Announcement

Collapse
No announcement yet.

Apple M1 Performance On Linux: Benchmarks Better Than Expected For Its Alpha State

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    This is pretty insane. Considering how small the development team is and how much work it is it implement support for a brand new M1 SoC, this is pretty amazing show. Compared to the lackluster Linux support on the last Intel-based Macs, this really blew my mind.

    Asahi team, you goddamn rock!

    Comment


    • #12
      Originally posted by schmidtbag View Post
      But what about the Mac OS results? If Mac OS is optimized and can tap into power states, why is it overall performing worse?
      Because open source applications themselves are almost entirely built and optimized for Linux. If you get someone willing to port and optimize for *BSD or MacOS then the performance gaps close with Linux. This is especially true of OpenGL which has been deprecated and largely unsupported on OS-X for many years (you have to use Metal).

      Also like to add that if Linux is treating all cores as the same, then there's no throttling going on either.
      Last edited by stormcrow; 23 March 2022, 12:31 PM.

      Comment


      • #13
        This is extremely shameful for Apple. The "powerful" macOS has been defeated or at least almost defeated on its own platform by a system made with reverse engineering.

        Apple has some excuses on Intel, but on its own ARM CPUs it doesn't have any excuse. What happens to drivers and firmware?

        Comment


        • #14
          Does the GPU on the M1 even have the necessary features for OpenGL? I figured Apple wouldn't bother wasting transistors on a technology that they've deprecated

          I was thinking that we'd see Vulkan first, because that's very similar to Apple's Metal, and then we'd use Zink to get OpenGL-via-Vulkan

          Comment


          • #15
            20 times faster than Raspberry Pi...
            When will there be a competitive product that's on the M1's price range?

            Comment


            • #16
              Originally posted by vladpetric View Post
              Microarchitecture matters hugely, and that's a reason why a comparison with the RPi4 (despite having similar instruction sets) is not that meaningful.

              RPi4 - toy-grade. Can issue up to 4 instructions per cycle.

              Apple M1 - best processor on the planet at this time. Can issue up to 12 instructions per cycle (+ 4 vector instructions ... ).
              Problem is Apple M1 has a monopoly on that market - neither Qualcomm nor Ampere are willing to enter that market despite being the only ones capable of providing a competitive product.

              Comment


              • #17
                Originally posted by tildearrow View Post

                Problem is Apple M1 has a monopoly on that market - neither Qualcomm nor Ampere are willing to enter that market despite being the only ones capable of providing a competitive product.
                Personally I don't think Qualcomm or NVidia are even remotely close to this level of microprocessor design (Intel, AMD, IBM - maybe). ARM tried to some degree, but their failure is one of the reason they were on sale as a company, essentially.

                Comment


                • #18
                  Originally posted by jokeyrhyme View Post
                  Does the GPU on the M1 even have the necessary features for OpenGL? I figured Apple wouldn't bother wasting transistors on a technology that they've deprecated
                  OpenGL is a big abstraction API. So yes, no problem to support OpenGL.

                  Comment


                  • #19
                    Originally posted by vladpetric View Post
                    Apple M1 - best processor on the planet at this time. Can issue up to 12 instructions per cycle (+ 4 vector instructions ... ).
                    Do you have some link with details? all I could find was this:
                    https://www.anandtech.com/show/16226...14-deep-dive/2
                    It says it can decode 8 at a time, and it is not clear how many it can issue at a time. Are you counting vectors instructions as multiple instructions?

                    Comment


                    • #20
                      Originally posted by schmidtbag View Post
                      Linux was overall performing very well; this is way better progress than I expected. But I'm confused...
                      I thought the M1 Macs had performance similar to an 11th gen i7, yet, in these tests it is barely keeping up with an i5. What am I missing here? Perhaps it depends on which application is being used.
                      No it was similar to a two generations older _mobile_ core i7 which was themally throttled. It was benchmarked against the last Intel MacMini not a modern equivalent.

                      Comment

                      Working...
                      X