idk who that guy is but I couldn’t think of how to word the title, sorry

︀︀• Custom AMD Zen 4 CPU

︀︀• 6 cores / 12 threads

︀︀• RDNA 3 GPU

︀︀• 28 Compute Units

︀︀• 8GB GDDR6 VRAM

︀︀• 16GB DDR5 RAM

︀︀• 512GB or 2TB NVMe SSD

︀︀• SteamOS

︀︀• Wi-Fi 6E

︀︀• Bluetooth 5.3

︀︀• Gigabit Ethernet

︀︀• HDMI 2.0 + DisplayPort 1.4

︀︀• microSD expansion

  • grue@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    1 day ago

    They say it’s a custom design, so surely they could’ve custom-designed it to be unified rather than discrete if they wanted. I guess maybe they were trying to make sure it would only be bought by gamers by deliberately making it less versatile for AI?

    • SavvyBeardedFish@reddthat.com
      link
      fedilink
      arrow-up
      6
      ·
      24 hours ago

      Custom in this case doesn’t really need to carry any weight either, it could be a simple voltage bump, clock bump, laser cutting cores etc. and they would still call it custom.

      It’s not a “from the ground up” custom chip. Unified still requires a significant amount of chip area per die, especially if they want to have a relatively beefy GPU (somewhere below Radeon 8060S, but above Radeon 780M).

      I would imagine this gives the best perf./buck from Valve’s POV, without costing an arm and a leg

      • grue@lemmy.world
        link
        fedilink
        arrow-up
        3
        ·
        23 hours ago

        Unified still requires a significant amount of chip area per die

        Even compared to having two entirely separate memory controllers, one for the CPU and one for the GPU?

        • SavvyBeardedFish@reddthat.com
          link
          fedilink
          arrow-up
          3
          ·
          22 hours ago

          I would assume the total area is larger for the separate CPU+GPU die when compared a single unified chip, sure. But the cost per millimeter doesn’t necessarily scale linearly either (larger chip, lower yields), so it might be cheaper to buy CPU+GPU rather than the unified chip even though the total area is larger.

          For reference, TechPowerUp lists:

          RX 7600M: 204 mm² @ TSMC 6 nm

          Strix Halo: 308 mm² @ TSMC 4 nm

          Not sure what kind of area one could expect for the CPU alone (without the integrated GPU) for this kind of process

          • grue@lemmy.world
            link
            fedilink
            arrow-up
            2
            ·
            21 hours ago

            larger chip, lower yields

            Oh right, I forgot about that part.

            Not sure what kind of area one could expect for the CPU alone (without the integrated GPU) for this kind of process

            I guess you could look up specs for a desktop Ryzen CPU that doesn’t have integrated graphics. Not sure which is the right one to pick, but I checked a few Zen 4 AM5 chips and they were all 71 mm2 @ TSMC 5 nm.


            BTW, what actually is “Strix Halo” anyway? I’m confused about whether it’s what they’re calling all the latest-generation APUs, or just the high-end ones, or Asus co-branding, or what.

            Are there not any lesser APUs (with smaller die size and higher yields) that aren’t “Strix Halo” but still have a similar architecture and decent gaming performance?

            • SavvyBeardedFish@reddthat.com
              link
              fedilink
              arrow-up
              2
              ·
              18 hours ago

              BTW, what actually is “Strix Halo” anyway? I’m confused about whether it’s what they’re calling all the latest-generation APUs, or just the high-end ones, or Asus co-branding, or what.

              Strix Halo is the “high-end” ones (and currently the latest), in terms of gaming they are closer to previous gen discrete laptop GPU (hence they use the naming scheme Radeon 8XXX series).

              There are smaller ones as well, the one that is “mid-tier” is Strix Point, which has the Radeon 800M series GPU, i.e. closer to what one had in previous generations of integrated GPUs.

              In terms of gaming performance, you can compare using Notebookcheck, as an example; Kingdom Come Deliverance 2 on low preset at 1080P:

              CPU GPU APU Average FPS
              Ryzen AI 9 HX 370 AMD Radeon 890M Strix Point 19.9
              Ryzen AI Max Pro 390 AMD Radeon 8050S Strix Halo 76.6
              Ryzen AI Max+ 395 AMD Radeon 8060S Strix Halo 101.7

              So, there’s a pretty big leap going from Strix Point (mid tier) to Strix Halo (high-end)

              • grue@lemmy.world
                link
                fedilink
                arrow-up
                2
                ·
                18 hours ago

                So, there’s a pretty big leap going from Strix Point (mid tier) to Strix Halo (high-end)

                Holy shit, no kidding! I guess maybe that’s the reason Valve didn’t go that way: they wanted to put their product right in the middle of that graphics gap. Also, even that first one apparently has 12 CPU cores, so the whole balance between CPU and GPU performance is just off.

                Still though, if we’re talking custom, it would’ve been cool if Valve could’ve had them build something equivalent to a “Ryzen AI 7” or “Ryzen AI 5”, but still with Radeon 8050S or 8060S graphics.