Intel is thinking of living it up a little

Subject: General Tech | December 20, 2017 - 02:22 PM |
Tagged: krzanich, Intel, cpu, gpu

Intel's recent releases have not been terribly exciting, the chips offer small incremental efficiency improvements each generation, with a few enterprise level features added which do not benefit enthusiasts.  The majority of the changes have been to the sockets, something many feel Intel could have left alone.  This may change, according to the memo from CEO Brian Krzanich who describes the coming years at Intel as akin to the mid-eighties when they risked everything to move from producing memory to producing CPUs and other chips.   This change will not focus solely on CPUs; he intends to capitalize on the IP which Intel has acquired recently as well as branching into the mysterious GPU which has been announced.  Drop by Slashdot for links.

View Full Size

"Anything that produces data, anything that requires a lot of computing, the vision is, we're there." The memo also underscores the dramatic change in the nature of Intel's business as it approaches its 50th anniversary in July 2018. "We're just inches away from being a 50/50 company, meaning that half our revenue comes from the PC and half from new growth markets," Krzanich wrote."

Here is some more Tech News from around the web:

Tech Talk

Source: Slashdot

December 20, 2017 | 03:55 PM - Posted by HighMarginAsAlwaysForChipolopy (not verified)

Wow looking at all the comments there and not much good to say. And CEO Brian Krzanich unloading some more Intel stock this time around. That mysterious GPU is probably more of an accelerator product than any sort of gaming product and Intel appears to be like an AIB partner with that semisemi-custom Radeon GPU added to the EMIB/MCM module and not very different than any other AIB card with the Intel SOC and its integrated graphics communicating with the custom Radeon die via PCIe lanes on that MCM. It looks like Intel is more about the same old data center focus where the best mark-ups are and that Intel will be sourcing semi-custom Radeon dies to place on that EMIB/MCM to make up for any Intel in-house graphics shortfalls on consumer devices. That's hardly any sort of Intel integrating Radeon graphics IP on the same die with Intel's CPU cores.

That EMIB is the only innovation and not any Redeon semi-custom die, as that's just to keep Intel inside of more consumer devices as AMD's Raven Ridge APUs will be taking more business back to AMD anyways. Intel is not going to be motivated to produce any consumer graphics discrete products and have to compete with both AMD and Nvidia for a slice of a shrinking pie that is mostly a low margin business in Intel's eyes. Raja is too busy working up some competition to Nvidia's Volta accelerator products with Intel not needing the TMUs/ROPs and such that are only of use for graphics workloads.

It's actually a less costly choice for Intel to source a semi-custom Radeon die and add that to an EMIB/MCM package so not much in that way of R&D expense to let AMD adapt one of its existing Polaris Micro-Arch IP based dies to an EMIB/MCM for Intel's consumer needs for graphics horsepower above what is offered by Intel's own in-house GPU designs.

AMD's Zen/Vega APU IP comes very tightly integrated via the infinity fabric and not PCIe! So until that Zen/Vega includes HBM2 Intel still has time to wait and see. Intel got a very good semi-custom deal on that Polaris/semi-custom die and that's something that Intel could never get/source from Nvidia, as AMD is more than willing to go very low margin.

That recently acquired Intel IP includes FPGA and AI processors and there really is no acquired GPU IP as that's just sourcing a GPU die in a semi-custom manner more like a discrete GPU AIB partner and less like a console device maker. Intel's Optane is new and Micron appears to be selling all of it's current share of that joint Intel/Micron XPoint fab's production to Intel rather than Micron producing any products under it's QuantX branding. So all of Intel's focus will continue to be on its data/cloud center and AI focused cash cow marketplace.

No one in the gaming market is really going to be able to best Nvidia at the top end of the performance charts and AMD appears to still be focused on the mainstream and compute usage of its GPUs, and that's working out great for AMD currently with its Vega GPUs being sold out before there is any chance to make it out of the stockroom and onto the retail shelves.

Intel must stem the Nvidia Tesla/Power9 tide with those New Tesla Volta micro-arch based SKUs that offer on one side some threat to Intel's real money maker in the server room/cloud centers around the globe. That and Intel's margins are just about ready to start taking an Epyc hit in the other broadside from AMD's newest Server focused designs that has AMD posed to retake at least that 20-25% share of the server market just as AMD had in the High point Opteron days.

