Announcement

Collapse
No announcement yet.

AMD Ryzen 7000 Series Make For Compelling Budget Servers, Leading Performance & Value Over Xeon E

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

  • #11
    Originally posted by S.Pam View Post

    Also, enabling different numa settings in bios, if available can make a huge difference too. On my EPYC server it makes cinebench 20% faster!
    I remarked the same.
    Enabling L3 as NUMA in the BIOS caused similar uplift for a Ryzen 7950X3D.
    I wonder why it is not on by default.

    Comment


    • #12
      Originally posted by Michael View Post

      Defaults. Not too worthwhile for multi-threaded workloads to segregate them since making use of all the cores will pan out better for said tests than limiting it to just a subset anyhow.
      That's what I figured since you didn't mention anything special.

      Looking at 7800x3d results from various places, I was noticing here with Linux tasks that the higher core counts seem to scale better with higher core workloads versus how limiting to a specific CCD seems to be more helpful with Windows gaming loads that don't use every core available so processes aren't cache thrashing between the CCDs. Perhaps that'll be something that shows up in future Linux x3d gaming benchmarks since games and other unoptimized processes are what seem to trigger it.

      Comment


      • #13
        Meh. Michael's "budget friendly" and my version of "budget friendly" are two different things apparently. Used PowerEdge or ThinkServer for under 400 USD, not $1000+.

        Comment


        • #14
          Typo:

          Originally posted by phoronix View Post
          as a very interesting 1U rackmount barebobes server

          Comment


          • #15
            Originally posted by grigi View Post
            There something seriously wrong with the joules/run calculations:

            For "Timed Godot Game Engine Compilation" example:
            Xeon E-2388G took 472.30 seconds, and used 92.86W avg.
            Ryzen 5 5600X took 425.74 seconds, and used 73.73W avg.

            Somehow that calculates to:
            Xeon E-2388G: 8729 Joules/run
            Ryzen 5 5600X: 31728 Joules/run

            So the Ryzen 5 does it faster and uses less power during the task, but uses nearly 4x as much power for the task?‚Äč
            I checked by hand and the AMD energy calculations sounds corect (J is just Ws ), but the intel ones are wrong (by 5x?)

            Comment


            • #16
              Michael Great report, thank you! Epic set of tests.

              Have you heard any indication from vendors that support for non-binary DIMM sizes with ECC will be coming? A box like this is tempting for my homelab but 128GB isn't enough for my workloads -- 192GB would be perfect. Yes, I could go with something older/Intel that allows for more RAM but the jobs I've got would really love a 7950X.

              I didn't notice idle power consumption being explicitly called out. Should we take that as the minimum of wattage shown, e.g. the 7950X minimum was at about 16 watts on the Phoronix test suite? Those numbers looked suspiciously low for idle, so I'm hesitant to rely on them.

              Thanks again.
              Last edited by Scramblejams; 05 September 2023, 08:46 PM.

              Comment


              • #17
                Would you please add barebones specs to the big table of CPUs-tested (after the words 'based on what I had available/access to testing')? We have not all memorized what a Ryzen 5 5600G is, nevermind a Xeon E-2336...!

                It'd be even better if all the individual graphs had tiny comments like '16c/32t 4.5-5.7GHz'...

                Comment


                • #18
                  Originally posted by stormcrow View Post
                  Meh. Michael's "budget friendly" and my version of "budget friendly" are two different things apparently. Used PowerEdge or ThinkServer for under 400 USD, not $1000+.
                  In fact, if you were to buy a brand new workstation or server from OEMs instead of DIY-ing it, it sometimes even ends up cheaper than the DIY build seeing that you get on-site warranty for the entire package, as well as the option to use certain processors that may not be available off-the-shelf. Especially if you live in an area where aftermarket server boards and server components like ECC memory are rarer than diamonds.

                  I stopped DIY-ing full systems years ago when I did the math. And now I only purchase prebuilds from Acer and Lenovo.
                  Last edited by Sonadow; 05 September 2023, 08:59 PM.

                  Comment


                  • #19
                    Originally posted by Sonadow View Post
                    In fact, if you were to buy a brand new workstation or server from OEMs instead of DIY-ing it, it sometimes even ends up cheaper than the DIY build seeing that you get on-site warranty for the entire package, as well as the option to use certain processors that may not be available off-the-shelf. Especially if you live in an area where aftermarket server boards and server components like ECC memory are rarer than diamonds.

                    I stopped DIY-ing full systems years ago when I did the math. And now I only purchase prebuilds from Acer and Lenovo.
                    Prebuilt prices here can be crazy, and often do not offer a lot of flexibility. I still like DIY-ing my systems, but know what you mean. Sometimes a prebuilt is more cost effective.

                    That said, my experiences of Acer have been very poor, and the Lenovo systems were were provided with some new equipment are... less than adequate, but I would suggest that that is the equipment supplier trying to save fractions of a dollar here and there (who in their right mind specs up a system with a 64-bit OS, 4GB of RAM and a 5400RPM HDD in the 2020s?)

                    Comment


                    • #20
                      Originally posted by Paradigm Shifter View Post
                      I would suggest that that is the equipment supplier trying to save fractions of a dollar here and there (who in their right mind specs up a system with a 64-bit OS, 4GB of RAM and a 5400RPM HDD in the 2020s?)
                      And yet sometimes, even after ripping out the bundled 4GB memory and 5400rpm HDD and replacing it with your own 8 or 16GB memory and a cheapass SSD, it still ends up cheaper than a DIY build.

                      The variations in prices between OEM prebuilts and aftermarket DIY builds can sometimes be staggering.

                      Comment

                      Working...
                      X