Jim Keller Leaves Tesla for Intel

Subject: Editorial | April 26, 2018 - 02:34 AM |
Tagged: Zen, tesla, raja koduri, Jim Keller, Intel, Conroe, Banias, amd

Update: The official Intel announcement can be found here.

For anyone that follows the twists and turns of the semiconductor world, the name “Jim Keller” is approaching legendary proportions. He was a driving force in AMD’s K7 and K8 development, he moved on to PA Semi which was acquired by Apple to produce their class leading SoC’s for the iPhone, and then went back to AMD to become lead architect of the Zen architecture which powers the latest Ryzen CPUs from AMD. He then moved on to Tesla to be in charge of chip development for their autonomous driving program.

View Full Size

Very little has been heard from Jim Keller while he was at Tesla. The assumption was that he continued to do his job there and worked hard to innovate the potential chip designs that would power next generation Tesla vehicles to have fully autonomous driving capabilities. While that program has been in its infancy, we have not heard of custom chips being utilized by Tesla in the latest cars.

Now we have confirmation that Jim has left Tesla and has in fact been hired by Intel. Some months back Raja Koduri was hired by Intel to be in charge of all core development with a special interest in GPUs. It looks as if Raja has persuaded Jim to hop on board and help with what appears to be a stagnant core development team on the CPU side.

Intel has a history of “not invented here” mentality that has in previous years caused massive problems with the company. The reliance on the Pentium IV and its further development allowed their primary competitor to sneak up on them and shake up the marketplace. It took a design group out of Israel to set Intel onto a better path with the Banias/Conroe architectures which then lead to the Core architecture that we have seen iterated upon for the past decade.

The company has stagnated again. While the current Core architecture is faster in terms of IPC than Zen, it is a company that has not pursued innovation in a manner that has kept its competitor at bay. Jim Keller went back to AMD and architected what would become the Zen family of chips. In the space of those years he was there, he took the best technology AMD had to offer and built from the ground up a new architecture that could compete against Intel for a fraction of the R&D costs that the semiconductor giant typically spends. Intel stands to lose some significant marketshare in mobile, desktop, and server with the latest offerings from AMD. Combine this with the issues that the manufacturing group have run into with their development of the 10nm process, Intel seems to finally realize that design is really what matters when manufacturing issues hit. We can remember back in the Athlon 64/Pentium 4 days when AMD was 18 months behind on process technology, but still held a power/performance edge over Intel. While manufacturing can give a large advantage to any chip, a great design will not have to rely as heavily on cutting edge process tech to be competitive. Intel should hold all the keys to creating a truly overpowering series of products for their primary markets, but AMD has shown up with the plucky architecture that could cause some serious perturbations throughout the mobile, desktop, and server markets.

View Full Size

It seems that Raja is “getting the gang back together” to revamp the design culture at Intel to more adequately deal with threats to their CPU dominance across the board. They also are probably looking more closely at the ultra-mobile market that ARM has dominated for the past decade. Previous Atom designs have not come close to the efficiency needed to address those markets, but perhaps with a change of leadership and architects we can see Intel successfully address this very important area with high performance/high efficiency chips that we honestly expect them to be able to design.

Jim Keller to Intel looks to be a transformational move. Not just because of his expertise in architecture, but also a shift in how Intel goes about its daily business. Bringing this kind of expertise into the company is a watershed moment that moves away from the “not invented here” mentality that seems to dictate decisions at the company when they are not facing serious competition. We will see what kind of power Raja and Jim can leverage in changing the culture of the company. What cannot be denied is that Intel has frittered away its advantages in core design by not implementing aggressive product and feature changes for the past decade to insure its dominance in the CPU world. Compound this situation with the manufacturing woes at 10nm and we can see that Intel needed a shakeup.

Consider Intel shook.


Source: Intel

Video News

April 26, 2018 | 03:59 AM - Posted by anon. (not verified)

consider today Intel Q1 '18 Earnings Repport.

April 26, 2018 | 05:19 AM - Posted by Henry (not verified)

Not quite so sure if this is a good development for consumers. Everyone knows Intel's been profiteering by making every subsequent generation require a new motherboard. And now that Ryzen is out, we can see how ludicrously priced the top end chips were. If Keller and koduri end up making intel chips perform better than amd, whose to say intel won't return to those practices.

April 26, 2018 | 09:31 AM - Posted by Anonymouse (not verified)

Every other gen, you mean. And Intel and AMD currently have price parity.

April 26, 2018 | 05:31 PM - Posted by PriceParityDaftnessFromAnonymouse (not verified)

Do they really! Let's us a price/per-core metric and see how that works out in AMD's favor! Then there is that included on the Box cooling soultion that AMD offers.

So for the Intel 8700K that's 6 cores/12 threads at Intel's MSRP of $359-$370 and AMD Razen 7 2700X at 8 cores/16 threads at 329.99 MSRP.

I'll take the current NewEgg pricing of the 8700K at 339.99 and the Ryzen 7 2700X pricing at NewEgg of 329.99.


Ryzen 7 2700X: 329.99/8 = $41.25(rounded) per core

Intel 8700K: 339.99/6 = $56.67(rounded) per core

That's not price parity on a per core basis and what about GN's gaming while streaming tests and that 8700K not doing so well with its 6 core relative to the 2700X and its 8 cores.

Now add the monay saved by having that cooler included in the Ryzen 2700X Box at no extra charge and that some extra sevings above and beyond.

What a spceial kind of Fool you are!

