Lucky number 7 for AMD

Subject: General Tech | October 2, 2018 - 01:24 PM |
Tagged: zen 3, rumours, amd, 7nm

There is something interesting going on at AMD according to the rumours coming out of The Inquirer today.  It seems that they will be moving from their current 12nm process node directly to a 7nm node for the next generation of Zen processors, both consumer and enterprise.  AMD has confirmed the next EPYCs will be on that node, but the news that Ryzen 3 will be was previously unknown.  The chip itself is expected to have a base clock of 4GHz with 4.5.GHz top boost, eight cores with 16 threads and performance close to Coffee Lake S.  The performance comparison is not all that useful without more information about what scenarios this would refer to, perhaps single thread performance which would be a nice jump.

As always, grab your salt shaker before heading over to follow the links back to the source as we will not know much more until next year.

View Full Size

"THINGS ARE ALL ZEN at AMD but probably not in the way you'd think, as leaked information has spilled the beans on what shape the chip maker's second-gen Zen architecture could look like."

Here is some more Tech News from around the web:

Tech Talk

 

Source: The Inquirer

October 2, 2018 | 01:43 PM - Posted by Anonymouszzzzz (not verified)

"4.GHz top" missing boost clocks there I think. (hoping 4.5 or so)

October 2, 2018 | 05:39 PM - Posted by Tim Verry

Rumor says up to 4.5 GHz ;-). Will update asap.

October 2, 2018 | 03:20 PM - Posted by CaedenV (not verified)

Would be some sweet justice if they break 10nm before Intel

October 2, 2018 | 04:11 PM - Posted by Maybe5GhzAfterReleaseToMarketForZen2BasedRyzen3SKUs (not verified)

So the Zen-2 micro-arch for Ryzen 3 gereration of that consumer branded "Ryzen" line of products.

So that's a Rumored Zen-2 at 7nm, 8 core engineering sample with a clock rate of 4.0 base and 4.5 boost. Then what where the clocks on the early engineering samples based on first generation Zen Micro-Arch at 14nm and how much did they improve over time before they where declared RTM.

It's not hard to guess that all of AMD's Server and Consumer SKUs will eventually get a Zen-2 at 7nm treatment because that will be the latest and greatest from AMD in 2019 what with Zen-3 still in active development. AMD would be foolish to not try and get any Zen-2/Zeppelin dies used for any Ryzen-3 consumer Branded desktop SKUs in addition to any first use as Zen-2 on AMD Epyc/Rome Branded server offerings.

But what is still not Known with any Zen-2 Zeppelin DIE layout for Epyc/Rome is will it come with that Same 4 CPU cores to a CCX-Unit and 2 CCX units per Zeppelin Die arrangement that was used for Zen-1 Epyc/Naples and also the Ryzen first generation consumer desktop SKUs.

Zen+(consumer only) is what Ryzen-2 desktop and TR2 HEDT is based on and Raven Ridge APUs are still based on 14nm Zen first generation designs.

AMD's various specilized Zen+(consumer Only) based CPU die tapeouts are not based off of any Pro Zeppelin DIE that is binned down for desktop usage like Zen/Zeppelin was. So now there is some confusion as to how long Zen+ will be in use but looking at Wikichip's Zen+ entry there appears to be some Ryzen Pro Branded SKUs under Wikichip's Zen+ entry so those parts will be around much longer than any of the non Pro Zen+ based Variants.

If for Zen-2/Zeppelin AMD is still going to be making use of some style of 8 core/2 CCX-unit based die designs then maybe Ryzen-3 desktop products can appear very shortly after the Zen-2 Epyc/Rome products are available as there will always be some defective DIEs available from any large Ramp up in Zen-2/Zeppelin DIE production.

But if AMD's Zen-2 Epyc/Rome Die tapeouts start to diverge in CCX unit layouts or Number of Cores Per CCX-Unit per Zen-2/Zeppelin die then things may be different for Ryzen-3 Desktop parts somewhat. AMD could still bin the Zen-2/Zeppelin dies down even if they have more than 8 cores to begin with but that's going to change things up a bit for AMD and maybe AM4 SKUs with options for more than 8 cores may appear.

