.

November 9, 2017 | 06:18 AM - Posted by Coppermine (not verified)

Again and again you were wrong. It started from tensions between Raja Koduri and Lisa Su as far back as 2016. And Kyle Bennet of [H] was spot on https://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_... quote "Koduri was able to wrestle control of the graphics division away during AMD’s last leadership transition after threatening to leave the ship and take a role at Intel, something he's not shy about telling his AMD colleagues". More quotes "Where the plot thickens is when you look at the Koduri’s unwavering ambition. Koduri’s ultimate goal is to separate the Radeon Technologies Group from its corporate parent at all costs with the delusion that RTG will be more competitive with NVIDIA and become a possible acquisition target for Koduri and his band of mutineers to cash in when it's sold. While Koduri is known to have a strong desire to do this by forging a new relationship with Apple on custom parts (no surprise there) for Macbooks, the real focus is on trying to become the GPU technology supplier of choice to none other than Intel."

November 6, 2017 | 02:57 PM - Posted by JohnGR

Intel gets security for the x86 platform.

Nvidia was once 1/10 of Intel, maybe less, if we consider their stock market valuation. Today Nvidia is about 60% of Intel. 3 years ago this news would have send Nvidia's share down without a parachute. Today Nvidia's share doesn't even fill it. It's going up.

If Nvidia gains a huge advantage over AMD, with it's GPUs being for example two times faster, 3 times faster compared to AMD's, then not Intel, but Nvidia decides where the enthusiast and gamer goes.

With modern ARM SOCs getting ready to run Windows and the only parts of the PC market that sees growth being the enthusiast and gaming market, x86 is vulnerable and AMD it's only shield.

Switch proved that you don't need a strong system, Playstation and XBOX that you don't need a strong CPU either. You only need a good GPU and a good API. Apple proved that you can have ARM cores much more powerful than what a standard ARM design offers.

If Nvidia gains a huge advantage over AMD, it can easily announce in a few years two ARM platforms. One running Windows with DX12 for ARM and one running Android with Vulkan. It can move all it's high end GPUs to those platforms and force the x86 PCs, as we know them today, to be limited in servers, and office PC solutions. It will be the beginning of the end for the x86 platform.

November 6, 2017 | 09:40 PM - Posted by OutWithTheDogFoodIntelGraphics (not verified)

Nvidia can license some Power9 core IP from OpenPower and spin up its own Power9 based SOCs with Nvidia Graphics on an MCM. But Actually Nvidia's Denver Cores were actually a little wider order superscalar than even Apple's A7 and the Nvidia Denver(Gen 1) could issue 7+ micro-ops per cycle compared to the Apple A7 Cyclone 6 M-Ops/cycle issue rate.

The Power8 core design is an 8 decode and 10 instruction issue design with 16 execution units and SMT8(8 processor threads per core) very wide order superscalar design and the Power9 comes in a SMT4 variant that can offer upto 24 cores per socket, so maybe Nvidia can license from OpenPower and get a 4 core/16 thread Power9 desktop variant with Nvidia graphics. Hell Nvidia Denver is a RISC ISA(ARMv8A ISA Running) custom design and IBM's/OpenPower's are RISC ISA based power7/power8/power9 designs. Nvidia's Denver2 design may be just a little bit more powerful but that's only used in automotive systems currently. There is nothing stopping Nvidia from beefing up a custom ARMv8A ISA running design that's a little wider order superscalar than any x86 ISA based CPU micro-arch just look that the power8 core design that's used in the server room.

AMD's Vega 10/Vega micro-arch designs only use more power because of all the extra shader cores that the Vega 10 based design has standard for the compute market mostly. So AMD can trim back the Shader count and Up the ROP to shader ratio similar to Nvidia's ROP to ratios are and ROPs are what allow Nvidia to fling the frames per second out there whatever quality they are. So AMD's Vega 56 is gstting there with the lower shader counts but AMD needs to up the ROP counts and trim back on Vega 56's excess shader count and that would probably be enough to reach Nvidia's power usage metrics. Vega 10 is a very power efficient design for bsing so shader heavy compared to Nvidia's consumer designs bit those Vega 10 dies used in the Radeon Pro WX9100 and Radeon Instinct MI25 SKUs are actually very efficient clocked lower and undervolted for professional compute and have great performance/watt metrics for the professional markets that can afford to properly pay for the better drivers and the top binned Vega 10 dies that are not used in any consumer SKUs.

November 7, 2017 | 03:39 AM - Posted by JohnGR

They don't need Power9. ARM will do just fine and Vulkan and Windows on ARM, can make their lives so much easier in promoting an ARM based gaming platform. And as you say they already have experience with Denver.

The problem with Vega is financial. AMD doesn't have the money to make major changes to their GPUs. Tonga was a more efficient Tahiti that could offer with a 256 data bus what Tahiti was offering with a 384 data bus, thanks to data compression.
Then Fiji was a double Tonga design, with HBM memory controller. Vega on the other hand was just a die shrink of Fiji, with extra compute power.
If you see all AMD's products in the last 5 years, with the exception of Ryzen, they all share the same common characteristics. They are all based on previous designs with some extra tweaks and they offer improvements only in one area. Either performance, or efficiency. A newer more power efficient product, will offer the same performance. A newer faster product, will not improve in power efficiency. On the other hand Nvidia did had the money to create new architectures that improved in both areas and the same can be said for Intel, at least until Skylake, because after Skylake they just increase frequency and recently number of cores.

November 7, 2017 | 11:40 AM - Posted by OhSuchDaftynessAboundsInSome (not verified)

"Windows on ARM", Ha ha speed spaffer, ha ha!

Nvidia is free to license from OpenPower as is AMD and Intel. OpenPower is what Google is looking at in addition to Epyc or Xeon.

I'll bet that Jim Keller's/Team's K12 is just as wide order superscalar of a design as Zen and K12 may even have SMT capabilities that Devner(V1, V2) lack, ditto for even Apple's A series CPU/SOCs lacking any SMT capabilities!

Power8 is a beast of an extra wide order superscalar RISC ISA running design and maybe Nvidia can make use of that at a lower cost than power9. It all depends on the development cost and the certification time frames of clean sheet designs as opposed to licensing some already developed/certified IP for Nvidia, that's up to JHH and Nvidia's BOD.

Hell Nvidia makes use of the RISC-V open instruction set architecture(ISA) for its Falcon = FAst Logic CONtroller IP used on many of Nvidia's products.

November 7, 2017 | 05:08 PM - Posted by JohnGR

You laugh now, but if you see a future laptop with a Tegra in it you will rush to buy it.

