news amd shakeup combines cpu gpu businesses
News

AMD shakeup combines CPU, GPU businesses

Ever since AMD’s purchase of ATI closed in October 2006, the former ATI has operated as AMD’s Graphics Product Group—a business separate from the microprocessor development operations. Well, AMD is shaking things up. The company has announced a new operating model, as part of which its graphics and processor businesses have merged into a single products group.

Here’s how the company’s internal structure looks now:

  • A products group led by Rick Bergman, 45: This new group is responsible for delivering all of AMD’s platforms and products and aligning the graphics and microprocessor product development groups into a single unified organization. Jeff VerHeul, 50, will head the Processor Solutions Engineering team, to deliver AMD’s platform silicon and improve time-to-market and innovation for near-term roadmaps.
  • An Advanced Technology Group, led by Chekib Akrout, 51: This new group will focus on developing AMD’s future technology innovation.
  • A marketing group, led by Nigel Dessau, 44: The singular marketing group will drive cohesive and consistent external messaging across all of AMD’s products and platforms.
  • A customer group, led by Emilio Ghilardi, 51: The sales organization is responsible for expanding AMD’s customer relationships globally.
  • Rick Bergman was previously the Senior VP of AMD’s Graphics Product Group, while Randy Allen headed the Computing Solutions Group. However, the same press release says Allen has chosen to depart the company, leaving Bergman in charge of the new, unified products group.

    AMD goes on to say the reorganization will help it “further integrate” its x86 microprocessor and graphics technologies. Not much integration has happened on that front yet, despite AMD promising CPU-GPU hybrids for “late 2008/early 2009” right after the ATI buyout. Those hybrids have since been postponed until 2011.

    0 responses to “AMD shakeup combines CPU, GPU businesses

    1. IMO the AMD IGPs get praised too much based on reviews.

      Their 3D capabilities are really iffy, and I mean that in a way that’s beyond just the limited GPU core’s spec sheet. I have been running 780G for over a year now and have found it to be a rather inconsistent performer due to severe RAM/HT bandwidth limitations. It is really, really bad with alpha texture overdraw (sucks up fillrate/bandwidth). IMO there isn’t an IGP out there that’s worth considering for any level of 3D gaming when there are ~$50 cards like Radeon 3850 to be had.

      There’s also the tiny issue of a never-solved Cool’n’Quiet issue in XP. If you play a video, CnQ gets disabled until you go into the Windows power management panel and reset the profile. CnQ can also cause bad 3D performance issues because the RAM bandwidth is impacted dramatically by the CPU clock.

      For 2D and GUI, an Intel IGP is almost as good. Even the crusty 945G can run Vista fine (first hand experience). AMD/NV’s advantage here is with their awesome control panels that offer superior configuration options for color correction, etc.

      However, if you want hardware AVC/VC-1, it is great to be able to buy an IGP that does it. You get the power savings of running just an IGP and lose the need to put more money into a discrete video card. No denying how nice that capability is in this day and age.

    2. Wow, I am honestly impressed by that accomplishment. Love or hate ATI, they’ve been more competitive in the last 2-3 quarters than they have been in a long time in the desktop discrete arena.

      Anyone think it is due to X54 supporting SLI?

    3. after enough re-orgs, there will be nobody left for the managers to manage or workers to produce widgets. It’s a big pyramid scheme, like Amway.

    4. No looking at the stock price of the company I am talking about it hasn’t been effective for them either…

      nice try

    5. AMD has never done well except when Intel makes mistakes the reasons abundantly clear and the weaknesses ready to be exploited whenever Intel stops making mistakes.

      AMD’s P2 is a great product and ATI’s 4890 is tip top in it’s price segment…. they aren’t doing anything wrong atm except remaining guilty of what they have always been guilty of when your company is 1/10th the size of the competition and lacking in resources to compete.

      the mentioned reduction in gpu share has little to do with desktop sales and highlights how easily the numbers can be misleading.

      I’m not confident that this merger of CPU / GPU will work… if anything is sacrificed then it’s bound to fail at the cost of one or the other taking down the whole.

    6. Due to integrated an mobile. The desktop couldn’t possibly be doing poorly.

    7. Like #26, I think the main problem was timing. If they’d have bought when the stock was $40, that would have been an enormous help.

      The business concept seems sound to me though. Become a one stop supplier, develop platforms, leverage one thing when the other ain’t doing so great. One of the saving graces for AMD before the P2 (yeah, I said P2) was that the 690g, 780g, 790g chipsets were better than anything available on the Intel side, as far as integrated graphics goes.

    8. Amen, brother. Silly geeks, pedantry is for… um, something something.

    9. If they had waited that long, we might not have had Radeon 4000s turn out so well, and ATI may never have re-established itself against Nvidia. And there may not have been “Fusion” processors in the works at all. Look how long it’s taking as it is.

