Announcement

Collapse
No announcement yet.

AMD Zen 4 AVX-512 Performance Analysis On The Ryzen 9 7950X

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

  • #11
    Originally posted by Michael View Post

    The colors are auto-generated and hard to sometimes to get it to behave robustly and correctly with having not yet come up with a great algorithm for always choosing great colors.
    Well.... for under four colors it should be simple to just default to some well spread out choices. https://www.dataquest.io/blog/what-t...visualization/ looks like a great reference. And I do realize that picking colors is hard, but I also feel that you could do better when comparing just two or three values so that they're better visually.

    Cheers,
    John

    Comment


    • #12
      Supports what I've been saying all along, which is that 512-bits is too wide.

      BTW, I'm pretty sure the only benches where AVX-512 uses more power are those which use fp16 or bf16, which equates to processing 2x the number of elements per cycle than you can do with AVX2. Otherwise, AVX-512 also seems a win on power!

      Comment


      • #13
        Originally posted by Michael View Post
        The colors are auto-generated and hard to sometimes to get it to behave robustly and correctly with having not yet come up with a great algorithm for always choosing great colors.
        It would help if the graphs were either bigger (higher resolution) or if there were some plotting option to make the lines partially opaque.

        ...or, render them in SVG and that way I can easily open them in another tab and zoom.

        Comment


        • #14
          Originally posted by MorrisS. View Post
          A test between AVX2 and AVX5 performance would be useful.
          I believe most of these tests are using AVX2, when AVX-512 is disabled.

          Comment


          • #15
            Originally posted by ms178 View Post
            Finally, AVX-512 on a mainstream CPU people actually want to buy.
            Hopefully, this spurs the adoption of glibc's hwcaps feature by more libraries & distros.

            Comment


            • #16
              Pretty good results for what is effectively a half-baked solution. I'm impressed the power draw didn't change much despite the performance increase.

              Comment


              • #17
                Originally posted by schmidtbag View Post
                Pretty good results for what is effectively a half-baked solution. I'm impressed the power draw didn't change much despite the performance increase.
                It's not half-baked, it's half-sized and perfectly baked

                In fairness we do also have multiple FP execution ports/pipes (6 vs 3 for Golden Cove) so there can still be a lot of work happening in parallel.
                Last edited by bridgman; 26 September 2022, 03:43 PM.
                Test signature

                Comment


                • #18
                  The videos used with dav1d aren't 10 bit? Performance gains should be much bigger vs. 8 bit video.

                  Comment


                  • #19
                    Originally posted by coder View Post
                    Hopefully, this spurs the adoption of glibc's hwcaps feature by more libraries & distros.
                    So AVX5 yields a significant increase.

                    Comment


                    • #20
                      Let me see if I got this right. Intel introduces AVX512, limits it to super-duper expensive server-grade CPUs and makes it a tricky tradeoff between performance, power consumption and performance-per-watt. AMD has AVX512 that pretty much just works. Intel needed a massive slap on the face after the 25th refresh of WeDontCareLake but this is getting ridiculous.

                      Michael: Is there any chance you could do a quick round of Zen4/ADL AVX512 benchmarks and compare the relative benefits of AVX512 on both uarchs?

                      Comment

                      Working...
                      X