April 27, 2018 | 03:52 AM - Posted by Anonymouse (not verified)

"Let's us a price/per-core metric"

Why? It has no value.

Instead, use price/perf filtered by the applications you actually use (i.e. no point comparing Blender performance of you do not use Blender).

April 27, 2018 | 03:59 AM - Posted by psuedonymous

Plus, outside of HPC workloads scale with Amdahl's law, not Gustafson's, so adding more cores does not come close to linearly scaling performance.

Bigger number != better. If you use predominantly lightly parallel workloads (e.g. desktop applications, gaming) then adding more cores just adds dark silicon.

April 27, 2018 | 09:01 AM - Posted by OhTheGrossMarginFallAndResultingShareValueDrop (not verified)

Those 2 extra cores for less total money are good for the game streaming bumpkins. so that's where the value of Ryzen 7 2700X can be found. So sure there is a marginal return as core counts get higher but for Ryzen 7 2700X where games may be using up more of any 6 core processor's total available cores/threads that's where the Ryzen 7 2700X with its 8 core/16 threads outshines the 8700K, in addition to any Blender/other workloads that like more cores threads.

Threadripper is like a dual socket configuration on a MCM with those dual Zen/Zeppelin dies operating in NUMA Mode so the game can run on one Zen/Zeppelin die and its near memory while the streaming/encoding could be done via the other die/its near memory and the little gaming narcissists can stream in very high resolution to all the other little gaming narcissists on the interwebs.

Threadripper is definitely what any high end game streamer would want if game streaming is their sickness, and the game's affinity set to make use of one TR's Zen/Zeppelin dies for gaming and the other Zen/Zeppelin die for streaming/encoding with each Zeppelin die on the TR MCM in NUMA mode and not fighting each other for memory latency in that NUMA cores/memory mode where each Zen/Zeppelin die gets its own local NUMA dual memory channel all to itself. Gaming is very memory latency sensative so one Zen/Zeppelin die for the game with its complement of memory channels and the streaming software on the other Zen/Xeppelin die's NUMA node dual memory channels and no game and streaming software fighting over memory queued R/W traffic and any extraneous latency affects that worsen the games's play while streamng.

AMD should be working TR for that very uasge and be helping the gemes developers and game streaming software developers take advantage of TR's dual Zen/Zeppelin dies and NUMA mode ability on that single socket TR MCM for the best game streaming experience at the higest streaming resolutions.

So game streaming on TR is really an example of a non HPC workload that could make good use of TR/TR4 and even Ryzen 7 2700X gets a great uplift from those 2 extra cores relative to the 8700K and all the Amdahl's law marginal return on extra cores does not really begin to effect processors at the 8 core level that's more for higher core count processors and clusters running a single application.

Poor Intel's market Cap is going to take a hit no matter if Raja/Keller or J. H. Crist is on Intel's team, as the AMD Epyc ecosyetem TCO and Price/Performance metrics are what is going force Intel's gross margins lower. Just you wait until the Wall Street Quants' algorithms start to mathmaticly notice Intel's gross margin slide, and those algorithms can sure do some fine future extrapolation on any conpany's gross margin trends.

AMD's Zen/x86 is performant enough to tip that TCO and Price/Performance balece in AMD's favor even fruther than Opteron ever could! And the only thing that Intel will be able to do is suck it up and lower its margins/markups on a per unit basis and that's really going to make that gross margin metric go south for the Winter Is Comimg moment for Chipzella's Market Cap!

April 27, 2018 | 10:07 AM - Posted by Bill Murray (not verified)

You are out of the loop. Intel just had a Q1 2018 record crushing earnings despite all your runaway hyping. Datacenter YoY compared to Q1 2017 was +24% with $5.2B. Makes Epyc look more like a bit player rather than a major competitor.

April 27, 2018 | 12:31 PM - Posted by JustWaitTillAfterThe2Hof2018QuartersAreOverAndReportedAndSee (not verified)

Intel's gross margins are down and those gross margins are what affect investment. Intel's larger revenues reflect Intel's large size but If Intel's margins start to shrink too much investors will flee. Intel's rather large market cap and continued share value is reflected more by investor hopes for increaed equity value even more so than only earnings per share.

Shrinking gross margins can wipe out billions in equity value if investors(Wall Street) fear any large scale gross margin shrinkage is on the horizon. Wall street and the short term investor market(Options Based) can do a number on any company's share value if those gross margin figures are moving in the wrong direction.

Record cushing earings only amount to a few cents per share and still qualify as a record and those new figures may not even be at or above the inflation rates so any constant dollars figures have to be taken into account as well. Sure positive revenue numbers are good if you want your lenders to keep lending to you then positive revenue growth is necessary. But higher gross margins are what generate that revenue above and beyond the break even point. And Wall Street weighs any gross margin shrinkage as a bad indicator for the future especially with that serious Server/HPC market competition returning to the Server/HPC marketplace.

That's Epyc(Much Lower Markup) on the lower cost side and even IBM/Nvidia Power9/Volta in the high end side for that OpenPower server competative price/performance pressures against all the x86 market to consider.

Intel's gross margins at around the 63% range are very high by any industry standards so with Epyc now in the server room and at around 1/2 the cost in some cases it's going to force Intel to have to lower its markups and that directly affects those gross margin numbers for the worse.

Wall Street watches(Computerized trading algorithms) those gross margin figure basis points like a hawk for the slightest signs of a shift towards declining margins(Lower Markups) pressures brought about by fierce competitive pricing pressures.

