Announcement

Collapse
No announcement yet.

Hands On & Initial Benchmarks With An Ampere eMAG 32-Core ARM Server

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

  • #21
    That is some really nice hardware, I hope FOSS firmware can exist for it one day.

    Something that caught my eye is that the BMC is on its own board. Should the end user not require it, it could always be removed for a price discount (if a configuration without the BMC is ever produced), but most importantly to avoid having unnecessary superprivileged hardware.

    Originally posted by torsionbar28 View Post
    The 32 cores, 8 channels of DDR4, and 42 PCIe lanes puts it firmly in Threadripper/EPYC territory (on paper anyways) and solidly ahead of anything from intel. It's pretty amazing they managed all this in just 125w TDP.
    AFAIK, ARM cores perform less than x86 counterparts, though they have a lower TDP. Considering that, it's not so impressive.

    Comment


    • #22
      Originally posted by juanrga View Post

      ThunderX2 has been tested by ServeTheHome on server worklaods and by Bristol University people on HPC workloads

      Interesting results from the Bristol report. Not sure why STH has any more "favor" on ARM than anyone else.

      No one is buying a Cavium Thunder X2 to run HPC in their house.

      Comment


      • #23
        Thunder X2 is 32 core (96 threads triple hyperthreaded). not sure why youre saying its 96 core.

        Comment


        • #24
          Originally posted by juanrga View Post

          ThunderX2 has been tested by ServeTheHome on server worklaods and by Bristol University people on HPC workloads

          Thanks a lot for this! I remember that the Bristol University and the ISAMBARD HPC system were also mentioned as a test platform with important implications for the European Processor Initiative. By the way, there were some important news about this European HPC ARM64 initiative some weeks ago: https://www.youtube.com/watch?v=1FX9XdNwTWw

          It was announced that they go the ARM route first, but will make the switch to RISC-V later. I hope that we will see a more open source related software initiative from the European institutions along the way. To this day the European Commission is heavily dependent on Microsoft.

          Comment


          • #25
            Originally posted by Th3Fanbus View Post
            AFAIK, ARM cores perform less than x86 counterparts, though they have a lower TDP. Considering that, it's not so impressive.
            What is the x86 counterpart of this chip? With zero data in hand, I think you'll have a difficult time defending your "not so impressive" assertion.

            Comment


            • #26
              Originally posted by torsionbar28 View Post
              What is the x86 counterpart of this chip? With zero data in hand, I think you'll have a difficult time defending your "not so impressive" assertion.
              I was speaking about ARM CPU cores vs. x86 CPU cores in general, not about a specific chip. Adding to that, I did not "assert" anything. This is all based on what I have seen. OTOH, I have not experienced the performance of any ARMv8 chip, so I might be somewhat outdated.
              If you do have data, feel free to share it.

              Comment


              • #27
                Originally posted by Michael View Post

                As written in the article, I don't have any ThunderX2 access. Even with the X1, it was just remote access for a short time.

                I did try getting access to X2 from Cavium, but they claimed PTS is biased against ARM which is clearly just crap.... Considering it's just collecting interesting open-source/upstream benchmarks and can be seen from the great Ampere performance here, etc.
                There is a Gigabyte made board with a mATX form factor and a single socket 48 core Cavium TX2, appears to be similar to the Talos Blackbird in format.

                I am reaching out to a distributor to see if I can acquire one. If so I will arrange access. If the distributor quotes me some ungodly amount for ordering just "1" well.....all I can say is I tried. I think Starship has gone through some these same discussions with distributors.

                I finished my research on Ampere today and all I can say is I am very excited what they are doing.

                - Execs formerly with Intel
                - Engineers recruited away from the Qualcomm Centriq team
                - Former Intel engineer who developed the Pentium Pro
                - Well funded via Carlyle Group
                - Lenovo is their manufacturing partner
                - CPU based on AMCC X-Gene 3, which they purchased the IP directly.

                There is a good story @ Business Insider about how the Ampere CEO got a lot of "no thanks" phone calls when starting up the company.

                With Centriq near death, the timing for Ampere is perfect.

                Comment


                • #28
                  Originally posted by LukePoga View Post
                  Thunder X2 is 32 core (96 threads triple hyperthreaded). not sure why youre saying its 96 core.
                  The TX2 core (formerly codename "Vulcan") has up to four-way multithreading (ST, SMT2, and SMT4). So the 32 core SKU can run a maximum of 256 threads.

                  Comment


                  • #29
                    Originally posted by Th3Fanbus View Post
                    AFAIK, ARM cores perform less than x86 counterparts, though they have a lower TDP.
                    That is a myth that refuses to die.

                    Comment


                    • #30
                      Originally posted by ms178 View Post
                      It was announced that they go the ARM route first, but will make the switch to RISC-V later.
                      Are you sure about that? The last notice I have is that they will use ARM for the central processor and RISC-V for the accelerators.

                      Comment

                      Working...
                      X