Everyone is free to license anything, and then throw a few billion dollars hoping that there will be interest. As for Google. Google looks everything.

November 8, 2017 | 01:01 PM - Posted by AMDisBeyondConsumerNOW (not verified)

I'll buy a Tegra with Denver cores provided that the Denver cores are close to what AMD's K12 custom Design is. The power8 RISC ISA based micro-arch can have its SMT8(8 processor threads per core) units paired down to SMT2 and still keep that power8 core's 8 instruction decoders feeding into 16 execution ports/units per core and make for a fine Desktop SKU.

Nvidia's CPU engineers are more than quilifed to make the Denver cores a little wider order superscalar than say the Intel/AMD x86 designs and the power8 core is the widest order superscalar of the bunch that I have listed. Nvidia needs to add SMT support on its Denver cores as even the first generation Denver cores are wider order superscalar than even Apple's A series cores.

The power8 core design using SMT4 is the optimal design for those SMT units to keep those 16 execution units on the power8 core fully utilized and that's part of the reason that IBM offers a Power9(SMT4) variant at 24 cores per socket and also offers a Power9(SMT8) variant at 12 cores per socket.

I'd really Like Nvidia to create a custom Laptop/mobile Core using the RISC-V ISA as RISC-V is an open ISA that can be used by all, no licensing fees required. Nvidia already uses RISC-V on their Falcon = FAst Logic CONtroller General purpose embedded processor!

November 8, 2017 | 01:29 PM - Posted by JohnGR

So, if you agree that Nvidia can create Laptop/mobile core, where is the problem with Intel's and AMD's collaboration? I know I know. 99% of the x86 market. Have you posted huge posts like this about VIA getting a bigger piece of the market? I bet you didn't. But for Nvidia. You lose all day online writing huge posts.

November 8, 2017 | 07:44 PM - Posted by AnnoyedTownMouse (not verified)

You love your Team and its Flag more than your own mama!
Your constant need for E-Peen enhancement via an ego driven need to have the hardware will the highest FPS metrics above an beyond what is really necessary has you twisted in knots mentally over your Team's performance.

Nvidia may have the money(Maybe) but does not have the time required! As that process to build a software/OS ecosystem around any CPU ISA takes Decades compared to the 4 or 5 years it takes to design a CPU micro-arch that runs a new ISA.

A power9/Power ISA needs a desktop/laptop SKU developed with maybe 4 core/4 threads per core power9 CPU micro-arch based variant and that may not take too much time to get out and certify the design. But getting those software/OS ecosystem ducks in order in time would put Nvidia at too much disadvantage in the laptop market to compete with that Intel/AMD partnership.

Intel's and AMD's collaboration can quickly become Intel's and AMD's collusion if Nvidia's GPU parts are locked out of that Intel MCM/EMIB nano-motherboard laptop x86 ISA/Graphics(Radeon) market that just started up with this new deal.

November 9, 2017 | 04:40 AM - Posted by JohnGR

LOL. Funny stuff.

November 12, 2017 | 09:00 PM - Posted by IntelGetsRajaAcceloratorsForProMarkets (not verified)

X86 is an ugly ISA that's a power hog just look ath the smart phone market there Elmer. Trim that monobrow!

November 6, 2017 | 09:51 PM - Posted by Anonymous5667899554444 (not verified)

Ok that's just dumb people have been screaming end of x86 for years but it's not going to Happen. They can't even utilize more than a few threads on modern software you think they will ditch x86? Lol what a laugh.

November 7, 2017 | 04:39 AM - Posted by JohnGR

No, it's a possibility and thanks to some things that happened these last years, it is turning from a "Lol what a laugh" to a possible threat to Intel and x86.

Intel has lost most of it's huge advantage it had on manufacturing.
Intel failed to enter the smartphone/tablet market.
ARM cores are becoming more powerful, or should I say powerful enough?
Vulkan runs on Android, making probably easier for developers to create top games for it.
Apple, Nvidia, Qualcomm and others, all try to create even faster ARM cores than the standard cores ARM designs.
Windows are getting ready to run on systems with Qualcomm processors.
Apple is rumored to consider changing from x86 to ARM for it's future Macs, some of them at least, and probably this collaboration between Intel and AMD was also done to delay that decision.
Probably there are more things to write and if there are not today, they might be tomorrow.

You can say that all these look like a big laugh, or speculation, or tin foil hat theories. But what is a fact is that none of these arguments where valid 5-6 years ago.

November 7, 2017 | 11:52 AM - Posted by OhSuchDaftynessAboundsInSome (not verified)

The x86 ISA a Frankenstein of an ISA! So very ugly and inelegant and legacy bloat laden of an ISA with so much ossified cruft in the micro-code form clogging up the works and prodigiously producing the Errata for both AMD and Intel.

November 6, 2017 | 03:30 PM - Posted by Can'tlogin (not verified)

Egon Spengler: 40 years of darkness! Earthquakes, volcanoes!

Winston Zeddemore: The dead rising from the grave!

Peter Venkman: Human sacrifice, dogs and cats living together, mass hysteria!

November 6, 2017 | 03:34 PM - Posted by Power (not verified)

Did Intel just announced that Nvidia is leaving x86 market?

November 6, 2017 | 04:41 PM - Posted by NiceSemiCustomForTheClientsLaptopSKUsONLY (not verified)

Well the EMIB/MCM becomes the new laptop mini-motherboard of sorts with Intel linking up its CPU/Cores die with AMD's GPU die and the HBM2 stack/s using whatever HyperTransport/Intel implementation of that HyperTransport standard that is most compatable with the GPU IP in AMD's Radeon GPUs. And the Infinity Fabric is HyperTransport+ according to Charlie over at S/A. HyperTransport is an open standard that is a subset of AMD's Infinity Fabric according to SemiAccurate so that should not be so hard for Intel to make use of. EMIB (Embedded Multi-die Interconnect Bridge) is a more affordable solution that uses small Interposer Bridge chips to splice everything together using small interposers. So Intel can for its semi-custom design make use of AMD's Radeon/Polaris, or newer GPU IP, with AMD just providing a GPU die to be integrated on that Multi-chip package along with some HBM2.

This whole project has Apple as a client written all over it with Intel providing Apple with a new space/power saving x86 CPU/Radeon custom GPU system on an EMIB module. So AMD will fab the semi-custom design most likely at GF/TSMC and that GPU die will be sent to Intel for packaging with Apple most likely the intended client with the deep pockets to fund the entire project, least Apple go with an entirely AMD design and Intel be completely cut out of the action.

