Announcement

Collapse
No announcement yet.

AMD's HD7970 is ready for action! The most effiency and fastest card on earth.

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

  • #76
    Originally posted by efikkan View Post
    For workstations high performance per core is a requirement
    try the Opteron 6204. right now its the AMD cpu with the highest single-thereat performance.
    Its a Quatcore with 3,3ghz and 16MB L3 cache and quatchannel 256bit ram per cpu.
    this means a full 64bit ram channel per core. and full 4,5mb cache per core.

    for a workstation you can get a dualsocket mainboard with this kind of cpu.

    but yes its expensive: 474,00 per opteron 6204 cpu.

    Comment


    • #77
      Originally posted by Qaridarium View Post
      try the Opteron 6204. right now its the AMD cpu with the highest single-thereat performance.
      Its a Quatcore with 3,3ghz and 16MB L3 cache and quatchannel 256bit ram per cpu.
      this means a full 64bit ram channel per core. and full 4,5mb cache per core.

      for a workstation you can get a dualsocket mainboard with this kind of cpu.

      but yes its expensive: €474,00 per opteron 6204 cpu.
      i found benchmark: http://www.spec.org/cpu2006/results/res2011q4/


      Opteron 6204 --> 151/8cores =18,87
      Opteron 6282 SE --> 355/32cores=11.09

      The Opteron is single threated 71,5% faster than the Opteron 6282 SE.

      the amd fx8150 makes 78,8 points with 8cores

      this makes the opteron 6204 single threated 91,6% faster .
      Last edited by Qaridarium; 01-01-2012, 11:08 PM.

      Comment


      • #78
        Btw. i did some benchmarks i7-2600 (4cores,ht,4gb ram) vs 2x Opteron 6180 SE (24 cores,32 gb ram) and the i7 kicks ass in most benchmarks. When you do a short one, like compiling mplayer2 then the i7 is twice as fast, when you do a longer one, then running the complete kernel compile in ram the opterons need 17min and due to missing ram 21 min for i7 on simple 500 gb hd a few years old. 7zip benched would win the opteron, but not much more real live benchmarks - or the diff is a joke compared to the price. Btw. i did not compile a minimal kernel, but a full u release 3.0.0-15 kernel on 64 bit.

        Comment


        • #79
          Originally posted by Kano View Post
          Btw. i did some benchmarks i7-2600 (4cores,ht,4gb ram) vs 2x Opteron 6180 SE (24 cores,32 gb ram) and the i7 kicks ass in most benchmarks. When you do a short one, like compiling mplayer2 then the i7 is twice as fast, when you do a longer one, then running the complete kernel compile in ram the opterons need 17min and due to missing ram 21 min for i7 on simple 500 gb hd a few years old. 7zip benched would win the opteron, but not much more real live benchmarks - or the diff is a joke compared to the price. Btw. i did not compile a minimal kernel, but a full u release 3.0.0-15 kernel on 64 bit.
          I don't get your point first of all the 2010 old Opteron 61xx is not the tropic.
          second of all the top speed is not the question there are many many faster systems.
          and the last joke: price dosn't matter because he ask about a workstation Intel Xeon computer and not a cheap super market I7-2600 price breaker.

          Show him a Workstation with Coreboot (not bios/UEFI) with TOP speed at SINGLE-TREATED performance.

          i just answers his question! your writing just do not answer this question!

          Comment


          • #80
            It's my understanding that Xeon E5-1620 would outperform Opteron 6204 in most single threaded uses, even when considering the slight overhead introduced by EFI. My personal concern regarding EFI is primarily regarding security and protecting company property. With EFI we might see the first real OS independent viruses, and with all this "surveillance" features embedded into EFI I fear this would become a target for economical criminality. So for servers I would consider this security very important, and maybe a little less important for the workstations, but that's just my opinion. One of the reasons for the requirement of high single threaded performance is graphical development. With Kepler's increased programmability the CPU might become less of a bottleneck tough...

            Comment


            • #81
              @Qaridarium

              The speed is exactly the same as it is the same silicon, just with different features enabled. So the benchmarks are absolutely correct. I had only ssh access to that Opterons (running Kanotix live) but i dont know anybody with the newer ones, do you? The gfx part is maybe different but thats not tested in that benchmark. You mainly need the Xeon flavour when you want to use ECC together with a workstation chipset like C206, the cpu itself would run in any desktop s1155 board as well, you just can not use ECC.

              http://ark.intel.com/compare/52214,52213,52277

              All i can say is: too many cores hurt performance as there is much more work to synchronize. A smaller source code like mplayer2 and the performance of quad with higher single core performance beats 24 cores by 100% difference! All compile tests resulted in debian packages, so the time where configure was done and creating deb packages was included as well, i did the tests with cached depends in pbuilder for the kernel, so dl speed differences are not tested. It does not help much when you try artificial workloads, a kernel so minimal that you can compile (but not package) within 60s is just useless when it has to run on lots of systems.