Just look a Zen/Ryzen on 8 cores and see how that forced Intel to lower its High Core count SKU pricing below the $1000 dollar mark so damn quickly for those 8 and 10 core parts. Now look at the even higher Intel server SKU markups and Epyc's competative markup pricing pressure with Epyc SKUs costing much less than Intel's costly server kit.

Epyc is so low cost that even Threadripper can not compete with Epyc/SP3 motherboards on a cost/feature basis and that's a 180 degree reversial compered to Intel and Intel's consumer SKUs that are always have lower cost/feature compared Intel's Uber Costly server grade kit.

April 27, 2018 | 04:49 PM - Posted by Bill Murray (not verified)

Your runaway daydreams are over. Wake up and feel the real sunshine. Already 3Qs has passed and yet too little Epyc contribution. AMD's EESC $532M including conslow chips versus Intel's datacenter $5.2B. Stark difference of nearly 10X disparity.

April 27, 2018 | 10:53 PM - Posted by IgnoramusMaximusBillTheBumpkin (not verified)

You are full of it, You Bumpkin and still you think of this as some size matchup and Epyc's rollout is right on time by server standards where it's some time after any server SKU's release to market before the server clients have fully vetted any new server CPU SKUs for production workload status.

You are another daft fool that thinks high technology is like some god awfully stupid sports matchup!

And the PC Enthusiast Press is currently in the process of embarrassing itself in confusing Jim Keller(Computer Architecture Engineer) with some of the Fab Process Technology materials/fab process engineering folks! And Keller is no materals engineer or Fab process technology engineer.

So Keller is not going to be able to fix any of Intel's current or future 10nm fab process node issues that's not even Keller's job or Keller's area of responsibility at Intel. Really ExtremeTech and a few other outlets, I do not see how your "reporters" can even look at themselvs in the mirror each day after some of those more egregious journalistic errors they make.

April 28, 2018 | 01:41 AM - Posted by Bill Murray (not verified)

According to latest Q1 2018 earnings, Intel still making a killing despite limping along with 14nm. The new Jim Keller hire is just going to pile more pain on top of AMD's situation in upcoming months or years. I'm sure you are squirming in rage at Keller for joining your most hated obsession.

April 28, 2018 | 09:41 AM - Posted by BlindBillTheIntelShillDoesNotSeeTheMarginsAsWallStreetDoes (not verified)

No Problems with Keller and Raja working for whoever they want. Intel's Gross Margins are going to be in question in a manner exactly that same as when AMD's Opteron forced a downward pressure on Intel's high margins.

Keller, Raja, or J. H. Crist on Intel's team is not going to stop AMD's Epyc from affecting Intel's gross margin figures over the next few years. Ditto for that IBM/Nvidia OpenPower Power9/Nvidia Volta competition and Intel's higher margins taking a dive!

Epyc is quite a bit less expensive than Intel's pricy server kit and Intel can not afford to lose market share so Margins will fall in the face of competition that costs much less to own.

April 28, 2018 | 03:30 PM - Posted by Bill Murray (not verified)

Low gross margins are AMD's specialty and not going to change anytime soon. Intel had always maintained a healthy and high gross margins throughout its fight with AMD even in those K8 era. AMD's Epyc still not doing enough. Quotes from Anandtech "As we saw in AMD’s earnings, EPYC hasn’t really made an impact on their earnings yet, but Intel continues to dominate in this segment."

April 26, 2018 | 05:19 AM - Posted by Henry (not verified)

Not quite so sure if this is a good development for consumers. Everyone knows Intel's been profiteering by making every subsequent generation require a new motherboard. And now that Ryzen is out, we can see how ludicrously priced the top end chips were. If Keller and koduri end up making intel chips perform better than amd, whose to say intel won't return to those practices.

April 26, 2018 | 06:18 AM - Posted by forgerone (not verified)

Meltdown does not affect the Keller designed secure Translation Lookaside Buffer in AMD x86 cpu's.

Likely Keller was brought back to solve that problem for Intel.

Any Meltdown solution using Bios or software patch from Microsoft will result in a massive performance hit.

April 26, 2018 | 10:43 AM - Posted by Anonymously Anonymous (not verified)

Most likely, it isn't any one specific reason why Keller was hired on, it was multiple good reasons adn the one you mentioned is just one of many.

April 27, 2018 | 02:29 PM - Posted by WhyMe (not verified)

The engineering on the hardware to sort out Meltdown would already have been done, if anything he's probably been brought on board to improve their multicore designs.

Intel have basically been stitching together what was a design based around a single core for over a decade and that's starting to show, especially with the way Zen is designed and everyone wanting more cores because of AI/machine learning.

April 26, 2018 | 06:24 AM - Posted by Hypetrain (not verified)

Likely to help with developing the Ocean Cove-architecture.

April 26, 2018 | 07:34 AM - Posted by John H (not verified)

Jim Keller is awesome in a way that reminds me of Jay Miner :). Respect to both.

April 26, 2018 | 09:37 AM - Posted by KellerWillBattlePower9RajaWillBattleVolta (not verified)

"Intel stands to lose some significant marketshare in mobile, desktop, and server with the latest offerings from AMD"

Intel is only going to lose some x86 server market share to AMD as that market can not be bought as readily as the consumer/PC market is by Intel.

Intel has no presence in the larger ARM based phone/tablet market and any OEMs in tha mobile market have their more power efficient ARM ISA SOC supplies that have the mobile devices OEM not dependent on any one SOC supplier like exists in the x86 market. The mobile market will remain ARMv8A ISA based and even Intel could not buy(Contra Revenue) its way into market domination there.