Intel and AMD both supply Apple with CPUs and GPUs respectively so this is not much more than an system on a module for a specific client that wants x86 from Intel and the GPU portion from AMD. Apple told Intel what its needs where and Intel knew that AMD's semi-custom division had the methods for such and undertaking owing to AMD's console design wins with XBox and PS4.

This will most likely not be using all of AMD's Infinity Fabric IP, and AMD is still developing a workstation grade APU on an Interposer design that will be priced for the professional markets and probably make use of an active interposer with the infinity fabric traces/circuits etched on a full silicon interposer. So there will be a Zen/Vega/HBM2 workstation APU on an interposer for the professional markets that is of an all AMD CPU/GPU IP design and that will be on an entire different level than this Intel custom design and priced accordingly for the Pro market by AMD.

But most likely/legally Intel and AMD have to be doing this for a client and this SKU may not find its way into the larger laptop market outside the end customer who commissioned the design from Intel/AMD's Semi-Custom. Both Intel and AMD can not get to closely involved on any x86/Redeon projects for the overall PC/Laptop market with AMD/Intel acting as a combined entity and not run afoul of the Justice Department and its anti-Trust division that is there to enforce the anittrust laws on the books for the past 100+ years.

Together AMD and Intel would represent about 99%(via is still out there) of the x86 based CPU/APU/SOC market and this semi-costom design for say Apple's exclusive use would not set off and bells/alarms at the US justice department. So do not look for any wide availabiliy of any Intel SKU for the larger Laptop market as that would get both Intel and AMD looking like a classic Trust arrangement that would not be allowed. Illegal trusts can be single companies or made up of more that one company(Trusts) and two big x86 players coming together for a wider market combined design would be an example of an illegal Trust sort of arrangement if any Intel x86/CPU and AMD/GPU design where made for the larger market place with both AMD and Intel collectively in control over too large an x86/Radeon APU/SOC market on PCs/laptops.

The market still needs both Intel and AMD in direct competition with eachother and not acting like a single Trust to the regulators so this custom Intel/Amd APU/SOC like system on EMIB module is for a third party client and will not be seen outside of that clients' hardware.

So Now is the Time fore Nvidia to break some news that it has Licenesd a Power9/SMT4 semi-custom variant with Nvidia getting a more powerful Power9/micro-arch based design to compete head to head with both AMD and Intel for the PC/Laptop APU/SOC market business. And Power9, or even Power8, would be up to the task of competing with any x86 based PC/Laptop x86 based SKUs.

"HyperTransport Consortium"

https://en.wikipedia.org/wiki/HyperTransport_Consortium

November 6, 2017 | 06:25 PM - Posted by Coppermine (not verified)

HyperTransport is outdated and pretty much dead in the water. Nobody else nor even the members in the consortium no longer use HyperTransport anymore (despite numerous revisions and additions such as Torrenza) except AMD themselves on their previous generation chips. Nvidia has their own NVLink, IBM uses their own CAPI, etc. Meanwhile the rest of the industry is using the standardized PCI Express (which is also a point-to-point interconnect).

November 6, 2017 | 07:16 PM - Posted by DRAMsAsWriteCacheIsNotSoSafeWithoutBatteryBackup (not verified)

IBM uses OpenCAPI and AMD along with IBM(CAPI's creator)/Others is a founding member of the OpenCAPI consortium. And read this article from S/A(1). And Charlie(S/A) states that at the AMD briefing:

"On the surface it sounds like AMD has a new fabric to replace Hypertransport but that isn’t quite accurate. Infinity Fabric is not a single thing, it is a collection of busses, protocols, controllers, and all the rest of the bits. Infinity Fabric (IF) is based on Coherent Hypertransport “plus enhancements”, at a briefing one engineer referred to it as Hypertransport+ more than once. Think of CHT+ as the protocol that IF talks as a start."(1)

So Even if Intel is not using HyperTransport then maybe it's using stright up PCIe, But there has to be some part HyperTransport there if AMD's GPU IP is involved even if it's a subset of HyperTransport and AMD's Infinity Fabric is a superset of HyperTransport. Standardized PCI Express is more ubiquitous but some standards like CCIX take that PCIe standard and strips out some of the PCIe Protocol parts and and keep other parts and makes use of the PCIe infrastructure using CCIX signialing/protcol instead of purely PCIe.

(1)

"AMD Infinity Fabric underpins everything they will make"

https://semiaccurate.com/2017/01/19/amd-infinity-fabric-underpins-everyt...

November 7, 2017 | 01:05 AM - Posted by James

If it is Polaris based, it is almost certainly just pci-express. All of the interesting interconnect stuff is in Vega. They don’t need any of that stuff connecting to an intel cpu.

November 7, 2017 | 07:41 AM - Posted by Coppermine (not verified)

And that's because AMD trying to get into the supercomputing scene on the IBM side. But the first systems with OpenCAPI from IBM are all using Nvidia GPUs unfortunately. Just like what happened to AMD in OpenCompute consortium as they bid to get Facebook business, but that didn't turn out well with Intel grabbing all of it.

Like I said, only AMD alone still uses HyperTransport. And if this Infinity Fabric is their new HyperTransport replacement then again only AMD uses it.

Everyone else have their own interconnects. In fact HyperTransport was derived from Intel's patents acquired from DEC Alpha buyout. That's why Intel has their own version called QPI, first seen in Nehalem generation.

If you look at most supercomputers and server racks, they are mainly all based on PCI Express on board for connection with other chips like gigabit Ethernet controllers. That's why Intel's Xeons and AMD's new Epyc has lots of PCI Express bus lanes.

November 7, 2017 | 09:35 AM - Posted by BustEmTrustsTeddy (not verified)

Do NOT ever think that IBM likes Nvidia over AMD/others, as IBM was the one that forced Intel to cross-license the x86 16/32 bit ISA to AMD and others in order for IBM to have a competitively priced second/third source of x86 parts for the IBM PC that started this whole PC market to begin with.

IBM would rather there be an OpenCAPI industry based CAPI(1) standard that all are free to include in their acceleorator products GPU/Other accelerator products. NVLink only works with Nvidia's GPUs and OpenCAPI is IBM's way of getting more second/third/more sorces of GPU/Other accelorator products into use with the power9 processors licensed via OpenPower. So IBM/Third party OpenPower licensees can use OpenCAPI and make use of AMD's GPUs and Other's Accelerator products with Power9 CPUs.

