Announcement

Collapse
No announcement yet.

AMD EPYC 7F52 Linux Performance - AMD 7FX2 CPUs Further Increasing The Fight Against Intel Xeon

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

  • AMD EPYC 7F52 Linux Performance - AMD 7FX2 CPUs Further Increasing The Fight Against Intel Xeon

    Phoronix: AMD EPYC 7F52 Linux Performance - AMD 7FX2 CPUs Further Increasing The Fight Against Intel Xeon

    AMD today is announcing three new EPYC 7002 "Rome" SKUs in the form of the 7F32, 7F52, and 7F72 processors. The AMD 7F52 processors we have been recently testing and offers some impressive performance potential as while it's a 16-core / 32-thread part it offers an impressive 256MB L3 cache (16MB per core). Here are our initial Linux benchmarks of the AMD EPYC 7F52 in 1P and 2P configurations up against various AMD EPYC and Intel Xeon processors.

    http://www.phoronix.com/vr.php?view=29080

  • #2
    I believe you mixed up the pricing.

    A strong showing, as expected.

    Edit: Oh crap, you didn't. The 16 core really is more expensive than the 24 core part. Apparently because of more cache and chiplets.

    Edit2: I just realized, there's no perf/W graph. I hope Michael is saving them for further reviews.
    Last edited by bug77; 04-14-2020, 03:43 PM.

    Comment


    • #3
      Wow. That 7F32 looks to be a nice CPU to double up as a workstation/gaming CPU. Very nice, AMD.

      Comment


      • #4
        Benchmarks doesn't matter

        Comment


        • #5
          Finally ... I was asking about these last July but AMD was not yet ready to quote them. But now .. I'll just wait for Milan

          Comment


          • #6
            Could you include performance-per-dollar figures too? It's hard to keep track of it manually with so many parts involved.

            Comment


            • #7
              These are not meant to be competitive in raw perf/$. They only become competitive when used by some expensive proprietary software with per-core licencing model.

              Comment


              • #8
                Originally posted by pegasus View Post
                These are not meant to be competitive in raw perf/$. They only become competitive when used by some expensive proprietary software with per-core licencing model.
                Except for the workloads out there that simply don't scale well from 16 cores / 32 threads to 64 cores / 128 threads.

                Comment


                • #9
                  Sounds like those workloads would benefit from a code reorg / rewrite then. Maybe money could be better spent on people than on hw ...

                  Comment


                  • #10
                    Kinda weird to see so much cache on so few cores.

                    Also weird for me to say 16 cores is "so few" haha.

                    Comment

                    Working...
                    X