Intel pays its way into most of its PC/laptop designs wins either with engineering help to the OEMs or Marketing assistance. So AMD has been forced into semi-custom dealings with Intel to at least get AMD's Graphics into that Kaby Lake series of offerings. The only true in that AMD has to get its Ryzen Prducts into consumers' hands is through the home system builders market where consumers can purchase Ryzen/Raven Ridge more directly via the direct retail channels. AMD lacks the funds to offer PC/Laptop OEMs that level of product engineering assitance and/or marketing assitance that Intel can offer.

With Keller on board I fail to see any x86 advantage for Intel as they already have the better x86 CPU core designs and Intel only lacks the willingness to lower its pricing and risk lowering its overall gross margins longer term.

Maybe Keller is there more besause of that OpenPower Power8/Power9 paired with Nvidia GPU accelerators market and Keller is maybe there for some More RISC like with higher processor threads per core competition against IBM/Nvidia and that OpenPower consortium that's netting many high profile Government Supercomputing design wins. Jim Keller does have plenty of RISC design experience under his belt. Keller has that co-authorship of the HyperTransport specification also and Intel needs and answer to Nvidia's NVLink and AMD's Infinity Fabric. Intel badly needs some form of Design that can compete with Power8/Power9 and Nvidia's NVLink that's a bigger technological threat to Intel than AMD's x86 is currently.

Intel's biggest threat from AMD is AMD's lower pricing that threatens Intel's traditionally higher server SKU markups that directly translates into Intel's rather high gross margins. And without those very high gross margins quarter to quarter Intel's share price would tumble.

Keller is not there to fix Intel's x86 Keller is there along with Raja to battle IBM/Nvidia and that OpenPower Power8/Power9 with NVLink CPU/GPU accelerator ecosystem that's coming at Intel from the top down while AMD eats away Intel's x86 server margins from the bottom up!

April 26, 2018 | 06:52 PM - Posted by Bill Murray (not verified)

AMD market share shrinking to almost zero in coming years now that Keller and Raja enjoy free reign and near limitless budget at Intel.

April 26, 2018 | 07:57 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

Really Intel's supercomputer market share is up against OpenPower(Nvidia/IBM and others) Power9s paired with Volta GPU acceleratrs on those Government Supercomputer design wins. And It's Intel's costly x86 based server pricing that will net AMD's Epyc SKUs more market share regardless of any Intel small lead in IPC over the mext few years. Keller has nothing to add to Intel's x86 as Intel already has its in-house engineering folks for that. Keller can not produce any radically new design for Intel when that process takes years to complete from conception to execution and on to verification and certification before the big server/HPC interests will even begin sampling any new unknown Intel Kit.

AMD's Opterons sure netted AMD around 23% of the server market sales back at that time and Epyc is way closer to Intel's current offerings in IPC and way lower in price compared to what Intel is charging for its server SKUs.

Intel's gross margins are right at 62.3% (-1 percentage point) as of the latest Q-earnings and Intel's gross margins will take a beating from Epyc as AMD begins to ramp up its server business over the 2H of 2018 and beyond. Watch AMD's Epyc take market share back frocing Intel to forgo those higher margins and watch how Wall Street reacts to any substantial Intel Gross Margin shrinkage.

Thoes Zen/Zeppelin DIEs are coming off of the GF diffusion lines at greater than 80% die/wafer yields and that's very low cost for AMD on a usable die/wafer basis allowing AMD to argessively price Zen/Epyc to take more server market share from Intel over the next few years it will take Intel to tweak their x86 designs. AMD's already looking at 7nm products so there goes any Intel Fab Node advantages and Intel's 10nm is nowhere currently with larger numbers of Intel 10nm production still uncertian!

Are you with Viceroy Research with that almost zero market share nonsense.

Keller on Not, and Raja Included, Intel is not welcome inside any of the ARM ecosystem devices market and even Intel wasted 14 billion trying to buy its way into the mobile phone/mainstream tablet market with that failed Contra Revenue scheme.

Keller will most likely be trying to Help Intel compete better against Power9's in the server/HPC markets paired with Nvidia Volta GPU accelerators! And Raja will be helping Intel with that GPU accelerator part of the CPU/GPU competition that IBM/Nvidia via OpenPower Power9/Volta are bringing to the market. AMD will via its much more affordable Epyc pricing latitude begin to carve out more x86 server market share over the next few years. And Intel can not get those large monolithic die yield intrensic issues to help with matters of usable Die/Wafer Yields so Intel's pricing latitude will not be as great as what AMD's Zen/Zeppelin modular scalable dies will allow.

Free Reign Design takes even more time and even Intel can not speed that process up for CPUs or GPU and that takes years.

Really you are not in possession of very much grey matter in that large lipid storage vessel between your very hairy ears!

April 26, 2018 | 08:15 PM - Posted by Bill Murray (not verified)

Desperate sounding you are. IBM supercomputer very few there are. Everything else is Intel x86 takes supercomputer share the most. Epyc too late already. In coming years AMD still laggard in second place behind Intel and even IBM.

April 26, 2018 | 08:24 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

Epyc is on time and priceed to win at that Price/Performance metric an a lower TCO also!

AMD's revenues will only go up while Intel's revenues and high margins will take a hit along with Intel's Share Prices, once those gross margins start to fall!

Google is testig power9's and even Epyc has a chance there also as Epyc is priced to take it back to that 23% server market share once again!

Cletus Spuckler, yo SWIFE(Sister-Wife) is a callen you down to dah cement pond.