AMD is a founding member of OpenCAPI while Nvidia(hedging it's bets) is just an assoicate member of OpenCAPI, so there will be OpenPower power9 licensees makeing use of AMD's GPU accelertor products because no PC/System producer wants to be tied to a simgle parts supplier and CPUs and GPU's/other processore are just parts in the end that need to be competitively sourced supplies of parts to the PC/Server/Embedded systems OEMs.

From the wikipedia entry on CAPI:

"OpenCAPI 3[edit]

OpenCAPI, formerly New CAPI or CAPI 3.0, is not layered on top of PCIe and will therefore not use PCIe slots. In IBM's CPU POWER9 it will use the Bluelink 25G I/O facility that it shares with NVLink 2.0, peaking at 50 GB/s.[8] OpenCAPI doesn't need the PSL unit (required for CAPI 1 and 2) in the accelerator, as it's not layered on top of PCIe but uses its own transaction protocol.[9]" (1)

"Coherent Accelerator Processor Interface"

https://en.wikipedia.org/wiki/Coherent_Accelerator_Processor_Interface

November 7, 2017 | 03:41 PM - Posted by Coppermine (not verified)

OpenCAPI is IBM's initiative. And do read directly from OpenCAPI website itself http://opencapi.org/2016/10/tech-leaders-unite-to-enable-new-cloud-datac... Some quotes from that webpage "Technology leaders AMD, Dell EMC, Google, Hewlett Packard Enterprise, IBM, Mellanox Technologies, Micron, NVIDIA and Xilinx today announced a new, open specification that can increase datacenter server performance by up to 10x, enabling corporate and cloud data centers to speed up big data, machine learning, analytics, and other emerging workloads". Also notice OpenCAPI is closely tied to IBM's hardware.

They are all founders, includes Nvidia also. Do note that Nvidia is also on the board of directors http://opencapi.org/about/board-of-directors/ Most of the current and future hardware released are using IBM's Power8/9 and Nvidia's GPUs. You can read this https://www.itjungle.com/2017/10/02/talking-power9-ibm-fellow-brad-mccre... the upcoming Summit and Sierra supercomputers will be using IBM and Nvidia

But where does AMD fit into all this? More like in the background than a bit part player actually. At least in OpenCompute, AMD had hardware to show but not here. Since its tied to IBM's hardware, thus IBM can dictates everything here including decisions on choosing GPU accelerators (which is still Nvidia).

November 7, 2017 | 04:22 PM - Posted by DemJokezOnYouzKnuckleheadz (not verified)

Nvidia is not a founding member of OpenCAPI Nvidia is only there in case its NVLink Vendor lock-in does a G-sync and And Gets OpenCAPI-ed FreeSync(VESA DP adaptive sync) style putting an end Nvidia's Vendor Lock-In NVLink ambitions. BOD membership and Founding Member are not always the same, BOD members can come from outside entities also!

That's Just Nvidia Hedging its bets with Nvidia in with OpenCAPI. That Government supercomputer contract is somewhat of a custom affar and Others are using Intel's offerings. AMD may get some wins there also with Epyc/Radeon GPUs and look at AMD's/Partner's Project 47 supercomputer in a single cabinet, That's 20, 32 core Epyc CPUs and 80 Radeon Instinct MI25(Vega 10 die) based GPU AI/Compute accelerators offeing up one petaflop in that single cabinet.

November 7, 2017 | 06:19 PM - Posted by Coppermine (not verified)

Are you plain ignorant? Nvidia was also a founder, as it was already present from the time of its inception. The true founder of OpenCAPI, as always, was IBM who has the presidency of the board. And here you go off the beaten path again. What does G-Sync and FreeSync have to do with GPGPU accelerator tasks in the server and supercomputer space? Those are related to display monitors, and not to GPU or CPU interconnection technology. Furthermore that NVLink feature is inevitable because the hardware is dictated by IBM (due to the close partnership with Nvidia as well) as mentioned earlier.

For your information, AMD's Project 47 supercomputer is 1 petaflop of single precision or FP32. Somehow that still that was not convincing enough (to potential supercomputer customers), because supercomputer Linpack performance metric is based on double precision or FP64. Unfortunately those Radeons does not have much double precision performance https://instinct.radeon.com/en/product/mi/radeon-instinct-mi25/ around 768GFLOPS only per card. And we have yet to hear top supercomputer manufacturers like Cray (who used to be AMD's customer long ago) announced any new system using either AMD's Epyc or Radeon GPUs.

November 8, 2017 | 12:29 PM - Posted by AMDisBeyondConsumerNOW (not verified)

1 petaflop stated without any quilification(FP16, DP/etc.) always defaults to FP32 and yes that DP is what the SC industry uses. But the MI25 is for infrencing workloads mostly and Vega 20 will have 1/2 DP to 1 SP(FP32) ratio.
The MI25's FP 16 rate is 2 Petaflops and AI/Infrencing makes use of 16 and 8 bit mostly for AI/infrencing workloads.

So Vega 20 is incoming and so is Vega 11 and AMD's semi-custom division can do Government Exascale contracts also so go and read this PDF.

(1)

"Design and Analysis of an APU for Exascale Computing"

http://www.computermachines.org/joe/publications/pdfs/hpca2017_exascale_...

November 8, 2017 | 03:16 PM - Posted by Coppermine (not verified)

Half precision FP16 is fine for games (heck GPUs inside ARM SoCs uses FP16 mostly), however no one uses FP16 in supercomputers for scientific work, simulations and calculations because just not enough precision. Single precision or FP32 isn't used much either (except in specific situations or tasks), again due to not good enough precision. Thats why double precision or FP64 is the gold standard in supercomputers.

Look at governmental and institutional supercomputer projects by IBM and Cray, those are mostly Nvidia cards because of double precision or FP64 prowess.

Weak response because that is just a research paper. Meanwhile still no high performance APU from AMD forthcoming. If AMD had wanted one then they would have done so much earlier (before Zen).

November 8, 2017 | 07:57 PM - Posted by AnnoyedTownMouse (not verified)

Games be damed that MI25 is for AI/Infrencing workloads and not gaming and the Vega 10 die variant is for the compute/AI markets first and formost with the runt Vega 10 dies slop bin for the gaming pigs that wallow in the feces filled muck and eat that mixed with the lower binned Vega 10 die slop!

JHH over at Nvidia mixes in a little extra ROPs in Nvidia's lower binned GP104 slop and that flings those filthy frames out there at higher ego boosting FPS metrics, quality be damed, to feed his high paying hogs their fill of that poo/slop also. JHH's Quadro markups are where his real bread is made.