Certianly TSMC has probably already realized that they had better hitch their Flagship Process node horses up to AMD's Epyc/Rome wagon as that will represent even more wafer sales for TSMC in addition to any Vega 20 and Navi production at 7nm/TSMC. That's even more potential with Epyc Rome than even Nvidia's fab business represents for TSMC. AMD's potential for raw numbers of Server CPU sales represents a bit more sales/revenues than any GPU only sales figures. The only bigger customer for TSMC currently is Apple but still AMD's potential Epyc/Rome based fabrication business for TSMC is potentially very large also.

Apple has been Known to change Chip fabs if that suits Apple's needs. So TSMC landing all of AMD's 7nm Epyc/Rome production in addition to 7nm Vega 20 GPUs and 7nm Navi also in addition to AMD's consumer CPUs at 7nm is a very good thing for TSMC. And that AMD business represents more opportunity for TSMC's longer term customer stability than totally relying on Apple who can and does change its mind so often, much to the detriment of any of Apple's suppliers for parts or fab services.

Apple has the funds to Build a little Apple fab capacity of its own if necessary or even help its suppliers fund expansion if Apple thinks that will help secure constant and secure parts supply streams for Apple's needs. But Really TSMC can not lose now that they have some of the most stable and longer term based server CPU production being done in TSMC's Fabrication lines from AMD that represents more longer term stability for any third party Fab company. Even GF will still be making First Generation Zen Epyc/Naples SKUs for a good while as server market customers stipulate that longer term SKU availability in their contracts. So extended parts supply availabitlty must be guaranteed by AMD.

October 3, 2018 | 10:53 AM - Posted by Anonymous1 (not verified)

I hope that there will be Apple computers with AMD cpu one day. :D

October 3, 2018 | 03:24 PM - Posted by MoreLikePayTheOEMsInOrderToGetInsideForAMD (not verified)

Considering how much Intel spends to do some expensive engineering for the laptop OEMs, AMD really needs to build up some Epyc Revenues so AMD can buy its way into the laptop market.

Intel will spend billions and pass that cost on to all consumers of Laptops in higher average CPU pricing for Laptop/Mobile that use Intel CPU/Graphics SKUs. So Most folks will have to wait for AMD to begin to offer such levels of OEM laptop end product engineering/other support where AMD helps out the laptop OEMs also.

AMD will not be able to offer as much incentives currently but that may change. Apple does all of its own engineering including Apple making use of not really the latest Intel CPU/Graphics products. So Apple is not really about bleeding edge Intel or AMD technology unless it's Apple's own in-house CPU/SOC designs. If AMD would design a custom Zen APU with its own 4GB stack of HBM2 in a thin enough(Z hight) module then AMD would get Apple's attention but Apple would have to supply a TB3 controller for the device or TB3 Controller IP to AMD to include on AMD's Chipsets.

I think that Apple is more interested in moving to its own custom ARM cores on its laptop devices and AMD may be just a little too late for that Apple x86 ISA related Laptop SKU business. AMD may be able to win some Apple Mac-Mini business with Raven Ridge APUs inside but that dependes on how much pull Intel still has with Apple over the next few years.

In the long Term for AMD I think that they would be better off becoming a Laptop/Mini-Desktop-PC OEM or Partnering with a Favored Laptop/Mini-Desktop-PC OEM with AMD helping with the engineering costs necessary to build a line of proper AMD Raven Ridge based laptops and Mini-Desktop-PCs.

AMD does need to begin to think about designing/building a Similar to Intel's NUC sort of Product line also and use that as a way to get its Raven Ridge processors/graphics into a larger share of that Mini PC form factor market.

Intel is doing a lot of Display reserch and as such just look at this Verge article(1):

"At Computex today, Intel made a brief but potentially momentous announcement: it’s developed new power management technology for mobile displays that can cut battery consumption in half under some circumstances. The company didn’t go into the details of how its Intel Low Power Display Technology works, but it’s a combined effort with Sharp and Innolux, who are manufacturing the 1W LCD panel required for it. Sadly, you wouldn’t be able to retrofit this battery life extender into your existing laptop. Another requirement is that you’d have to use Intel’s graphics instead of Nvidia or AMD’s superior alternatives"(1)

So that should reveal the lengths that Intel goes to keep its CPU/Graphics Products inside laptops. And AMD has some design wins with Raven Ridge APUs but Intel is still keeping AMD outside of Zotac's product offerings and limiting the other Laptop OEMs from supporting better AMD Raven Ridge based OEM laptop options. Intel will even go as far as contracting with AMD's semi-custom division for a Graphics DIE to place on Intel's Kaby Lake G line of SKUs. So any Sorts of "NUC" like products are going to require AMD, or AMD and MB-partner, to engneer a mini form factor Motherboard(Socketed AM4) and Case/Power supply design.