April 26, 2018 | 08:37 PM - Posted by Bill Murray (not verified)

Epyc too little and still too late to market. Server market share still under 1% at most. AMD's revenue and gross margin did not reflect Epyc contribution much. Not enough Epyc sold. So AMD stock price could not rocket higher. Last checked only 11, still below the highest. No one wants to compete with Power9, too expensive, power hungry and super miniscule niche. Phoronix benched Power9 recently, lame performance compared to Xeons for the high price and power consumed.

April 27, 2018 | 09:41 AM - Posted by OhTheGrossMarginFallAndResultingShareValueDrop (not verified)

Epyc is here and in that being deployed stage now for some rather large cloud server intersts that have done their Epyc vetting and are making orders as I post. 2H of 2018 is where the Epyc/Server Revenue trends and market share potential will be sussed out as soon as the first quarterly reports close in 2019 and beyond. AMD will be well on their way towards getting back to those Opteron Levels of server market share and beyond. Zen/Epyc is so very close to Intel in IPC, closer than any Opteron/older AMD CPU Micro-Arch ever was for server workloads and Epyc/SP3 MB's with those 8 memory channels per socket and 128 PCIe lanes standard at no extra charge is going to totally fustigate Intel's high margins in the server market.

Epyc's TCO and Price/Performance is more attractive and that includes the total up front hardware costs for Epyc/SP3 being Half/less than half of what Intel charges. And Theose big cloud services intrests do have quants/actuaries employed because they have to finance their purchases via the bank at the going intrest rates.

And if your up front costs for the hardware is less to begin with then that up front equipment financing is going to be lower, as is the total amortized loan cost over time on a smaller amount needed to up front purchase the hardware. That's figured in to any TCO cost along with power effency and other factors that the big cloud services make use of in such large numbers that every littly cent saved adds up to millons in savings over an entire data center.

Oh the Epyc Margin Fustigation is about to begin and the billions in stock share and market cap value that will evaporate for the big chip iterest that is so accustmed to those high margins these so many many years. The Giant beast's High Margin Diet will have to be pared back and that built up blubber can only last so long when those very cold winds of lower gross margins begin to blow. Gross Margins are Directly related to Share Price and market cap by extention. So much so that they are like conjoined twins shareing the same vital organs those Gross Margins and Share Price sisters.

April 27, 2018 | 09:54 AM - Posted by Bill Murray (not verified)

Epyc is under EESC. But EESC margin was only 14/532*100%=2.63%. Not enough Epyc volume being sold. Still swamped by sales of conslow chips.

April 27, 2018 | 11:09 PM - Posted by IgnoramusMaximusBillTheBumpkin (not verified)

Epyc has just started being deployed in production server workload status after going through the proper amount of client vetting and certification and you are out of your mind, you daft fool!

Epyc sales are included alongside the semi-custom division sales that have gone down it that ususl semi-custom market influnced cyclical fashon. With any further losses in that EESC accounting unit figures actually showing that Epyc sales actually buoyed that seni-custom revenue decrese somewhat and made the EESC accountng less worse than if there where not any Epyc sales revenues incoming at all. So Epyc sales revenues actually tempered any greater losses in that EESC accounting unit that includes Epyc sales and semi-custom/other sales also.

AMD even menitoned that fact in their SEC 10k quarterly income/guidence statment. Really you are all Viceroy Research levels of foamimg at the mouth insane with your diatribes on the matter.

April 28, 2018 | 01:31 AM - Posted by Bill Murray (not verified)

This is the 3rd Q after launch, roughly already 9 months but Epyc not making much headway. Very few specific deployments and trials here and there, while new Xeons still lit up roads everywhere you look. Epyc revenues did keep EESC afloat from making another loss but EESC margins clearly shows ultra low margin conslow chips continue to dominate.

April 28, 2018 | 09:54 AM - Posted by BlindBillTheIntelShillDoesNotSeeTheMarginsAsWallStreetDoes (not verified)

Epyc is being deployed for production workloads and those Zen/Zeppelin Wafer/Dies have been comeing off the diffusion at GF's chip fabs with greater than 80% Dies/Wafer yields. So AMD's locked stocked and loaded to meet that Epyc demand with no problems.

What AMD's Opteron did to Intel's fat gross margins expect that Epyc will at a minimum do at least as well and Epyc/SP3-MBs are a bit more performant and feature packed than Opteron/MBs ever were.

The only "conslow" thing around here is your single enfeebled cell of grey floating in a vast ocean of lipids that exists between your large jug ears under that thick layer of bone.

April 28, 2018 | 03:25 PM - Posted by Bill Murray (not verified)

Yeap, you are squirming in rage alrite. Your hero has gone to the dark side and joined the Sith lords. There can only be two! The other was Darth Raja. Soon AMD will only be known for its conslow chips in Sony PS and Microsoft Xboxes. Enjoy writhing around in your own sea of vomit consisting of incoherent ramblings.

April 30, 2018 | 09:17 PM - Posted by 2901bitslice

It was reported in the Wednesday Conference call that EESC Revenues took a hit from a downturn in SemiCustom returns that are affected from a cyclical downturn in the Gaming Console Business.

In terms of Epyc adoption we are still very early in the game. Customers are still testing and evaluating and most are waiting for the 7nm products which will feature substantial performance increase over the current generation of Epyc. Epyc 2 (Rome) will be released at year end. We should see a modest uptick in Epyc Sales that will explode in 2019, I would expect AMD to be booking 50% of all new Datacenter orders by the start of Q3 2019.