November 9, 2017 | 06:01 AM - Posted by Coppermine (not verified)

Yet I see more rants, yadda yadda. Meanwhile https://www.top500.org/news/server-vendors-jump-on-volta-gpu-bandwagon/ quote "The world’s largest server OEMs announced they will be soon be shipping systems equipped with NVIDIA’s latest Volta-generation V100 GPU accelerator. Included in this group are Hewlett Packard Enterprise (HPE), Dell EMC, IBM, Supermicro, Lenovo, Huawei, and Inspur, all of which took the opportunity to reveal their Volta-powered servers at this week’s GPU Technology Conference in China".

This is just a sample of why Nvidia became a very successful and profitable company (also reflected in its share price breaking the $200 barrier).

November 12, 2017 | 09:04 PM - Posted by IntelGetsRajaAcceloratorsForProMarkets (not verified)

Pro markets is where the money is and gaming Mooks don't pay so JHH, Lisa, Brian, all go for CPU with GPU compute and the gamers get the slop dies. Eat that slop gaming pigs becuase the filet mignon is the professionals who will pay the proper markups!

November 7, 2017 | 09:59 AM - Posted by BustEmTrustsTeddy (not verified)

There is more built into AMD's CPU and GPU products fabric controllers that support other Open Standards Protocols like OpenCAPI and if you look at some of the Block Diagrams of the AMD CPU's and GPU's there is More than Infinity Fabric IP functional blocks included and that more includes support the OpenCAPI and other standards. So those Zen/Zeppelin dies that are binned down into Summit Ridge platform dies(For the consumer market) have a lot of IP functional blocks that already support many of the Open Standards Protocols and assoicated PHY layers than only PCIe or Infinity Fabric.

You can be assured that any Open Standard Fabric protocol standards body that AMD is a member will already have that standard body's Open IP included on the Zeppelin Die and the Vega 10 Die that are currently on the market. So OpenCAPI 3.0 is most likely already there and OpenCAPI 4.0 is supposed to be avaliable for the Power9's as well as PCIe 4.0(The Power9 platform will be the first to support PCIe 4.0).

November 7, 2017 | 03:20 PM - Posted by Coppermine (not verified)

This OpenCAPI is still a small niche (created by IBM) in the sea of other interconnect solutions (from Cray, Intel, Mellanox, etc). Likewise with OpenPower (also created by IBM) as an alternative to x86 and ARM, but that is still a small niche (which yet to even take any significant share). As mentioned above, this is the "IBM side" of the market which accounts for less than 1% of the total server and datacenter market share. IBM's main targets here are those big contracts for custom supercomputers and mainframes.

Nonetheless at the end of the day this OpenCAPI is will still be mainly dominated by IBM and Nvidia mostly, even with AMD on the board of members. That's because most of the current and future hardware released are using IBM's Power8/9 and Nvidia's GPUs. You can read this https://www.itjungle.com/2017/10/02/talking-power9-ibm-fellow-brad-mccre... the upcoming Summit and Sierra supercomputers will be using IBM and Nvidia chips.

And where does AMD fit into all this? More like in the background than a bit part player. In OpenCompute, at least AMD had hardware to present (and demonstrate) to potential clients but worse still in OpenCAPI space there is none.

November 7, 2017 | 04:23 PM - Posted by DemJokezOnYouzKnuckleheadz (not verified)

AMD's server products will support OpenCAPI and AMD's GPUs interfaced to Power9's!

November 7, 2017 | 05:49 PM - Posted by Coppermine (not verified)

And none of them is going to get into those big projects in IBM's grasp.

November 8, 2017 | 11:29 AM - Posted by AMDisBeyondConsumerNOW (not verified)

IBM Licenses the Power8/9 IP Via its OpenPower consortium in a similar manner to the way ARM holdings offers up its IP/ISAs for licensing. Google will be using its own Power9s obtained by licening from OpenPower so that's a bigger potential market right there than any US government supercomputer contracts. IBM Started OpenPower just to get its Power9 designs out into a larger market of users the same way Arm Holdings does and DITTO for The IBM started OpenCAPI consortium so IBM/OpenPower licensees would not suffer any GPU/Accelerator Vendor lock-in.

Those Government supercomputer procurement contracts were done via OpenPower and IBM/Nvidia on a limited contract basis only and IBM is free do do the same with AMD or even with others and use OpenCAPI and not NVLink!

Coppermine, what neolithic cave dwelling did you emerge out from just to ride the turnip cart, that you fell out of, to get to a library branch with internet connected PCs with which to spraff your FUD!

This ain't no Football Game there Joe SixPack, it's business and IBM never has a single source of parts supplies ever in BIG Blue's history of mainframes or the IBM PC that started the whole Personal Computer market. Back then PCs that were not made BY IBM where called IBM PC clones. IBM is no longer a PC powerhouse and even it's mainframe business is limited and IBM is not even in the CPU fabrication business or hardware business for that matter. IBM is a fabless IP company and IBM's reaearch division is tops in the world but the majority of IBM's non IP revenues come from software/consulting services and its mainframe Z series management contracts.

P.S. both OpenCAPI and NVlink run under IBM's BlueLink PHY/Link layer IP on IBM's systems and BlueLink is just a codename for CAPI/OpenCAPI.

"IBM is running the NVLink 2.0 protocol over its BlueLink ports" (1)

(1)

"Opening Up The Server Bus For Coherent Acceleration"

https://www.nextplatform.com/2016/10/17/opening-server-bus-coherent-acce...

November 8, 2017 | 03:25 PM - Posted by Coppermine (not verified)

And those government and institution contracts goes thru IBM, not your precious AMD. Furthermore when the hardware choice is already dictated by IBM for those contracts (signed and sealed), there is little or nothing AMD can do. Like I said earlier, AMD is more of a background player than a bit part one.

November 12, 2017 | 09:15 PM - Posted by IntelGetsRajaAcceloratorsForProMarkets (not verified)

Ha ha ha AMD is just another dirty parts supplier same for Nvidia and Intel. IBM licenses its CPU designs via OpenPower and IBM, Nvidia and got the supercomputer contracts via OpenPower and that's IBM, Nvidia and others all in with OpenPower. IBM is not in the hardware market it's in the services business. IBM sells hardware to make the majority of its money from software services so hardware is just to IBM as a means to its services end of the business. IBM's research division its tops in the world for all compute but IBM makes more on the services and OS/Software support side.

And gaming gits are just there to eat the slop dies and allow the filthy parts suppliers to recoupe some of their fab costs. The real businsee is for the professional compute/AI markets for CPU and GPUs with the consumer gaming market there to eat the slop dies and offset some expenses!

