5 ms·

Can anyone explain what prevents AMD from making x86_64 chips competitive with ARM on the lower end like in mobile phones? I doubt it's about ISA.

by shmerl 11mo ago

wmf 11mo ago
Their lowest end chips are probably competitive already. I think x86 support was removed from Android though.
- shmerl 11mo ago
  So why did for example Valve decide to use Qualcomm Snapdragon for Steam Frame and not some AMD APU?
  - kube-system 11mo ago
    I have seen speculation that mobile app architecture compatibility was part of it
    - shmerl 11mo ago
      I see. But aren't they emulating x86_64 on ARM64 there anyway? Can't they emulate ARM64 on x86_64 the same way?
      - rincebrain 11mo ago
        Yes, in theory. But there's not a huge market (and therefore, FOSS dev time spent on it) for emulating AArch64 on x86 the way there is x86 on AArch64, so if your options are to build your own AArch64 emulation for x86 (or drop a fortune into an existing FOSS option), or building something based on AArch64 and using the existing x86 emulation implementations, one of these has much more predictable costs and outcomes today.
        shmerl 11mo ago
        Sounds like a project for Valve to back then.
        rincebrain 11mo ago
        They could, but why? If an ARM device both suits the goals and has lower risk, there's little upside other than forcing the project to exist. And since there's very few pieces of AArch64-exclusive software that Valve is trying to support, that's not a goal that benefits the project. (If I were guessing without doing much research, Switch emulators might be the largest investment of effort in open source on x86 systems running AArch64 things performantly, but that's certainly not a market segment Valve is targeting, so...)
        shmerl 11mo ago
        To have flexibility and not being tied to CPU choice due to simply lacking software side. Besides, such kind of project would be beneficial to more scenarios than their immediate need. They didn't need to back a bunch of projects they backed (radv / aco are a big example where you could claim there was redundancy so they weren't strictly necessary), but results paid off even to the point where AMD is dropping their amdvlk in favor of using radv/aco. They didn't need to strictly speaking back zink either (OpenGL on Vulkan), but they decided to and it gives them a long term ability to have OpenGL even if native drivers will be gone, as well as an upside of bringing up OpenGL on any Vulkan available target out of the box. It's just something in their style to do when there seems to be an obvious gap.
        rincebrain 11mo ago
        It's in their style to do when they think there's a benefit to them from it, short or long term. I'm not claiming they're not beneficial or acting in bad faith, but their past contributions are "making enormous improvements in the ecosystem when they think this is the best option for them" - things like radv+aco you could see as "we looked at AMD's solution for this and decided it was lower risk to build a better one ourselves" - and I can't really argue with their logic, even if AMD hadn't given up and gone for it, they have a history of switching solutions every so often and then god help you if you built around the old one. Zink makes sense because it closes a unique gap (supporting older games if OpenGL support sunsets) in their platform, not just for its own sake. AArch64 emulation on x86 is useful, but doesn't fill such a gap for them, IMO. Nothing is uniquely targeting AArch64 in the library of things they want to support on Steam platforms except non-game software targeting only macOS/AArch64. I would predict, since they seem to be adding some level of Steam game support for handling Android games [1] [2], that you might see them financing such a thing in the next couple years if there's enough things exclusively targeting Android or AArch64 that they want to de-risk people complaining they can no longer play them in 5-10 years. [1] - https://www.gamingonlinux.com/2025/11/steamworks-sdk-adds-support-for-linux-arm-and-android-and-it-seems-we-know-the-first-android-game-on-steam/ https://www.gamingonlinux.com/2025/11/steamworks-sdk-adds-su... [2] - https://www.theverge.com/news/818672/valve-android-apps-steam-frame https://www.theverge.com/news/818672/valve-android-apps-stea...
axiolite 11mo ago
Just price, I'd say. AMD / Intel are used to a certain margin on their products, and the low barrier to entry to create ARM CPUs, and fierce competition from giants like Broadcom, keeps margins very thin in this market. The original smart phones like the Nokia Communicator 9110i were x86 based. AMD previously had very impressive low-power CPUs, like the Geode, running under 1-watt. Intel took another run at it with Atom, and were able to manage x86 phones (eg: Asus Zenphone) slightly better than contemporary ARM based devices, but the price for their silicon was quite a bit higher than ARM competitors. And Intel had to sink so much money into Atom, in an attempt to dominate the phone/tablet market, that they couldn't be happy just eeking out a small sliver of the market by only being slightly better at a significantly premium price.
- shmerl 11mo ago
  I see, but why others like Qualcomm are doing it then? They are OK with low margins?
  - ACCount37 11mo ago
    Qualcomm has a massive "value add" because they own the modem. As well as a doom stack of patents on all things cellular. You need a modem if you want to make a smartphone. And Qualcomm makes sure to, first, make some parts of the modem a part of their SoC, and second, never give a better deal on a standalone modem than on a modem and SoC combo. Sure, AMD could make their own modem, but it took Apple ages to develop a modem in-house. And AMD could partner with someone like Mediatek and use their hardware - but, again, that would require Mediatek to prop up their competition in SoC space, so, don't expect good deals.
    - shmerl 11mo ago
      Not every scenario for such chips is a smartphone, but as you said, AMD could as well develop their own modem. I would prefer them to start with WiFi though, since Intel made their latest chips impossible to use with AMD CPUs.