April 26, 2018 | 09:50 AM - Posted by JohnGR

AMD fixed... AMD, so now it sends (ex AMD) people at Intel to fix... Intel.

April 26, 2018 | 10:52 AM - Posted by KellerWillBattlePower9RajaWillBattleVolta (not verified)

The broken part of Intel is not any broken part that Keller/Raja are there to deal with. The really broken part with respect to Intel is the broken US government part that never dealt Intel's illegal market practices properly in the first place. And that was ever allowing Intel to buy its way into the third party x86 based Independent third party OEM PC/laptop market. Ditto for Nvidia and that GPP, that ironically even affects Intel in the very same manner that Intel has affected AMD to the detriment of the free and fair market place.

The Third Party x86 based OEM PC/laptop market is too supplier dependent to ever be heathy and properly competative. And any market that becomes dependent on any parts supplier to the detriment of the Indipendend Third Party PC/laptop OEMs is a sick market that needs government intervention.

I hope you all have enjoyed that Terrible Intel graphics that where forced onto the larger PC/Laptop market for over a decade, a period of time when Intel had plenty of money to hire Keller, Raja, and their dogs, to get Graphics/Other things better. And Really it was easer for Intel's lazy management to buy Intel's way into OEM PC/Laptop design wins in the OEM consumer PC/Laptop market than actually make an effort at producing better graphics, a longer term task with no short term managerial profit perks!

Raja and Keller are not there for any direct consumer market assistance for Intel! Raja and Keller are there to help Intel compete with Power9/Nvidia Volta!

April 26, 2018 | 06:47 PM - Posted by Bill Murray (not verified)

With Raja and Keller on board then AMD will kaput much sooner than Nvidia, guranteed. AMD will get relegated to console chip maker with bottom margins.

April 26, 2018 | 08:04 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

Are you with Viceroy Research? It sure sounds like that special form of shorting madness that no one took seriously! You must have lost millions on that little scheme!

April 26, 2018 | 08:19 PM - Posted by Bill Murray (not verified)

Yet AMD had to patch the flaws in the end. Still desperate defending to no avail.

April 26, 2018 | 08:32 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

AMD/chipset partners fixed that even though they only had 24 hours notice instead of the usual months notice from that scam outfit "Lab" for hatchet job hire!

Hah Ha You are the Dictionary Definition of a Daft Fool.

Just look at that high forhead of yours! Even the Hubble Space Telescope can not see far enough to pick out just where that forhead ends!

April 26, 2018 | 08:42 PM - Posted by Bill Murray (not verified)

Flaws are real which is why AMD had to to patch them, you are just still in denial. Flaws originally from early ASMedia chipsets but carried on into AMD chipsets when AMD gave ASMedia contract to make the chipsets. Denial is a river in Egypt.

April 26, 2018 | 08:56 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

Fixed those issues were! And you are just making yourself look more the product of incest going all the way back through the generations to the stone age!

Denial is what you do when you repeat the same nonsense over and over and fail everytime only to do that very same all over again.

Now your SWIFE is in the mood to ad more childranewts to your ever expanding brood! Just look at them critters scamper about!

April 27, 2018 | 12:31 AM - Posted by Bill Murray (not verified)

All AMD chipsets made by ASMedia are flawed. Same flaws that company discovered before AMD started using ASMedia. They found the same flaws with AMD chipsets made by ASMedia. They had to wait until flaws are verified on all AMD chipsets. Especially the Epyc ones which are critical. Too bad Epyc was late and trickling out in very low volumes, so with no early access to Epyc they had to wait until they could get one to test and verify. Late to reveal this is the reason.

April 27, 2018 | 04:22 AM - Posted by Johan Steyn (not verified)

Yip, this guy is part of the Viceroy/Intel rubbish. He clearly knows nothing about what these bugs were about. Strangely enough, he is silent about the seriousness of Meltdown, which Intel managed to evade as well. Intel made rubbish CPU's when it comes to security and now he wants to talk about AMD's few small issues. There is so much ignorance and blind loyalty to an evil comapany.

April 27, 2018 | 05:58 AM - Posted by Bill Murray (not verified)

Flaws were verified by other security experts https://www.bleepingcomputer.com/news/security/researchers-who-found-amd... Already earlier known PSP also had flaws before this disclosure http://seclists.org/fulldisclosure/2018/Jan/12 So Keller gone evil, joins Intel. Raja also now in dark side. Human sacrifice, dogs and cats living together... mass hysteria!

April 27, 2018 | 09:51 AM - Posted by OhTheGrossMarginFallAndResultingShareValueDrop (not verified)

No it's more you and your SWIFE(Sister wife) having more of those childranewts that's harbinger of the apocalypse! And all the way back in your family history of Husbands and SWIFEs and that all that de-evolving that's been going on.

Intel's high margins are heading south there is no stopping the gross margin disaster that's about to befall ChipFatza, no matter the engineering talent on board!

It's an Epyc storm of lower margin competition that's just getting started!

April 27, 2018 | 09:57 AM - Posted by Bill Murray (not verified)

Epyc storm never materialized. EESC margin still dominated by conslow chips. Reality strikes down your day dreams.

April 27, 2018 | 11:15 PM - Posted by IgnoramusMaximusBillTheBumpkin (not verified)

You are at your wits end with that fuzzy logic you total moron! Server rollouts take several quarters to even begin in earnest after a proper server market longer period of client vetteing before any new CPU product is rated ready for server producton workloads.

You are a most egregious fool, and a daft one at that!