November 6, 2017 | 08:24 PM - Posted by stalin (not verified)

I agree, a new Mac Mini was promised recently, forgo the disk option, stick USB-C TB only and I bet you could get the FF down to original ipod or smaller.
But yes smells very Apple
(Think a Mac about the size of a PI in a case...)

November 6, 2017 | 08:42 PM - Posted by OutWithTheDogFoodIntelGraphics (not verified)

A Mac Mini with Radeon Graphics and TB3 support on the chipset from Intel and that CPU better be 4 cores and 8 threads because dual cores are not enough. Could there Be a Mac Mini server SKU once again because folks were wanting that for a good long while.

RIP 2014 Mac Mini Server( The Mini-Server's SOC parts were a little old by that time but still better than any Intel U series dual core i7/i5 crap), please come back from the dead Mac Mini Server!

November 6, 2017 | 05:27 PM - Posted by NiceSemiCustomForTheClientsLaptopSKUsONLY (not verified)

"To reiterate, this is Intel as a customer for AMD’s Semi-Custom group rather than a licensing agreement between the two companies. They are working hand in hand in developing this solution and then both profiting from it. AMD getting royalties from every Intel package sold that features this technology will have a very positive effect on earnings. Intel gets a cutting edge and competent graphics solution along with the improved software and driver support such a package includes."

No it's Intel/AMD providing a third party client with a custom x86(Intel)/Radeon(AMD) APU/SOC system on an EMIB module so Intel is not the direct client Intel is the primary subcontractor for that third party client and Intel will probably not be allowed to sell this design to the larger PC/Laptop market. Both Intel and AMD would have too much collective control over the x86 based PC/Laptop market if this design was for the larger market with AMD and Intel in an x86/Radeon Graphics Trust agreement that would set off alarms bells at the US Justice Department's AntiTrust division.

AMD and Intel need to be in direct competition with each other for APUs vs SOCs and AMD's grahics vs Intel's graphics for the wider market or the consumer will lose in the long run. Both AMD and Nvidia need to watch out becaues their collective GPU IP patents will not last forever and once Intel no longer has to license from Nvidia or AMD then Intel can go it alone and brush both AMD and Nvidia aside. AMD needs to make sure not to allow Intel to fab its custom GPU design AND needs to simply supply the die in the same manner that AMD does for an AIB partner with AMD in charge of its GPU IP.

If Apple is in fact the Client for this custom mash-up of AMD and Intel then Apple will write the drivers and use Apple's in house Metal Graphics API. I really want an all AMD version for the laptop market and AMD making full use of its Infinity Fabric IP and CPU/GPU IP.

If AMD and Intel even think that they can collectively come together and become and even bigger RedZilla mutent for any SKUs beyond this semi-custom design for the client's hardware usage only then the US Justic Department's Anti-trust division needs to be doing their job to force Intel and AMD to compete or strip them both of their x86 IP/GPU IP exclusive rights and let a larger market compete using the x86 ISA/and Radeon GPU IP.

There are some very broad GPU patents that both AMD and Nvidia control due to the USPTO's ineffectual granting of patents with prior art not properly vetted by the USPTO. And when those Unified Shader/Other patents expire both AMD and Nvidia will have problems from more than just each other. Intel can not be allowed to use this design in a larger market without AMD having the rights to a Zen/Radeon Vderson of the same/similar design and AMD will probably have an Interposer basd APU for the workstation market in 2018 and the consumer market also a little later, but those GPU patents are approaching their natural end in not to many more years and AMD can not be sharing any GPU IP with Intel, or even Apple above what is necessary in the software/drivers level sorts of IP abstraction that all GPU makers practice to protect their IP.

November 6, 2017 | 06:00 PM - Posted by James

Curious as to whether this design can show up elsewhere. The EMIB tech is quite similar to silicon interposers. The height differences seems like it will be quite small. You are essentially embedding an upside down chip in the package with the interconnect. Apple may care about that for their super thin x86 devices. For regular laptops, an APU plus a dedicated GPU would not be that large. Most of the space savings comes from using HBM, so if AMD or Nvidia makes an HBM mobile dedicated GPU, then those could be almost as small. If this is Polaris, then this does indicate that AMD has essentially made an HBM based Polaris design, but it is unclear if they can or will use it anywhere else. A Ryzen APU with a larger Vega GPU or a just Ryzen APU with HBM on an interposer would be better. For a laptop, I would rather have a Ryzen mobile part plus a dedicated Vega HBM GPU. This would be very compact with the GPU using HBM.

November 6, 2017 | 06:37 PM - Posted by SincereAnonymous (not verified)

what the fuck just happened

November 6, 2017 | 07:55 PM - Posted by BigMoneyAppleBarrelTeraBucks (not verified)

Apple told Intel that they did not want Intel's dogfood graphics and Apple said to Intel Get some Radeon from AMD's semi-custom division and use your EMIB (Embedded Multi-die Interconnect Bridge) IP to marry AMD's Radeon dies with your CPU die along with HBM2 for our Apple MacBook/Macbook Pro SKUs or we will go with AMD only for or laptops.

This development was started before Zen was to market so Apple went with Intel's CPUs and AMD's Radeon GPUs paired with HBM2, because Apple is rather obsessive with that thin and light laptop form factor stuff that Apple brought to market first long before there was any Intel Ultrabook(TM) marketing plan. So Apple was looking at those AMD GPUs with that HBM/HBM2 space saving/power saving HBM memory design and wanting a system on a module(EMIB or full fat Interposer) based, whatever was lower cost to implement.

So Apple has enough petty cash to purchase both AMD and Intel lock stock and barrel and Apple's money talks loudest. So Intel will dance and dance hand in hand with AMD's semi-custom division all under Apple's bidding to produce that custom Intel-CPU/AMD-Radeon-GPU mash-up for Apple's needs. Apple is so Flush with that dosh that Intel will dance cheek to cheek with AMD for an Apple only custom design.

Intel and AMD are Apple's Beotches and Apple will even have them sucking face while twerking. So Apple just bought out GOD and now iTat is sold in both heaven and hell and Apple looks to be the first trillion dollar iTAT selling megacorp. Apple is both a retailer and a hardware maker with that closed ecosystem top to bottom of the Apple OS/API software stack and iTAT App store.

November 7, 2017 | 01:15 AM - Posted by James

To be clear, the cpu probably connects to the GPU with just pci-express. If the mock-up is accurate, which it probably mostly is, the GPU and cpu are way too far apart for EMIB. The EMIB connects the HBM and the GPU. With EMIB they embed an upside down silicon chip for the interconnection. The chips pretty much have to be right next to each other for this to work.