Short of that the only way would be Crowdfunding a Raven Ridge based "NUC" like offering that runs Linux. And That's a very hard thing to do.

The entire OEM Laptop/Mini-Dasktop-PC market is like the Endcap Part of the Display Aisle at the grocery store and Product Makers have to pay the Store for that high visibility product placement. Really the entire OEM based PC/Laptop market is that way.

(1)

"Intel claims it can cut display battery use in half with new tech"

https://www.theverge.com/2018/6/5/17427180/intel-display-battery-power-s...

October 3, 2018 | 10:54 AM - Posted by Anonymous1 (not verified)

And NO i am no ROBOT! Why i have to verify that over and over and over again?

Are you guys afraid of ROBOTS?

October 3, 2018 | 11:24 AM - Posted by Naruto Uzumaki (not verified)

Chill out Anonymous1 how are you going to pass the Turing test with that attitude, Nvidia did.

October 3, 2018 | 08:30 PM - Posted by KevinM (not verified)

I wonder if Ryzen 3 will be able to work in X370 motherboards like Ryzen 2 can? That would be great if so!

October 3, 2018 | 10:41 PM - Posted by MaybeSomeCaveYouHaveBeenLivingWithoutInternetAccess (not verified)

Did not AMD state that AM4 would be supported until 2020 and that date sort if implies that Ryzen 3(2019) will still be AM4.

So whatever the X or the A to the B to the whatever Motherboard if that MB is socket AM4 well that's until 2020. So I can't really see AMD releasing any consumer CPU/APU, that's NON Threadripper/SP3 related, on any other consumer Socket but AM4 until 2020.

I'd love for AMD to Increase in its Next Socket AM[Whatever] pin counts and offer more PCIe 4.0 connectivity(PCIe 5.0 even better) and more total PCIe lanes instead of so few PCIe 3.0 as the current AM4 offers. Laptops are in more need of PCIe 4.0, by the way, as that would allow the Laptop OEMs to offer more bandwidth without having to increase the numbers of PCIe lanes on their always custom laptop MB designs.

Part of the reason that TB3/Newer will never be that wide spread currently for laptops is because Laptop's have less MB real estate for PCIe lanes so PCIe 4.0/5.0 can not arrive soon enough on laptops. DDR5 memory also would be a nice all around improvment for any Integrated Graphics if the makers could not begin offering APUs/CPUs-Int-Graphics with any HBM 2/3 options available that fit on the socket. Maybe even some smaller 1GB of some sort of On Die/On MCM memory to feed the integreted GPU that's attatched to the GPU's HBCC(High Bandwidth Cache controller) and used as a HBC(High Bandwidth Cache) like AMD's Vega graphics offers. but that can not be useful for current AMD/Raven Ridge APUs without some form of higher bandwidth VRAM/Cache memory to act as HBC.

Zen2/Next Generation APUs need some form of eDRAM even if its only 500MB in size so that the HBCC/HBC IP in Vega could be of use and that eDRAM/Smaller HBM memory could be used by Vega as a last level VRAM cache to a larger pool of virtual VRAM paged out to regular system DRAM.

I'd expect that AMD should begin to discuss it's CPU socket roadmap/timeline at CES 2019 and give consumers some previews of what's going to be replacing AM4 in 2020. And that better include the newer PCIe standards and DDR5 memory at least. Maybe even TB3 IP on the chipset also as Intel appears to be maybe trying to open that up at least. But I'm not hopeful on that TB3 so maybe USB 3.2 chipset support and 2 channels of USB 3.1 Gen2(10Gbs) link bonded to give a total of 20Gbs of bandwidth over the same USB Type-C compatable cable's already existing USB 3.1 Gen 2 link and the USB cable's Type-C alternative channel that can support the other USB 3.1 Gen 2 siginaling that comes with the USB 3.2 standard.

If the USB Type-C cable if following the proper USB Type-C standard then all the wiring is already there for any Type-C cable to already have insluded that alt-mode cable pair if the cable's maker is not cheaping out and not following the USB Type-C standard like they should. So USB 3.2 should be able to work on existing Type-C cables(Maybe).

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

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.