Alongside today’s profitable-but-uneasy earnings report from Intel, the company’s earnings presentation also offered a short update on the status of their discrete GPUs. As of today, Intel’s DG1 GPU is now shipping. Meanwhile the company announced their next GPU, appropriately named DG2, which is based on their upcoming Xe-HPG architecture. This GPU is now back from the fab and is in Intel’s lab, and is now far enough along to have been powered on.

First and foremost we have DG1, or as it’s better known by its commercial product name, Iris Xe Max. Intel’s first discrete GPU in over two decades, the company has since the beginning of this year been touting it as a companion to their Tiger Lake CPUs, pitching it as an upgraded graphics option for thin & light notebooks, and a successor of sorts to Intel’s GT3e and GT4e iGPU configurations from past generations. Until recently, we weren’t quite sure when it would show up in commercial products, but recent OEM notebook reveals along with Intel’s earnings announcement are now confirming that the GPU is shipping to OEMs. According to Intel, DG1-equipped notebooks are expected later in Q4. In the meantime, there are still scant few details on DG1 itself, such as expected performance and power consumption; so hopefully Intel will be getting ahead of its OEM partners on this one to set some expectations.

Meanwhile, today’s notes also announce for the very first time the next discrete GPU to come out of Intel, DG2. While obviously still some time off, Intel has completed tape-out and fabbing of the initial alpha silicon, with the company reporting that they’ve powered-on the GPU in their labs.

Somewhat surprisingly, CEO Bob Swan has also confirmed that this isn’t just a DG1 successor, but instead is a higher performing GPU based on the company’s forthcoming Xe-HPG(aming) architecture. First revealed this summer, Xe-HPG is Intel’s enthusiast/gamer-focused architecture, incorporating marquee features found in similar dGPUs like ray tracing. It’s also being manufactured completely external of Intel; while the company hasn’t said which fab and process node is being used, it’s none of Intel’s nodes. So this is the first major piece of external fabbed silicon that we know of to be up and running at Intel.

But like all teasers/financial disclosures, Intel isn’t saying too much more at this time. Nothing new was revealed about the Xe-HPG architecture, and Intel hasn’t clarified whether DG2 is a big, flagship-grade chip, or a more modest, high-volume part. For now, the company is simply saying that DG2 will “take our discrete graphics capability up the stack into the enthusiast segment.”

Source: Intel

POST A COMMENT

50 Comments

View All Comments

  • lorribot - Saturday, October 24, 2020 - link

    Don't forget LinTel, but you can also get LinArm, LinPPC, LinSparc (among many others), WinArm and MacTel and MacArm.

    For those that like their OSes a bit more exotic, VMS is moving from IA64 (Intel's only true 64 bit processor) to x86 so that VMSTel to err... VMSTel having been VMSVAC and VMSAlpha (there was also WinAlpha and WinIA64 too).
    There are always choices you just have to look for them.
    Reply
  • JayNor - Friday, October 23, 2020 - link

    Intel is also in production of SG1, which is a 4 tile version of DG1 aimed at servers.
    DG1 and SG1 are both based on the Xe-LP GPU in Tiger Lake.
    Reply
  • valinor89 - Friday, October 23, 2020 - link

    If one of the tiles says "Indeed" every so often I'm sold! Reply
  • pixieluc - Friday, October 23, 2020 - link

    Tek'ma'te, Brother Reply
  • wolfesteinabhi - Friday, October 23, 2020 - link

    Intel will easily replace the MX250/350 kind of high volume chips with this ... its easier for OEM's to buy complete chipset from Intel and some good discounts perhaps. Reply
  • bill.rookard - Friday, October 23, 2020 - link

    Agreed. Much easier for an OEM to click an extra box on the order sheet from Intel for something with a higher power integrated graphics than to coordinate shipments between two manufacturers... Reply
  • TheinsanegamerN - Friday, October 23, 2020 - link

    Maybe. The ryzen 4800u and 4700u already perform at MX 350 level, and intel's new Xe enables 11th gen chips will be in the same ballpark. I'd see those as far better replacements, yet the MX150/250/350 live on. Reply
  • wr3zzz - Friday, October 23, 2020 - link

    If the rumor that discrete Xe can work together with iGPU Xe were true then Intel has a pretty nice in to sell some discrete chips. Reply
  • tipoo - Friday, October 23, 2020 - link

    They'd have done better than 16 years of AMD and Nvidia trying and failing out of multi GPU scaling for games, if they did achieve this. So I'm kind of doubting it.

    My OpenCL loads can target multiple GPUs just fine though
    Reply
  • JayNor - Friday, October 23, 2020 - link

    Intel demoed 4 tile scalability for Xe-HP, but that is using something other than pcie interconnect between tiles according to David Blythe's Hotchips 2020 presentation.

    I wouldn't be surprised to see decent use of both GPUs by oneAPI, but I would be surprised to see the linear per tile performance increases that were demonstrated for the 4 tile Xe-HP.
    Reply

Log in

Don't have an account? Sign up now