November 8, 2017 | 11:49 AM - Posted by AMDisBeyondConsumerNOW (not verified)

If Intel/AMD gets too much of the laptop CPU/Graphics market share then Intel will be forced to offer SKUs with Nvidia GPUs/HBM2 also. Nvidia can sue right now before that happens and force Intel to offer an Intel/Nvidia solution because Intel is just EMIB nano-motherboard maker in this respect. So Intel will be forced to offer Nvidia based solutions or both AMD and Intel could be forced to license the x86 32/64 bit x86 ISA to Nvidia.

If it were up to me I'd force AMD and Intel to license their respective x86 ISA IP rights to the whole market the same way that ARM holdings does and look at all the innovation/competition that ARM ISA based markets offer.

Nvidia could go with some Power9/variant based SOC for the laptop market or try to beef up its Denver custom ARM ISA rinning IP but really the x86 ISA has had to much PC/Laptop market domination for 35+ years and I'd rather the x86 32/64 bit ISA be declared out of patent protection for anti-trust reasons. AMD has had 20 years with the x86 64 bit ISA while Intel's x86(16/32 bit) ISA is 10-15 years before that. So let's open up the x86 ISA like the ARMv8A/Other ARM ISAs are and let the entire worlds PC/Laptop market have at it.

November 8, 2017 | 09:13 AM - Posted by Orcblood (not verified)

This made had me LMAO. Wish I could hear the talks between apple and intel.

November 6, 2017 | 08:29 PM - Posted by OutWithTheDogFoodIntelGraphics (not verified)

"Update: We have been informed that AMD is producing the chips and selling them directly to Intel for integration into these new SKUs. There are no royalties or licensing, but the Semi-Custom division should still receive the revenue for these specialized products made only for Intel."

Yep AMD's selling them to Intel just like Intel is an AIB partner but Intel's client(Apple) funded the project with Intel the main contractor and AMD the subcontractor for the GPU part/HBM2 to GPU integration part. Apple funded AMD's semi-custom division indirectly, via Intel, to design the GPU part with AMD's semi-custom engineers working with Intel on the HBM2 to Radeon GPU integration part of the process. Intel can be thought of as Apple's main semi-custom partner with AMD's semi-custom division along for the their part of the deal.

If Zen were to market maybe one year eariler then this may have been an all AMD semi-custom project but deals like this where made a few years back just to be ready for 2018. For sure this will not include any of AMD's HBCC/HBC IP and any other Vega related IP that was not ready in time to be included in this deal. Just follow the Apple Seed trail back to where the real money came from and Intel had no choice in the matter if Intel still wanted some of its Intel IP inside Apple's Lappies!

November 7, 2017 | 07:52 AM - Posted by Coppermine (not verified)

As the spokesperson said, its mainly targeted for gaming, which are mainly laptops running Windows. But gaming is not Apple's target market. And also stop making up delusional hypothesis.

November 7, 2017 | 12:06 PM - Posted by OhSuchDaftynessAboundsInSome (not verified)

"spokesperson said", Ha ha, Marketing Monkey's boss said to spin this SKU for gaming also, but really gaming is not a large enough market compared to other laptop usage.

Gaming is mostly PC/console based, this has Apple's hands all over with that thin and light laptop usage being targeted by the anorexic laptop mavens at Apple!

Gamers piffle, that's PC's mostly for the freaks that need that FPS ePeen fix. Now Mobile gaming is not hard even for Intel's brand of dogfood. But for real graphics arts uses that AMD graphics on a laptop that's right up Apple's client base's ally. Blender 3D likes them higher shader counts on AMD's and Nvidia's Graphics while Intel's dogfood is not where is at for Blender 3d/Adobe/Maya graphics workloads!

November 7, 2017 | 02:44 PM - Posted by Coppermine (not verified)

Again, you did not read much news. Here it is, straight from the lion's mouth https://www.bizjournals.com/sanjose/news/2017/11/06/amd-intel-chip-partn... quote "Asked about the rumored announcement, an AMD spokesperson told the Journal that the new chips would “appeal to serious gamers” and would not compete with the company’s forthcoming line of Ryzen Mobile chips".

November 8, 2017 | 11:59 AM - Posted by AMDisBeyondConsumerNOW (not verified)

"serious gamers" is just another name for deadbeats with no means to pay the proper amount of markups for the advanced IP that those gaming gits so whine for.

For Both AMD and Nvidia it's those professional markets that will pay the proper markups and the gaming gits get the binned dies that fail to make the grade! And that's only to help AMD/Nvidia to have a place to dump their trash dies instead of the trash can! Nvidia/AMD can also make the gaming leaches tas BETA testers for Driver issues so both GPU makers can tune their professional drivers for maximim QA/QC to those enterprise customers that generate AMD's/Nvidia's real revenues!

November 8, 2017 | 03:57 PM - Posted by Coppermine (not verified)

Then explain the plethora of new laptops (not just gaming ones) that comes with discrete GPUs nowadays. Like Microsoft's new Surface Book 2 which uses Nvidia's discrete GPUs like GTX1050 and GTX1060. Heck even Chinese laptops like Xiaomi Mi Notebook Air uses GeForce MX150. Certainly these companies recognizes that there is a market for users who wants gaming on the go (and not just for productivity only). All your replies are looking more like whining and rants (using words like "deadbeats" for example).

As for professional GPU market (for datacenters, HPC and supercomputers), that is totally dominated by Nvidia currently. Additionally as evidence, just look at Nvidia's earnings performance which is dominated by their professional GPUs (Tesla, Quadro, etc) which can demand higher asking prices and batch volumes (especially for large supercomputer projects). For your information, typically low end to midrange GPUs don't earn as much. Sure, they may sell in large volumes (in the consumer desktop PC market) but the net profit is very small per card. You can see this in AMD's situation (especially with the gross margin).

November 6, 2017 | 08:51 PM - Posted by 96cores (not verified)

Ryan, please take some time on Wednesday's show talking about the difference EMIB and a standard interposer. I would also love to hear your opinions on a theoretical 8700x/Vega 64 Desktop chip :-).

November 6, 2017 | 10:12 PM - Posted by OutWithTheDogFoodIntelGraphics (not verified)

EMIB are just made with small silicon interposers that are embedded in the resin and are narrower and are used to connect the dies via higher density traces with the rest of the substrate made of organic resin/plastic. So only little strips of silicon are only needed for the chip to chip interface where silicon can have much denser traces etched than a substrate made of organic resin/plastic that can only have less dense traces.

