Eric Demers leaves for Intel after 14 years at Qualcomm — father of Radeon and Adreno GPUs now sits at Lip-Bu Tan's table

News
By published

GPU design genius switches teams to build AI accelerators for Intel

Intel logo and Gaudi chip
(Image credit: Getty Images)

The talk of the town in the world of AI right now is almost exclusively about the amounts of money shifting from one company to the next, but the latest event might have farther-reaching effects than most business deals. Eric Demers, who designed ATI's best GPUs and spearheaded almost all of Qualcomm's Adreno designs, has now joined Intel's GPU team "with a focus on AI."

The blue team's GPU efforts are all but guaranteed to be significantly bolstered by Demers, a particularly welcome development in these troubled times for the company. According to Moor Insights and Strategy, this move is "bigger than people realize", as "[Demers] is an executive, but also he is a GPU architect, of which there are not that many that are at the level that he is at because he can basically build a GPU architecture from the ground up.”

Google Preferred Source

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Bruno Ferreira
Bruno Ferreira
Contributor

Bruno Ferreira is a contributing writer for Tom's Hardware. He has decades of experience with PC hardware and assorted sundries, alongside a career as a developer. He's obsessed with detail and has a tendency to ramble on the topics he loves. When not doing that, he's usually playing games, or at live music shows and festivals.

6 Comments Comment from the forums
  • bit_user
    The article said:
    The blue team's GPU efforts are all but guaranteed to be significantly bolstered by Demers
    He said he's going to Intel to work on "AI", not "GPUs".

    I'm just pointing that out, because Intel GPUs face a rather uncertain future and this news really doesn't change my point of view on that.

    Keep in mind that Nvidia's Bx00-series Blackwells and AMD's MI-series processors don't have any of the hardware acceleration needed for graphics. They are purpose-built AI/HPC accelerators. Hence, Intel building AI chips does not necessarily mean they're investing in their dGPU product lines.

    The article said:
    When AMD absorbed ATI, Demers became the company's graphics Chief Technical Officer, a position he held until 2012 when he joined Qualcomm
    Oh, weird. I'd have assumed he went to Qualcomm years earlier, when they bought AMD's mobile GPU business.
    Reply
  • SirStephenH
    bit_user said:
    He said he's going to Intel to work on "AI", not "GPUs".

    I'm just pointing that out, because Intel GPUs face a rather uncertain future and this news really doesn't change my point of view on that.

    Keep in mind that Nvidia's Bx00-series Blackwells and AMD's MI-series processors don't have any of the hardware acceleration needed for graphics. They are purpose-built AI/HPC accelerators. Hence, Intel building AI chips does not necessarily mean they're investing in their dGPU product lines.


    Oh, weird. I'd have assumed he went to Qualcomm years earlier, when they bought AMD's mobile GPU business.
    AI work is typically done by GPUs. The article makes it pretty clear that he will be working on commercial-grade GPUs for AI, not consumer-grade GPUs that would typically be used for gaming.

    Here's just one example.
    the reports so far predictably indicate that Demers will be in charge of designing AI accelerator GPUs
    Reply
  • thisisaname
    bit_user said:
    He said he's going to Intel to work on "AI", not "GPUs".

    I'm just pointing that out, because Intel GPUs face a rather uncertain future and this news really doesn't change my point of view on that.

    Keep in mind that Nvidia's Bx00-series Blackwells and AMD's MI-series processors don't have any of the hardware acceleration needed for graphics. They are purpose-built AI/HPC accelerators. Hence, Intel building AI chips does not necessarily mean they're investing in their dGPU product lines.


    Oh, weird. I'd have assumed he went to Qualcomm years earlier, when they bought AMD's mobile GPU business.
    I often wonder what they bring, anything they have developed is owned by another company. So they would have to come up with something total new, which would be very hard
    Reply
  • Pierce2623
    SirStephenH said:
    AI work is typically done by GPUs. The article makes it pretty clear that he will be working on commercial-grade GPUs for AI, not consumer-grade GPUs that would typically be used for gaming.

    Here's just one example.
    I don’t think you’re getting the point that those data center AI “GPUs” aren’t actually GPUs because they don’t have ROPs. They can do all the math but they can’t actually render an image.
    Reply
  • bit_user
    SirStephenH said:
    AI work is typically done by GPUs.
    I'm not sure where the disconnect is. I thought I made it pretty clear what distinction I was drawing. Nvidia's B200 is not a GPU in the same sense as what you've got in your PC - it lacks tessellation engines, ROPs, TMUs, RT cores, and a display engine. The things they have in common are basically just the ISA, Tensor cores, and a variety of other IP that holds it all together. And that much is mostly just because Nvidia backed its way into building AI chips. It really shouldn't even be called a "GPU", because it contains nothing to do with any "G"!

    When you look at purpose-built AI chips, they tend to deviate even more from GPUs, often preferring DSP-like VLIW cores over the narrow-issue + SMT approach used by GPUs. Intel's Gaudi 3 was detailed at Hot Chips '24 and uses 4-way VLIW cores, with no mention of SMT:
    https://hc2024.hotchips.org/assets/program/conference/day1/60_HC2024.Intel.RomanKaplan.Gaudi3-0826.pdf
    SirStephenH said:
    The article makes it pretty clear that he will be working on commercial-grade GPUs for AI, not consumer-grade GPUs that would typically be used for gaming.
    It sure sounded to me like the author was reaching to connect Demers' prospective work with Intel's broader GPU products, with which I was taking issue. You can see the quote I used. If you interpreted differently, then so be it.

    That said, I do hope Intel stays in the dGPU market. If Demers turned out to have some role in that, I think he would be an asset.
    Reply
  • bit_user
    thisisaname said:
    I often wonder what they bring, anything they have developed is owned by another company.
    Well, he's covered by non-disclosure agreements that he signed with Qualcomm. Anything of Qualcomm's that's trade secret, he could get into big trouble for disclosing or utilizing. Theoretically, he could get Intel to license some of Qualcomm's patents that he's aware of, but Qualcomm probably wouldn't want to play ball, unless they already have some kind of patent cross-licensing agreement established.

    thisisaname said:
    So they would have to come up with something total new, which would be very hard
    He'll be in an executive position. So, he can just hire people who know their stuff and let them do the actual design & implementation work.

    Also, while Qualcomm builds DSPs (Hexagon) and has its own little line of server-based AI inferencing cards, they really aren't established in the market for training hardware. To the extent he's only focusing on AI, there should be a lot less risk of him violating his confidentiality agreements with Qualcomm.
    Reply