Announcement

Collapse
No announcement yet.

Linux 4.12 Ubuntu Benchmarks With AMD Ryzen, Intel Kabylake - 12 Systems

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

  • Linux 4.12 Ubuntu Benchmarks With AMD Ryzen, Intel Kabylake - 12 Systems

    Phoronix: Linux 4.12 Ubuntu Benchmarks With AMD Ryzen, Intel Kabylake - 12 Systems

    While waiting for my motherboards to arrive for the new Core i7 7740X and Core i9 7900X, I've been re-testing many of my AMD/Intel boxes with Ubuntu 17.04 on the latest Linux 4.12 kernel for comparison to Intel's new high-end processors. Here is a look at 12 of the existing systems when running on the Linux 4.12 kernel as well as all of the systems have the latest BIOSes, etc.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    If you included a CPU usage graph for each test, it would be really useful.

    Comment


    • #3
      have you disabled hpet on amd motheroboard before the tests?

      Comment


      • #4
        "Even with Zen CPUs having AVX2, they come noticeably behind the Intel CPUs in the Himeno pressure solver test."

        While I'm a huge AMD Ryzen fan, there is unfortunately some kind of limit in regards to fused multiple ads or so, basically halfing the thruput compared to the Intel CPUs, ... :-/

        Beside that Ryzen is of course awesome, a welcome change to kick Intel's ass and innovate again ;-)!

        Comment


        • #5
          I guess some of these (eg scikit-learn) depends on lapack and blas. Are these libraries tuned for AMD?

          Comment


          • #6
            That's what I was wondering - some of the benchmarks have started linking with Intel's MKL, which AFAIK runs different code paths for non-Intel CPUs.
            Test signature

            Comment


            • #7
              "The smallpt path tracer is a race about the number of CPU cores."
              What? The Ryzen 7 1800X is performing basically the same as a CPU with two more cores than it, and you're saying it's a core race?

              Comment


              • #8
                When results are like this it's usually a race about cores x IPC x clocks. When the software doesn't scale well with cores of course the i7-7700K is usually the fastest then. Because then IPC x clocks matter.

                But normally you have mixed workloads and when the 7700K is 30% faster in 10 scenarios which don't interrupt your workflow taking 2 seconds while you save 5 minutes of rendering time with the 1800X(or even just 1700) the Ryzen CPU is still much better in sum.

                Finally I'm always quite happy to see the Linux benchmarks. It shows how optimized open source software is while the widely used commercial windows equivalents mostly scale with just one core. For Linux users Ryzen is even better for example with Darktable.
                Last edited by oooverclocker; 23 June 2017, 03:59 PM.

                Comment


                • #9


                  Comment


                  • #10
                    About this second picture in the article... Are the readers supposed to take notice of the hard liquor at the work place? I cannot say I like it.

                    Comment

                    Working...
                    X