Just look at the Graphic provided in this article and that's about it and Intel is not the only one that can make muli-chip designs this way.

November 6, 2017 | 11:08 PM - Posted by SuchConfusionOutThere (not verified)

I don't Know what to thA☐nk about why folks think that thA☐s CPU A☐s for the mass market but Intel's probably not goA☐ng to be able to sale to the wA☐der market beyond Apple's hardware.

Apple's A☐OS needs to be even more lA☐ke MacOS before Apple can begA☐n to make use of A☐t's ARM cores A☐n laptops.

November 7, 2017 | 07:47 AM - Posted by Coppermine (not verified)

The target market is gaming as mentioned by the spokesperson, which means laptops running Windows especially. But gaming is not a priority in Apple's target market.

November 7, 2017 | 12:14 PM - Posted by OhSuchDaftynessAboundsInSome (not verified)

"spokesperson said", Ha ha, Marketing Monkey's boss said to spin this SKU for gaming also, but really gaming is not a large enough market compared to other laptop usage.

Gaming is mostly PC/console based, this has Apple's hands all over with that thin and light laptop usage being targeted by the anorexic laptop mavens at Apple!

Gamers piffle, that's PC's mostly for the freaks that need that FPS ePeen fix. Now Mobile gaming is not hard even for Intel's brand of dogfood. But for real graphics arts uses that AMD graphics on a laptop that's right up Apple's client base's ally. Blender 3D likes them higher shader counts on AMD's and Nvidia's Graphics while Intel's dogfood is not where it's at for Blender 3d/Adobe/Maya graphics workloads!

Oh! Loads more shaders for those Ray Traced renderins that some time per frame to complete biut look oh so sweet!
FPS flicking ability(ROPs) is not needed for some Graphics workloads that like them shaders. We no need no stinking excessive ROPs, just more shaders and TMUs for some graphics workloads!

November 7, 2017 | 02:40 PM - Posted by Coppermine (not verified)

Don't take my word for it, get it direct from the lion's mouth here https://www.bizjournals.com/sanjose/news/2017/11/06/amd-intel-chip-partn... quote "Asked about the rumored announcement, an AMD spokesperson told the Journal that the new chips would “appeal to serious gamers” and would not compete with the company’s forthcoming line of Ryzen Mobile chips"

November 7, 2017 | 04:30 PM - Posted by DemJokezOnYouzKnuckleheadz (not verified)

That's marketing speak that “appeal to serious gamers”, seriously no one cares only about gaming MOOKs! That real professional/Graphics design money Talks and Gaming Mooks can walk! No one needs any loss leader revenue black hole gaming market that badly! The most fickle consumer market is the consumer gaming market, little e-Peen man/children who can not/will not pay the proper markups!

November 7, 2017 | 07:08 PM - Posted by Coppermine (not verified)

More here, directly from AMD themselves http://www.amd.com/en-us/press-releases/Pages/semi-custom-graphics-chip-... quote "Together, we are offering gamers and content creators the opportunity to have a thinner-and-lighter PC capable of delivering discrete performance-tier graphics experiences in AAA games and content creation applications". See that "discrete performance-tier" and "AAA games" that comes ahead of content creation part? Clearly is stating that gaming was the first/main target market.

November 8, 2017 | 12:15 PM - Posted by AMDisBeyondConsumerNOW (not verified)

Content creators are the productive class that should get the professional drivers while the gamers should be hitting the beaches on landing craft and earning the good investors of Amalgamated Bodybags some good returns on their investment!

Let's put backdoors in all the FPS/War relsted Games/console hardware tied directly into good old U-Sam's military conscription systems. Let's get those gaming deadbeats hitting the sands/fields/snow-ice as US Govt prime GI grade beef! Top FPS/War game scorers shoud be awarded with some very high resolution top FPS(Frames Per Second) based war gaming right down the ropes hanging from that UH-60!

November 8, 2017 | 04:06 PM - Posted by Coppermine (not verified)

And now what are you blabbering about? What's this about "Amalgamated Bodybags"? Looks like you're in meltdown mode, judging by your weird and incoherent replies sounding more like whining and rantings.

November 6, 2017 | 11:54 PM - Posted by Billiwam (not verified)

Neat.
Although seeing the CPU also take advantage of HBM2 would have been neater.

November 8, 2017 | 01:03 AM - Posted by James

A cpu probably wouldn’t benefit much from HBM. It is high bandwidth, but it isn’t really low latency. Modern CPUs are quite good at caching and prefetching. Connecting up HBM might not actually make much difference for the cpu. AMD’s rumored HPC APU looks like it is essentially an Epyc package (may not be same socket though) with two of the CPUs replaced by a Vega part with HBM. The CPUs will not have direct access to the HBM. This architecture would allow the cpu to connect to the Vega package with 4 interprocessor links. It will be exceptional for the GPU with fully virtualized memory (high bandwidth cache controller), since that allows a huge amount of bandwidth from main memory. Although, it would be interesting if they put some memory channels on the GPU die also to allow 8 channels to still be used.

November 7, 2017 | 01:02 PM - Posted by Mr.Book

I find it comedic how the Intel graphic render clearly displays the Intel branding on the right, while being devoid of AMD branding over the GPU pieces.

November 7, 2017 | 04:31 PM - Posted by James

To reiterate, while this is interesting, this is for very small devices only. It is intel cpu plus some Polaris based HBM graphics (single stack). For not so tiny systems, I would much rather have a Ryzen mobile APU and a Vega GPU with HBM (hopefully 2 stacks). It would be almost as small and Vega will be quite power efficient at much lower clock/voltage. It seems that the Ryzen mobile part will be quite good also since it can run at higher sustained clocks in a power limited form factor. Intel knew what Ryzen mobile could do, which is why we got the preemptive 4 core mobile part release. We just need to get Vega mobile parts for AMD based systems. I don’t want to pair Nvidia parts with Ryzen. Nvidia parts don’t age well anyway, which is bad for a non-upgradable laptop. Intel can have the old Polaris parts. HBM is actually a good fit for mobile with the high bandwidth at low clock and compact size, so hopefully we will see mobile Vega soon.

November 7, 2017 | 07:32 PM - Posted by Coppermine (not verified)

Whoah! The microchip world is turning topsy turvy.

More meltdowns coming soon. Raja resigns from AMD!

And one article says Raja is heading to Intel?

November 8, 2017 | 04:50 PM - Posted by Anonymously Anonymous (not verified)

nearly got a headache just from scrolling through the comments and not even reading them!

This collaboration is all about $$$.

Didn't Lisa Su say they don't intend to enable competitors?

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.