Intel's Talks Next-Gen GPUs: Xe2-HPG and Xe2-LPG

Intel Arc
(Image credit: Intel/YouTube)

When Intel announced its Xe family of graphics processors in 2020, it introduced as many as four microarchitectures, which eventually translated into major delays, cancellation of Xe-HP datacenter GPUs and issues with drivers. The company now says it has learned its lesson; with its Xe 2 'Battlemage' family it will offer fewer microarchitectures. Still, there will be Xe2-LPG and Xe2-HPG microarchitectures for different kinds of GPUs. 

"There is a Xe and there is a Xe 2 and in that Xe 2 generation there is a Xe-LPG and there is a HPG (…) and there a slight variations (…) which is our big learning," said Tom Peterson, an Intel Fellow, in an interview with Hardwareluxx. "The idea was we needed to optimize for each segment and build separate chips and do separate verifications. And I think now the real learning is we would be better off concentrating our focus and really thinking of it like a really solidly, hard IP business." 

On its Battlemage generation of GPUs, Intel will stick to Xe2-LPG and Xe2-HPG microarchitectures. It is unclear whether Xe2-HPC is still planned, but it officially Intel's next-generation GPU for HPC codenamed Realto Bridge is based on 'Enhanced Xe-HPC' cores, not on Xe2-HPC cores.  

"Now as we go forward in our roadmap, we realized this is a very, very expensive – the QA process and the segmentation. The Thinking was we needed to differentiate our IP and customize it per each segment," said Peterson. "[…] We are going to just have one thing and it goes everywhere unmodified. That's more the strategy we are looking at going forward. And that's because, that's really the only way to get IP reused to really work." 

Intel's original Xe family included four microarchitectures: Xe-LP for integrated and low-end standalone GPUs, Xe-HPG for discrete desktop graphics cards, Xe-HP for cloud datacenters and Xe-HPC for high-performance computing. Developing vastly different GPU microarchitectures has its advantages when it comes to performance and die sizes. For example, a slightly smaller iGPU that lacks features like ray tracing translates into hundreds of millions of dollars of cost savings as Intel sells boatload of client CPUs. Meanwhile, redesigned Xe-HPC cores translate into performance advantages. 

But to deliver on its promise, Intel needed to design, verify, and produce as many as nine different GPU variations, which is a pretty enormous number for a single product family even for a giant like Intel.  

Eventually, the company had to axe its Xe-HP because it decided that the datacenter vertical could be addressed by Xe-HPG and Xe-HPC GPUs instead. But the company lost precious time developing Xe-HP hardware and software stack. Furthermore, because Xe-LP and Xe-HPG are so different, the company had to tune integrated and standalone GPUs separately and then develop different drivers for built-in and discrete graphics processors. As a result, while Xe-LP iGPUs arrived in time in 2020, Xe-HPG and Xe-HPC GPUs were at least a year late. 

While the company says it has learnt its lessons with the first generation of its Xe graphics processors, it remains to be seen whether its Xe2 Battlemage GPUs will come out in time to compete against the best graphics cards from AMD and Nvidia. 

Anton Shilov
Contributing Writer

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.

  • evdjj3j
    "There is a Xe and there is a Xe 2 and in that Xe 2 generation there is a Xe-LPG and there is a HPG (…) and there a slight variations (…) which is our big learning,"

    That would have probably made a lot more sense if you wouldn't have left the .... out.
    Reply
  • bit_user
    evdjj3j said:
    "There is a Xe and there is a Xe 2 and in that Xe 2 generation there is a Xe-LPG and there is a HPG (…) and there a slight variations (…) which is our big learning,"

    That would have probably made a lot more sense if you wouldn't have left the .... out.
    When you're quoting someone and you want to omit parts of it, you need to indicate where you edited their statement. That's standard journalistic practice.

    The reason for such edits is to clarify or bring out the essence of a particular statement. So, you're basically affirming that.

