Announcement

Collapse
No announcement yet.

Ryzen 9 3900X/3950X vs. Core i9 10900K In 380+ Benchmarks

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

  • #11
    phoronix-test-suite analyze-run-times 2005283-PTS-BIGCOMPA49 for all the info in full.

    #####
    ACCUMULATED TIME:
    Core i9 10900K 1 Day, 16 Hours, 45 Minutes, 5 Seconds
    Ryzen 9 3900X 1 Day, 15 Hours, 45 Minutes, 40 Seconds
    Ryzen 9 3950X 1 Day, 15 Hours, 37 Minutes
    Average: 1 Day, 16 Hours, 2 Minutes, 35 Seconds
    #####


    Michael Larabel
    https://www.michaellarabel.com/

    Comment


    • #12
      Michael In the article there is missing image https://www.phoronix.com/misc/comet-18.png

      Krzysztof

      Comment


      • #13
        Originally posted by kcybulski View Post
        Michael In the article there is missing image https://www.phoronix.com/misc/comet-18.png

        Krzysztof
        Fixed thanks
        Michael Larabel
        https://www.michaellarabel.com/

        Comment


        • #14
          Originally posted by Michael View Post

          Fixed thanks
          Wait please there is typo

          Originally posted by phoronix View Post
          When looking at the breakdown of tests by all those that are C/C++ codebases like GraphicsMagick, libvpx, TSCP, FFTW, Stcokfish, FLAC audio, etc

          Comment


          • #15
            Also, why are the graphs in bitmap format instead of vector?...

            At least not DejaVu please... you could have saved a ton of space by not using that "Default Linux Font"... :<

            ...it just gives a bad image of how Linux looks like...

            Microsoft: evolved from Tahoma (Verdana-looking, which in turn is a lot like DejaVu) to Segoe UI
            Apple: evolved from Lucida (which is fat too) to Helvetica and then San Francisco
            Linux?: yeah, it's 2020 and we are still stuck on DejaVu

            Comment


            • #16
              Originally posted by tildearrow View Post
              Also, why are the graphs in bitmap format instead of vector?...

              At least not DejaVu please... you could have saved a ton of space by not using that "Default Linux Font"... :<

              ...it just gives a bad image of how Linux looks like...
              OpenBenchmarking.org currently doesn't expose the functionality for the per-suite geometric means, so I screenshotted them from my desktop with the local PTS result viewer and just uploaded that.
              Michael Larabel
              https://www.michaellarabel.com/

              Comment


              • #17
                Originally posted by mlau View Post
                What's the runtime of some of these benchmarks? That Intel 5.3GHz boost is really impressive: the 3950x has 60% more cores, and it's still only 15% faster in (i believe longer running) compilation tests.
                That can mean many things. One interpretation is that the majority of the tests do not scale well to high core counts but tbh I did not look at the individual results. The composite scores for workstation, HPC, and rendering loads could indicate this, though.

                My takeaway is: If your workload is threaded well, get the AMD chip. Otherwise, get the Intel chip, if you can afford the required mainboard, PSU, and cooling. In that case, I'd go for the 8-core version, though. No point in paying for 2 additional cores, if your workload isn't threaded well anyway.

                Comment


                • #18
                  Originally posted by Michael View Post
                  phoronix-test-suite analyze-run-times 2005283-PTS-BIGCOMPA49 for all the info in full.
                  Thanks, that would have sufficed

                  Last edited by mlau; 28 May 2020, 03:12 PM.

                  Comment


                  • #19
                    Originally posted by mlau View Post



                    that looks wrong. the 3950x isn't that slow, even my poorly-cooled,ddr4-2400 memory equipped 3950x builds x86_64_defconfig in 50 seconds.
                    Note it's due to the test repetition.... due to standard deviation being above 3.0%, it ran extra times. On the benchmark result you can see the actual average, this is the elapsed time.
                    Michael Larabel
                    https://www.michaellarabel.com/

                    Comment


                    • #20
                      Originally posted by phoronix
                      Meanwhile for thread-happy ray-tracing workloads with C-Ray, Tachyon, rays1bench, and YafaRay, the Ryzen 9 3990X was 7.6% faster than the i9-10900K.
                      That would be an impressive showing by Intel!

                      Comment

                      Working...
                      X