April 28, 2018 | 01:12 AM - Posted by Bill Murray (not verified)

Results are results. Everyone see that except yourself still full of incoherent ramblings. SkylakeX took to skies right after launch, while Epyc still limping along the sewer underworld despite having a head start.

April 28, 2018 | 10:00 AM - Posted by BlindBillTheIntelShillDoesNotSeeTheMarginsAsWallStreetDoes (not verified)

Intel's high gross margins are in the Skies currently, watch those High Gross margins take a little dive with that ground coming up rather quickly as Epyc begins to take off.

The Fatter they are the bigger that Fat Splat when they hit the ground at high speed. Lipids Lipids everywhere!

April 28, 2018 | 03:19 PM - Posted by Bill Murray (not verified)

Dear dickless peck, some quotes from Anandtech "Intel is always a company built on strong margins, and although they were down 1.3% from last year, at 60.6% they are still quite strong."

April 26, 2018 | 09:52 AM - Posted by pohzzer (not verified)

Jim Keller was brought to Intel to implement an AMD style modular/scalable design.

Until that is implemented Intel cannot compete on cost/performance all other considerations aside.

April 26, 2018 | 10:29 AM - Posted by KellerWillBattlePower9RajaWillBattleVolta (not verified)

Intel has the engineering chops to do that without Keller's help. Keller and Raja are there to stop OpenPower/IBM paired with OpenPower/Nvidia because really Power9 beats the x86 designs in the server/HPC markets and that NVLink with Volta accelerators is a bigger threat to Intel than little old AMD's x86.

AMD had damn well better keep Keller's K12 custom ARM work handy as the ARM maket is totally not under any Intel Influnce and never will be. Lisa Su be advised that K12/Vega graphics may just net more revenue potential from new markets compared to any x86 based CPU cores than can never beat any RISC ISA cores in low power usage metrics.

AMD lacks the funds to buy its way into the larger x86 based PC/Laptop market wit funded design wins like Intel Can, and buy its way in to those wins Intel really does.

Keller/Raja have bigger fish to fry for Intel and that's not so much consumer oriented as it is professional oriented designs to compete with Power9/Nvidia and that growing OpenPower threat that's not x86 based.

April 26, 2018 | 06:49 PM - Posted by Bill Murray (not verified)

K12 is ded long ago. No one else competes with IBM's Power9 because their niche market too miniscule.

April 26, 2018 | 08:16 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

Power9 via IBM/Nvidia and OpenPower sure got some good Government Supercompute design wins and Intel is in such fear of Volta in the server room that Raja was hired out from under AMD PDQ!

Intel's high margin days are numbered and once Wall Street sees Intel's margins going south that Market Cap will follow right along downwards.

K12's blueprints are still available for AMD's semi-custom folks to job out for some customer's needs if that will bring AMD some return on its K12 investment. So Maybe a gaming tablet client that wants to compete with the Nvidia based Nintendo Switch.

You are looking a little green around your actual gills and those webbed hands and feet are looking a little dry. So maybe it's time for you to crawl back under that rock under that green pond scub and hydrate your de-evolved self for a bit!

April 26, 2018 | 08:27 PM - Posted by Bill Murray (not verified)

K12 still ded, nobody wants it. In its place, Qualcomm Centriq and Cavium ThunderX gets customers. Many big supercomputer contracts by government and big institutions still under Intel's belt. IBM only a very few, only one government can afford them right now. Thats why IBM niche super miniscule I said.

April 26, 2018 | 08:47 PM - Posted by LipidsOfFattyWhiteAndNotGreyCellsToBeFound (not verified)

Not really Pea For Brains under loads of Lipids.

OpenPower has more growth and than Just IBM's/Nvidia's OpenPower Power9 supercompute busines and jusy you wau and see.

K12's blueprints remain for AMD's use and things change while Intel's has to suck it up and lower its fat margins no matter what as Epyc based system take market share.

x86 is not the only ISA game in town and Intel's charging too much relative to what AMD charges for its better performing Epyc platform standard options with 128 PCIe lanes and 8 or 16 full memory channels on the Epyc/SP3 1P and 2P MB SKUs.

Only One govnernment can afford my a$$ with Google, Facebook and others looking at power9/Nvidia Volta and for AMD that project 47 pataflops supercomputer in a single cabinet that takes 20 Epyc 32 core CPU/MB SKUs and 80 Vega 10 base die based Radeon Instinct MI25 per cabinet.

What Intel chargse for its Product will have to come down in price and Intel high Margins are in the cross hairs with Wall Street watching. So Long billions in market cap when those high margins swing low for Chipzilla!

April 27, 2018 | 12:39 AM - Posted by Bill Murray (not verified)

K12 is ded and buried six feet under for good. Thats why AMD let Qualcomm and Cavium rule in ARM servers. Radeon market share more likely less than 0.1% for datacenter. Nvidia dominates all. Q1 2018 earnings don't paint good picture for Epyc, still not enough. Anandtech states "The revenue decline is attributed to lower semi-custom revenue, but somewhat offset by higher server and embedded revenue. EPYC processor revenue has helped, but not enough to offset the loss of revenue in semi-custom."

April 27, 2018 | 11:28 PM - Posted by IgnoramusMaximusBillTheBumpkin (not verified)

K12's blueprints and IP are there for AMD to make use of if needed and Epyc has only now been certified for production ready workload status by the server clents. So 3 quarters is not a long time at all by server market rollout timeline standards before Epyc can even start to fully begin generating the reverues to make semi-custom sales a non factor for AMD's total health.

