Announcement

Collapse
No announcement yet.

AMD Ryzen 7 1800X Linux Benchmarks

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

  • #51
    I have Ryzen 7 1700 from about an hour's worth of testing, I am going to suggest anyone wanting to get maximum performance use the performance governor. The on-demand governor sucks and will not allow the CPU to clock as high as it should.

    Comment


    • #52
      Thanks for a great, informative review. Seems like Ryzen will provide the development workstation upgrade I'm looking for.

      The conclusion point out three benchmarks that are problematic for Ryzen; however, only one of those (Himeno) shows Ryzen losing to the closest 8-core Intel part (5960X). In general the "challenging" benchmarks for Ryzen are ones that seem single-thread heavy and clock-speed dependent, as it's still able to outperform Intel parts costing twice as much. Outside of Himeno none of the results show particularly negative results for Ryzen. Every time it loses to the 7700K, Intel's more expensive 6 and 8 core parts lose to Ryzen, so that seems to match expectations. Similar IPC. More cores for less. Still can't overcome giant clock speed disadvantages. Fair trade off in my book, but wanted to point out you seem to be dinging Ryzen for poor performance against the "cheaper" Intel part and not acknowledging it trouncing the more expensive Intel parts (6800K + 5960X) in the same tests.

      Originally posted by LinuxID10T View Post
      I have Ryzen 7 1700 from about an hour's worth of testing, I am going to suggest anyone wanting to get maximum performance use the performance governor. The on-demand governor sucks and will not allow the CPU to clock as high as it should.
      Can you expand on this, please? What is the performance governor and the on-demand governor and how can I switch between these? Obviously they control clock speed but it's not clear if you are referring to BIOS facilities, kernel facilities, or some userland utility. I pre-ordered the same CPU and want to be as prepared as possible.

      Comment


      • #53
        This CPU is really amazing for professional applications. I am very interested how it will perform with games.

        Michael
        AMD has confirmed that Ryzen has ECC enabled. Some motherboard manufacturers like ASRock also have ECC capable boards. Is there any chance you could test the ECC capability - perhaps with ecc_enable_override=1? I haven't read of anyone who got ECC actually working yet.

        Comment


        • #54
          Originally posted by hiryu View Post
          Does anyone know if the Ryzen still support ECC? Could be a great chip for home servers if they do.
          It does -> http://www.overclockers.ua/news/memo...-for-rezen.jpg

          Comment


          • #55
            Originally posted by existensil View Post
            T
            Can you expand on this, please? What is the performance governor and the on-demand governor and how can I switch between these? Obviously they control clock speed but it's not clear if you are referring to BIOS facilities, kernel facilities, or some userland utility. I pre-ordered the same CPU and want to be as prepared as possible.
            Those are the kernel drivers... as he stated it wasn't abiguous at all. The performance govenor leaves the CPU throttled up as much as possible, the on-demand govenor throttles up quickly if CPU load increases only.

            On most distros you can configure which one is used from userspace using the cpufreq-set program from the command line. Or a graphical utility as well (usually a tray pluging etc..)

            Comment


            • #56
              Michael Could you order/label the graphs more clearly? So that it is easier to see if the benchmark is multi or single threaded. E.g. sort them in pages or with another heading. Or maybe label in the graph if the bench is mult or singlethreaded.

              Thanks

              Comment


              • #57
                Originally posted by lightonflux View Post
                Michael Could you order/label the graphs more clearly? So that it is easier to see if the benchmark is multi or single threaded. E.g. sort them in pages or with another heading. Or maybe label in the graph if the bench is mult or singlethreaded.

                Thanks
                Usually in the test notes on the graph will reference pthreads or OpenMP they are linked against if they are multi-threaded. If you have any particular ideas, the code is open and hosted on GitHub, pull requests happily accepted: https://github.com/phoronix-test-sui...nix-test-suite
                Michael Larabel
                https://www.michaellarabel.com/

                Comment


                • #58
                  Originally posted by TitanFigter View Post
                  This image is misleading.
                  While the board supports putting ECC in, it will run in Non-ECC-mode (which ECC-Ram can do).
                  Looking in the more detailed specs, most mainboard manufacturers make this clear, so it's highly unlikely that the current desktop Ryzen platform will have ECC support at any point.

                  We'll likely have to wait for "Pro"-CPUs or boards - or even Opterons.

                  Edit: It might be different after all, according to AMDs tech marketing people: https://www.reddit.com/r/Amd/comment...other/def6vs2/
                  I guess we'll have to find out if the boards indeed do support it.
                  Last edited by tg--; 02 March 2017, 06:42 PM.

                  Comment


                  • #59
                    Thanks for the info cb88. I will try to learn more about the cpufreq kernel drivers.

                    Originally posted by cb88 View Post
                    Those are the kernel drivers... as he stated it wasn't abiguous at all.
                    The word "kernel" literally doesn't appear anywhere in LinuxID10T's post, nor any other obvious hints to what he was referring for those of us who don't already know the answer. I'm a careful reader. I didn't miss anything. He literally never said anything like that and it was the definition of ambiguous for someone without the requisite knowledge.

                    Comment


                    • #60
                      I expected this sort of review across the board. This is a completely new architecture, and I expect that a lot of performance issues will probably get ironed out over the next 6 months to a year. Anyways I was looking at the review and have two questions about the review.

                      Question 1: I know that the review is using 16GB of memory for the Ryzen system, but this tells me nothing. What ram kit is being used on the system, and what are the ram settings?

                      Question 2: I noticed that this review, and other reviews of it's kind do not mention how sensitive the processor is to ram timings. Is there any chance to run some further testing on ram speeds and timings. I created a list of timings that should be easily obtained with a ram kit...
                      • 14-14-14 Cas latency test speeds: DDR4-2133, DDR4-2400, DDR4-2666, DDR4-2800, DDR4-3000, and DDR4-3200.
                      • 15-15-15 Cas latency test speeds: DDR4-2133, DDR4-2400, DDR4-2666, DDR4-2800, DDR4-3000, and DDR4-3200.
                      • 16-16-16 Cas latency test speeds: DDR4-2133, DDR4-2400, DDR4-2666, DDR4-2800, DDR4-3000, and DDR4-3200.
                      • 17-17-17 Cas latency test speeds: DDR4-2133, DDR4-2400, DDR4-2666, DDR4-2800, DDR4-3000, and DDR4-3200.



                      Note: In I would expect performance differences where the higher the timings, the worst the performance, but The real question is how much performance is really lost.

                      Question 3: Any Chance to repeat the tests from Question 2 on a DDR4 enabled intel system?

                      Comment

                      Working...
                      X