      They have made and continue to make a lot of mistakes, but they’ve fallen behind enough. Sitting around and waiting any more wouldn’t have helped them in the long run.

    10. Or too brain damaged to realize that people probably aren’t talking about Pentium 2 in an AMD-related discussion.

      Its 2009. Nobody gives a shit about Pentium 2, especially not AMD. Using PII/P2 to refer to Phenom 2 is just fine to everyone with more than a handful of brain cells.

    11. I think the problem was just with timing. If they had just waited until after the 2900 had come out they would have saved a lot of money.

    12. It is fortunate, then, that the “high-end” graphics market has become nearly irrelevant.

    13. Doing well isn’t just selling more products. I could setup a bakery and sell bread for nothing and it would look like I was doing well from watching the amount of people walking in and out of my shop. Shame that I’m not making any money.

      I bet AMD is hurting bad on the PII’s

    14. Why? They had a SSE engine on their CPU’s so they know how to achieve throughput.

      A CPU is far more complicated than a GPU, I see no reason why they would not be able to develop their own GPU anyway. I’d say the most difficult part of a GPU is the software/driver side, especially with regard to IGP graphics.

      Also by the time IGP graphics are mainstream SOC devices, a basic CPU SSE engine will probably be more than powerful enough to render the UI..etc in software, not that that is an elegant solution.

    15. Really? Intel’s integrating nivida or s3 or ati graphics cores into their upcoming cpu+gpu chip? I thought they were using their crappy existing integrated graphics. And if they licensed tech, why did they develop Larabee?

      What am I missing?

    16. The first rule of re-orgs: every manager has at least one re-org in him; every re-org produces at least one additional manager.

      You will note this describes an unbounded process.

    17. Actually, it looks like the ATI guys are increasingly in charge. This could end up being like the Boeing-McDonnell Douglas merger, where Boeing was the surviving corporation, but it was mostly ex-MDD managers who ended up running the company.

    18. Their Server Technology broadcast showed a graph which, if you play with the numbers, shows a 35% improvement per core* over Deneb in both integer and FP. Core-for-core it would beat the fastest Nehalems today, and assuming an 8-core chip it would beat an 8-core 32nm Nehalem.

      The problem is that it’s 2011 and will therefore face Sandy Bridge launching late 2010. How it will match up to that is unknown.

    19. They could have done a simple cross license agreement with ATI, NVIDIA or even S3/VIA.

      That would have saved them billions.

      That’s pretty much what Intel did.

    20. In the long term it would have been death, with no in-house graphics technology for integration into the chips in the next decade. Of course if AMD had waited another year or two they could have got ATI cheaper.

      AMD went a bit awry in 2005, got bitten by Core 2 Duo (when they could have MCMd a quad core to compete) fell behind, had a dud release, and it looks like it will only be this year that they’ve sorted themselves out, with some products coming earlier than expected.

    21. Whatever. Just get Bulldozer right and out the door on schedule. No TLB crap, no undersized cache crap, good performance, good power consumption.

    22. Buying ATI is probably one of the biggest mistakes in computer history (after buying AOL that is).

      We are starting to see AMD cutting off parts (like the fabs) in order to stay alive.

      I think this move will eventually take their graphics division out of the high end market all together.

    23. So, the ATI brand will disappear now? or is this just a behind the scenes thing like with Activision-Vivendi.

    24. Shuffling the chairs on the deck of the titanic?

      I’m not sure how many reorgs the company I have been at has gone through over the last 10 years but I can recall one of them that actually was effective.

    25. Agreed, Ph2 is much more acceptable for those truly too lazy to write Phenom2.

    26. I think the point was that they’re still in the red. Having good products, but sacrificing profit margins (and thus their funds for next products), to get market share, isn’t really working for AMD. They lost GPU market share and even though they did regain some in the CPU market, they did it in the same quarter that Intel released their Nehalem based Xeons. That doesn’t bode well for their market share in the next quarters.

    27. Posh, a Pentium 2 will never hold its own against a Core 2.

      Plus, Pentium 2’s aren’t even made by AMD. What did you have for breakfast?

    28. Except all else isn’t failing right now. Radeons are doing well. PIIs are doing well against C2s.

      My guess is that this has been planned for a while, and they just wanted to wait until the pre-buyout projects on both sides were finished before changing things.

    29. Letting the graphics people take over the running of the CPU division is a good idea. The merger (despite the cost) was the best thing AMD has done since K8. They got great chipsets and the engineering talent that will increasingly be needed as CPU and GPU functions converge.

    30. Interesting, I hope this doesn’t fuck up what they’ve got going on with the Radeon’s right now.