Anandtech was quoting what AMD said in their quarterly 10K income/guidence statment so it's not what Anantech stateed it's what Anandtech quoted and that's good news for Epyc even with AMD's semi-custom sales revenues going down, that's gaming console influnced with semi-custom's sales. Epyc sales helped make those losse less worse so that's good news as far as Epyc's generated revenues are concerned.

But you are paid to spin negative you Viceroy Research Minion!

April 28, 2018 | 01:17 AM - Posted by Bill Murray (not verified)

Didn't you see the tombstone with "R.I.P K12" written on it? AMD also abandoned Seattle. Both missing from all future roadmaps. Conslow low margin and sometimes loss making sales still dominates EESC. Epyc could not raise margins because just too little sold.

April 28, 2018 | 10:17 AM - Posted by BlindBillTheIntelShillDoesNotSeeTheMarginsAsWallStreetDoes (not verified)

CPU designs never die they just get filed away!

It's not AMD that has to worry about margins as they are only going to rise. It's Intel's Fat Gross Margins that will be taking a dive and along with that billions in share holder value. AMD's Epyc and Radeon Pro WX/instinct markups are relatively high in relation to AMD's consumer markups! So as the Epyc and Radeon Pro WX/Instinct sales volumes increas they will become a higher percentage of AMD's total unit sales and then by default AMD's gross margins will go up.

Intel's fat margins will have to go down in order for Intel to attempt to keep AMD's server market share low there is no denying that fact. Epyc and Radeon Pro WX/Instinct will become AMD's largest sources of unit sales and revenues in short order and that comes with a professional market markup unlike AMD's consumer SKUs that are lower markup SKUs.

AMD's Epyc SKUs are lower cost to acquire than Intel's pricy server kit and that fact will give Intel's High Gross Margin Metrics some terrible fits!

April 28, 2018 | 03:38 PM - Posted by Bill Murray (not verified)

Summit supercomputer chose Nvidia's Volta not AMD's Radeon. Same with majority of supercomputers, Nvidia dominates. Intel supercomputer market share still undisputed, relegating everyone else including IBM into miniscule niches. As for Epyc's insignificant showing didn't even do much at all. Intel still maintained around 60% gross margins as usual. Let me requote Anandtech's findings "Intel is always a company built on strong margins, and although they were down 1.3% from last year, at 60.6% they are still quite strong."

April 26, 2018 | 01:01 PM - Posted by Anonymouss (not verified)

Josh, add a link to the newsroom article

April 26, 2018 | 01:09 PM - Posted by Josh Walrath

Will do. Thanks for the head's up.

April 26, 2018 | 06:13 PM - Posted by Subsailor

Josh- write more articles please. Nobody can touch your CPU related knowledge and ability to explain to us.

April 27, 2018 | 01:46 PM - Posted by James

Looking to spec out some Xeon machines makes intel’s prices look absolutely ridiculous. Intel has to make a very expensive chip to compete with Epyc. The large, low yield die are significantly more expensive. They also need to route a huge number of signals to a single chip. Six channel memory coming from one die is a routing nightmare. There is also the PCI-express. AMD can scale that to huge numbers easily with the distributed system. Intel doesn’t look competitive at all even with current 32 core / 64 thread Epyc. AMD will scale up to 64 core / 128 threads per socket with Zen2 making Intel parts look even more ridiculously underpowered and overpriced.

Intel will need to deal with that soon, otherwise AMD could grab a large segment of the server market. It is too bad so many people who make the purchasing decisions are essentially intel can boys. I suspect Keller’s first task will be in the server market, not the mobile market. Intel is going to face huge competition from ARM in the mobile space. That really looks like a losing battle for them. When they think that ARM processors can’t scale up, they are forgetting that Intel’s core architecture original came out of developement for a mobile chip; the Pentium 4 was the high powered desktop chip that got out performed by a mobile design.

The thing is, the distributed systems like Epyc are much cheaper overall. This means much lower selling price and much lower margins than Intel’s current monolithic die CPUs. Intel stands to take an ASP hit either way. The meltdown vulnerability hit at a bad time for intel also. Cloud providers must install the patch and take the performance hit. Epyc isn’t vulnerable to that, so it looks even better in that context. This also gives cloud providers a reason to diversify the machines that they have. They don’t want an intel bug to shutdown their entire business. Even if Intel comes out with a competitor soon, cloud providers still have a reason to maintain diversity that isn’t just price. Intel doesn’t seem to have adjusted down Xeon prices at all; even though their current parts are mostly not competitive. There is a lot of inertia in the server world, so they can probably maintain their ridiculous prices for quite some time. It is in the consumers best interest to have more than one cpu provider though. With the way things are going, if it isn’t AMD that provides a check on Intel, it will be ARM.

May 5, 2018 | 12:20 AM - Posted by Veer (not verified)

If anything lets hope that Keller and Raja are legally unable to reproduce work done for AMD for Intel's benefit.

Thats counterproductive for the industry.

Honestly, the time Raja spent at AMD, there were few to no GPU wins.

I honestly think Raja and Keller are doing it for some obscenely huge paycheck that no one could pass up.

As someone noted, Keller played a small part on Zen's design and he may just be at Intel to fix their gaping security problems.

Raja, bless his soul, will have to reinvent the wheel at intel for their discrete gpu plans that already failed once before.

I hope Lisa's new GPU team can do more than Raja was able too and their driver team will be more responsive to fixing bugs. The new Radeon rewrite was just full of problems still is.

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

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