So look at Intel's focus on that EMIB packaging IP and it's focus on cost savings is a very good indicator that Intel is preparing to cut costs as much as possible in order to respond to some very competative downward price pressures from Both AMD and Nvidia/OpenPower.

Brian Krzanich can say anything but he, at this point in time, has access to that golden ripcord that can be pulled to unfurl an equally golden parachute to float down safely to bank accounts overflowing with years worth of CEO perks and options.

December 21, 2017 | 08:28 AM - Posted by quest4glory

Why do you post your blog entries under new non-verified usernames all the time? Is Wordpress challenging?

December 21, 2017 | 12:51 PM - Posted by DaftFoolzNeedToFocusOnWhatWasWritten (not verified)

Up yours Peckerwood!

December 21, 2017 | 08:28 AM - Posted by quest4glory

tldr

December 21, 2017 | 01:10 PM - Posted by DaftFoolzNeedToFocusOnWhatWasWritten (not verified)

Enjoy Your Doogfood Intel graphics except where there is some Better discrete Radeon/Polaris die via PCIe(And in no way integrated) on That EMIB/MCM module. Looks like Intel got forced by Apple to create that EMIB or risk losing Apple's business. Hey anyways Apple has more caah on hand in its cash accounts then the entire market caps of AMD and Intel COMBINED! So how's that non x86 ARMv8A ISA based business making Apple damn near a trillion dollar market cap company!

AMD better be rethinking its custom K12 ARM cores looking at where Apple may be going and Apple makes more money off of the ARMv8A ISA than it does off of that x86 ISA. I'd rather see AMD get its K12 custom ARMv8A ISA running micro-arch out rather than AMD making the Intel mistake and focusing too much on that x86 only market.

I want there to be AMD ARM/Vega based APUs also as that will do better in the teblet markets for low power usage and Microsoft appears to be wanting to get some windows/win32 applications running on the ARMv8A ISA also.

But Apple is so damn big that their cash stash is bigger than the whole AMD and Intel x86 dog and pony show, Ha Ha Ha! You hear that AMD it's best to not throw away Keller's K12 design just yet!

December 21, 2017 | 06:19 AM - Posted by JohnGR

Krzanich has done everything wrong, so probably he is playing his last card here to step down latter with his head looking up, not down.
Intel looks like losing it's manufacturing advantage, managed to let AMD do a strong come back, needing to rush to update it's mainstream and HEDT platforms. Fell on it's face in the mobile market, left behind in AI, failed in creating a GPU. Even for Intel's shareholders, Intel's share price remained a boring thing to watch for many years.
After latest buyouts in AI sector and hiring Koduri, Krzanich probably is trying to create something before he gets the boot, so he doesn't end up as a failed CEO whose mistakes where covered up by AMD's inability to compete and the billions of dollars in Intel's pockets.

December 21, 2017 | 01:56 PM - Posted by OhLookyHereVernDatsNotGood (not verified)

Koduri is working on a Intel made Volta compute competitor with no focus on gaming and no ROP/TMU resources that's for compute/AI workloads only. Intel's getting it from Epyc/Radeon on one side and OpenPower/Power9/Volta from the other and Intel's IOT is DOA!

Amd go read what Charlie over at S/A has to say about that FPGA/HBM2 announcment from Intel:

"Intel’s trend of increasingly less credible press releases accelerated yesterday when they “Unveiled” an FPGA with HBM2 memory. SemiAccurate spent the last 12 hours or so digging into this and the more we dug, the less truth we found." (1)

And that 10nm process that never arrives:

"From the time we exclusively told you about 10nm Cannon Lake’s return from the fab, it was clear something was off. To say yields were bad was understating things to a degree that even the classic British humorists would not dare to delve. Normally from tapeout to product on the shelves, Intel takes ~12 months for server SKUs, less for consumer. According to SemiAccurate moles, it has been ~16 months and counting." (2)

(1)

"Intel’s claims on FPGAs with HBM don’t hold water"

https://semiaccurate.com/2017/12/19/intels-claims-fpgas-hbm-dont-hold-wa...

(2)

"What is the state of Intel’s 10nm process?"

https://semiaccurate.com/2017/12/20/state-intels-10nm-process/

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.