    Update: upon visiting the original HardwareLuxx article to get more of the context of these quotes, I noticed the "(...)" are actually present in their article.
    Reply
  • cyrusfox
    Another day, another article suggesting Intel dGPU business is safe for now, I hope MLID got this wrong, demise/exit of Intel discrete graphics.

    As battlemage is so far along, maybe it would be more likely for Celestial or Druid for the exit point though. Also what if dGPU does disappear for the mid to low end thanks to direct die attach of future GPUs (Meteor Lake)?

    Glad to see price drop on A750 as well as driver work maturing. I hope they keep all the fixed funciton accelerators on the small gpu die for BattleMage and I am very interested in the rumored half stepping ACM-G12 as the die is too large for ACM-10 to justify a product @ the $250 mark, Good for Intel to clear inventory but I would imagine the Margin is 0 or negative on these products for them.
    Reply
  • bit_user
    Tom Peterson: "... we would have been better off have we rigorously said 'you know what, we gonna give up something', like give up some differentiation in the high end or we are going to have some overhead in the low end. We are gonna just have one thing and it goes everywhere unmodified."Um, good luck with that. Both AMD and Nvidia have decided to have one microarchitecture for datacenter and a separate one for consumer. Granted, they don't have the three variations he mentioned, but I think 2 makes sense.

    One reason it's probably wise to unify iGPU and consumer dGPU is that it lets game developers target only one microarchitecture for the brand/generation.
    Reply
  • bit_user
    cyrusfox said:
    Glad to see price drop on A750 as well as driver work maturing. ... I am very interested in the rumored half stepping ACM-G12 as the die is too large for ACM-10 to justify a product @ the $250 mark,
    I am very interested in an Alchemist refresh. I hope it doesn't get canceled.
    Reply
  • InvalidError
    cyrusfox said:
    Glad to see price drop on A750 as well as driver work maturing. I hope they keep all the fixed funciton accelerators on the small gpu die for BattleMage and I am very interested in the rumored half stepping ACM-G12 as the die is too large for ACM-10 to justify a product @ the $250 mark, Good for Intel to clear inventory but I would imagine the Margin is 0 or negative on these products for them.
    The A750 is the outlet for A770 dies that don't make the grade, can't really have a separate die for it. What would make more sense for a refresh lineup would be a die that more cost effectively covers the A380-580 gap.

    IMO, DG2-128 aimed too low to be viable beyond basic desktop graphics. Intel really needed a DG2-256 to turn some heads at the $200-ish mark, knocking the RX6500 and anything weaker than that at the time out of the park.
    Reply
  • bit_user
    InvalidError said:
    What would make more sense for a refresh lineup would be a die that more cost effectively covers the A380-580 gap.
    Or, maybe there are chip bugs that are holding back performance. That's my hope.
    Reply
  • -Fran-
    Learning lessons is great. Let's hope that is actually true.

    Now the important matter is going to be the "when". Given how badly they fumbled all of the ARC launch dates and the expectations, I really really hope they learned their lesson.

    Regards.
    Reply
  • InvalidError
    bit_user said:
    Or, maybe there are chip bugs that are holding back performance. That's my hope.
    I wouldn't mind a more palatable sub-$200 GPU based on a DG2-256 die free from whatever known hardware shortcomings may have been identified in "Alchemist-gen1" assuming it becomes available before I decide (or am forced) to retire my GTX1050.
    Reply
  • usertests
    cyrusfox said:
    Another day, another article suggesting Intel dGPU business is safe for now, I hope MLID got this wrong, demise/exit of Intel discrete graphics.

    What he has been saying lately (or from the beginning idk) is that there will be a low-end Battlemage die, not multiple, not competing with high-end. And Intel will obviously never cancel integrated graphics. That's consistent with Intel's PR.

    It would be nice to see more competition under $200. The Arc A380 is already an interesting alternative to the RX 6400 and 6500 XT, with more VRAM, more PCIe lanes, more video support, but worse performance. If AMD makes a 7500 XT with 6 GB, PCIe 4.0 x8, and the missing decode/encode, there would be nothing to complain about except the prices.
    Reply