Announcement

Collapse
No announcement yet.

AMD EPYC 4364P & 4564P @ DDR5-4800 / DDR5-5200 vs. Intel Xeon E-2488

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

  • AMD EPYC 4364P & 4564P @ DDR5-4800 / DDR5-5200 vs. Intel Xeon E-2488

    Phoronix: AMD EPYC 4364P & 4564P @ DDR5-4800 / DDR5-5200 vs. Intel Xeon E-2488

    With the AMD EPYC 4004 series that was announced in May and we have delivered benchmarks of the entire EPYC 4004 stack from the 4-core SKU up through the 16-core model with 3D V-Cache, there are many advantages over Intel's Xeon E-2400 series competition. In addition to going up to 16 cores versus 8 with the Xeon E-2400 series, the more competitive pricing, the 3D V-Cache SKUs, and 28 PCIe lanes rather than 20, the AMD EPYC 4004 models also support DDR5-5200 memory where as the Intel Raptor Lake E-2400 models are bound to DDR5-4800. In this follow-up testing is a look at the AMD EPYC 4004 performance both at DDR5-4800 and DDR5-5200 speeds for showing the performance difference.

    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
    Michael thanks for the effort spent on this analysis. Your site is one of few that actually produces and shares facts about new products.

    However, I want to gently express some criticism or rather express my opinion about opportunity for improvement.

    You clearly state that you tested the configurations with DDR5-4800 and DDR5-5200 memory. I have missed any discussion about why you chose these two sets of memory. Why not DDR5-6000 or DDR5-6400, which are supported on Threadripper platforms or, more appropriately for these CPUs, AMD Ryzen (at least for a 2 DIMM configuration)? Yes, because the new CPUs have not improved the memory controller and faster speeds cannot be supported in 4 DIMM configurations, that are probably more appropriate for EPYC CPUs.

    Also, I am watching with interest your selection of benchmark tests in each of the hw review articles. The Phoronix suite has grown to more than just a small swiss army knife of benchmark tests, allowing, no requiring, you to make a manageable selection that is appropriate for this article.
    Most normal human beings (and therefore most of your readers I suppose) really will struggle understanding the technical intricacies of each individual test delegating and putting their trust in you to identify an appropriate selection of tests.
    Can you please do us all a favor and add explanations why the set of tests were chosen? This will reduce potential criticism and improve the quality of the effort.
    Also, it would help you noticing the selection of a PostgreSQL read-write test that is IO-bound and exchanging it for a read-only test that is CPU/memory bound for a memory test.

    Thanks for all the work over the past 20 years, I am eagerly looking forward to the next 20!

    Comment


    • #3
      Originally posted by jochendemuth View Post
      You clearly state that you tested the configurations with DDR5-4800 and DDR5-5200 memory. I have missed any discussion about why you chose these two sets of memory. Why not DDR5-6000 or DDR5-6400, which are supported on Threadripper platforms or, more appropriately for these CPUs, AMD Ryzen (at least for a 2 DIMM configuration)? Yes, because the new CPUs have not improved the memory controller and faster speeds cannot be supported in 4 DIMM configurations, that are probably more appropriate for EPYC CPUs.
      EPYC 4004 only supports up to DDR5-5200, not anything faster... Installing other DIMMs will just result in DDR5-5200 speeds.
      Michael Larabel
      https://www.michaellarabel.com/

      Comment


      • #4
        Originally posted by Michael View Post

        EPYC 4004 only supports up to DDR5-5200, not anything faster... Installing other DIMMs will just result in DDR5-5200 speeds.
        Insightful. Thanks for sharing.

        Comment


        • #5
          Really wish these benchmarks included "homelab on budget" contender (e.g. Ryzen 7950X) for comparison of value

          Comment


          • #6
            Its honestly shocking how much AMD is ripping Intel in the server space

            Comment


            • #7
              I'm not really surprised faster memory doesn't really do much anymore, especially as caches get larger. I could see DDR5 being sufficient for CPU-only purposes for the better part of a decade. Still a little insufficient for iGPUs but that's not anything additional memory channels couldn't fix.

              Comment

              Working...
              X