Can anyone explain what prevents AMD from making x86_64 chips competitive with ARM on the lower end like in mobile phones? I doubt it's about ISA.

by shmerl 11mo ago

Can anyone explain what prevents AMD from making x86_64 chips competitive with ARM on the lower end like in mobile phones? I doubt it's about ISA.

wmf 11mo ago

Their lowest end chips are probably competitive already. I think x86 support was removed from Android though.

shmerl 11mo ago
So why did for example Valve decide to use Qualcomm Snapdragon for Steam Frame and not some AMD APU?
- kube-system 11mo ago
  I have seen speculation that mobile app architecture compatibility was part of it
  - shmerl 11mo ago
    I see. But aren't they emulating x86_64 on ARM64 there anyway? Can't they emulate ARM64 on x86_64 the same way?
    - rincebrain 11mo ago
      Yes, in theory. But there's not a huge market (and therefore, FOSS dev time spent on it) for emulating AArch64 on x86 the way there is x86 on AArch64, so if your options are to build your own AArch64 emulation for x86 (or drop a fortune into an existing FOSS option), or building something based on AArch64 and using the existing x86 emulation implementations, one of these has much more predictable costs and outcomes today.
      - shmerl 11mo ago
        Sounds like a project for Valve to back then.
        rincebrain 11mo ago
        They could, but why? If an ARM device both suits the goals and has lower risk, there's little upside other than forcing the project to exist. And since there's very few pieces of AArch64-exclusive software that Valve is trying to support, that's not a goal that benefits the project. (If I were guessing without doing much research, Switch emulators might be the largest investment of effort in open source on x86 systems running AArch64 things performantly, but that's certainly not a market segment Valve is targeting, so...)
        shmerl 11mo ago
        To have flexibility and not being tied to CPU choice due to simply lacking software side. Besides, such kind of project would be beneficial to more scenarios than their immediate need. They didn't need to back a bunch of projects they backed (radv / aco are a big example where you could claim there was redundancy so they weren't strictly necessary), but results paid off even to the point where AMD is dropping their amdvlk in favor of using radv/aco. They didn't need to strictly speaking back zink either (OpenGL on Vulkan), but they decided to and it gives them a long term ability to have OpenGL even if native drivers will be gone, as well as an upside of bringing up OpenGL on any Vulkan available target out of the box. It's just something in their style to do when there seems to be an obvious gap.
        rincebrain 11mo ago
        It's in their style to do when they think there's a benefit to them from it, short or long term. I'm not claiming they're not beneficial or acting in bad faith, but their past contributions are "making enormous improvements in the ecosystem when they think this is the best option for them" - things like radv+aco you could see as "we looked at AMD's solution for this and decided it was lower risk to build a better one ourselves" - and I can't really argue with their logic, even if AMD hadn't given up and gone for it, they have a history of switching solutions every so often and then god help you if you built around the old one. Zink makes sense because it closes a unique gap (supporting older games if OpenGL support sunsets) in their platform, not just for its own sake. AArch64 emulation on x86 is useful, but doesn't fill such a gap for them, IMO. Nothing is uniquely targeting AArch64 in the library of things they want to support on Steam platforms except non-game software targeting only macOS/AArch64. I would predict, since they seem to be adding some level of Steam game support for handling Android games [1] [2], that you might see them financing such a thing in the next couple years if there's enough things exclusively targeting Android or AArch64 that they want to de-risk people complaining they can no longer play them in 5-10 years. [1] - https://www.gamingonlinux.com/2025/11/steamworks-sdk-adds-support-for-linux-arm-and-android-and-it-seems-we-know-the-first-android-game-on-steam/ https://www.gamingonlinux.com/2025/11/steamworks-sdk-adds-su... [2] - https://www.theverge.com/news/818672/valve-android-apps-steam-frame https://www.theverge.com/news/818672/valve-android-apps-stea...

axiolite 11mo ago

Just price, I'd say. AMD / Intel are used to a certain margin on their products, and the low barrier to entry to create ARM CPUs, and fierce competition from giants like Broadcom, keeps margins very thin in this market. The original smart phones like the Nokia Communicator 9110i were x86 based. AMD previously had very impressive low-power CPUs, like the Geode, running under 1-watt. Intel took another run at it with Atom, and were able to manage x86 phones (eg: Asus Zenphone) slightly better than contemporary ARM based devices, but the price for their silicon was quite a bit higher than ARM competitors. And Intel had to sink so much money into Atom, in an attempt to dominate the phone/tablet market, that they couldn't be happy just eeking out a small sliver of the market by only being slightly better at a significantly premium price.

shmerl 11mo ago
I see, but why others like Qualcomm are doing it then? They are OK with low margins?
- ACCount37 11mo ago
  Qualcomm has a massive "value add" because they own the modem. As well as a doom stack of patents on all things cellular. You need a modem if you want to make a smartphone. And Qualcomm makes sure to, first, make some parts of the modem a part of their SoC, and second, never give a better deal on a standalone modem than on a modem and SoC combo. Sure, AMD could make their own modem, but it took Apple ages to develop a modem in-house. And AMD could partner with someone like Mediatek and use their hardware - but, again, that would require Mediatek to prop up their competition in SoC space, so, don't expect good deals.
  - shmerl 11mo ago
    Not every scenario for such chips is a smartphone, but as you said, AMD could as well develop their own modem. I would prefer them to start with WiFi though, since Intel made their latest chips impossible to use with AMD CPUs.