              Comment


              • #82
                Originally posted by efikkan View Post
                It's my understanding that Xeon E5-1620 would outperform Opteron 6204 in most single threaded uses, even when considering the slight overhead introduced by EFI. My personal concern regarding EFI is primarily regarding security and protecting company property. With EFI we might see the first real OS independent viruses, and with all this "surveillance" features embedded into EFI I fear this would become a target for economical criminality. So for servers I would consider this security very important, and maybe a little less important for the workstations, but that's just my opinion. One of the reasons for the requirement of high single threaded performance is graphical development. With Kepler's increased programmability the CPU might become less of a bottleneck tough...
                are there any benchmarks for your claim: "It's my understanding that Xeon E5-1620 would outperform Opteron 6204 in most single threaded uses" ????

                and you ask for Coreboot support and the Opteron 6204 boards do have coreboot support!

                i just give you an alternative viewpoint why not buy a single Opteron 6204 workstation and test it and benchmark it?

                or pay michael/phoronix to do this.

                in my point of view an opteron solution is cheaper than a Xeon Solution.

                Comment


                • #83
                  Originally posted by Kano View Post
                  @Qaridarium

                  The speed is exactly the same as it is the same silicon, just with different features enabled. So the benchmarks are absolutely correct. I had only ssh access to that Opterons (running Kanotix live) but i dont know anybody with the newer ones, do you? The gfx part is maybe different but thats not tested in that benchmark. You mainly need the Xeon flavour when you want to use ECC together with a workstation chipset like C206, the cpu itself would run in any desktop s1155 board as well, you just can not use ECC.

                  http://ark.intel.com/compare/52214,52213,52277

                  All i can say is: too many cores hurt performance as there is much more work to synchronize. A smaller source code like mplayer2 and the performance of quad with higher single core performance beats 24 cores by 100% difference! All compile tests resulted in debian packages, so the time where configure was done and creating deb packages was included as well, i did the tests with cached depends in pbuilder for the kernel, so dl speed differences are not tested. It does not help much when you try artificial workloads, a kernel so minimal that you can compile (but not package) within 60s is just useless when it has to run on lots of systems.
                  o man you don't get it! the Opteron 6204 fit in your claim because its only a quatcore and not a 16core. AND the question was a Workstation CPU with support for Coreboot!

                  your only problem is you don't have a system do make your benchmarks,

                  Comment


                  • #84
                    It's based on bechmarks of Sandy Bridge vs Bulldozer (consumer models), I know there are differences between the consumer models and the enterprise models.

                    I've failed to find any motherboards from Tyan or Supermicro with coreboot, can you please show me one? I don't think anything will be ordered for another six months, in case anything show up.

                    Comment


                    • #85
                      Originally posted by efikkan View Post
                      It's based on bechmarks of Sandy Bridge vs Bulldozer (consumer models)
                      the AMD FX8150 is the consumer model and the Opteron 6204 is 91% faster in singlet-treated tasks.

                      you claim that the xenon is faster based on the benchmarks with the 91% slower consumer version.

                      in fact you are wrong here!

                      Originally posted by efikkan View Post
                      , I know there are differences between the consumer models and the enterprise models.
                      yes BIG differences... FX8150=8mb L3 cache for 8 cores and the Opteron 6204=16mb L3 cache for 4 cores.

                      but you prefer to claim based on the "consumer version"


                      Originally posted by efikkan View Post
                      I've failed to find any motherboards from Tyan or Supermicro with coreboot, can you please show me one?
                      you can watch the core-boot compatibly list all chip-sets for the Opterons are supported.
                      this means you can just buy a motherboard and flash your coreboot version on it.

                      there are no known pre installed mainboards.


                      Originally posted by efikkan View Post
                      I don't think anything will be ordered for another six months, in case anything show up.
                      if you buy stuff from Intel without a competitor without a alternative then you pay the highest price but this is clueless.

                      professionals always check alternatives.

                      Comment


                      • #86
                        I don't think you got my point. First of all, I'm not set on buying Intel, but selecting the best overall solution, in fact I'll be willing to sacrifice a little performance if needed to get a solution with coreboot, that's why I want to discuss it. Secondly, I did not claim Xeon X5 was faster than Opteron. If you have a link showing the performance of Opteron 6204 vs the alternatives I'll be more than happy to consider it.

                        Comment


                        • #87
                          Originally posted by efikkan View Post
                          I don't think you got my point. First of all, I'm not set on buying Intel, but selecting the best overall solution, in fact I'll be willing to sacrifice a little performance if needed to get a solution with coreboot, that's why I want to discuss it. Secondly, I did not claim Xeon X5 was faster than Opteron. If you have a link showing the performance of Opteron 6204 vs the alternatives I'll be more than happy to consider it.
                          ok!... this is the only benchmark link i found: http://www.spec.org/cpu2006/results/res2011q4/

                          sorry there are no other benchmarks in the web.

                          AMD should send a Opteron 6204 system to michael to bring more benchmarks to the WEB.

                          Comment


                          • #88
                            As far as I can see, the Xeon X5680 scores at 241/12=20. I'm curious how this compares to general x86 performance.

                            Comment


                            • #89
                              Originally posted by efikkan View Post
                              As far as I can see, the Xeon X5680 scores at 241/12=20. I'm curious how this compares to general x86 performance.
                              i don't know. sorry. but 18,87 compared to 20 is not so bad for amd !

                              and the Xeon X5680 is much more expensive 1.739 LOL!!!

                              you really compare a 470 cpu to an 1739 cpu LOL!!!..

                              the amd cpu is 6% slower (single-threated) but you save 1269 LOL!

                              (Ironic)ALL HAIL to Intel!

                              Comment


                              • #90
                                You know I was thinking about Xeon E5-1620 ($294 4 cores, 3.6 GHz (3.9 GHz turbo), 4x256 kB L2, 10 MB L3 (2.5 MB per core)), suggesting this might perform well against Opteron 6204, since X5680 (3.33 GHz, 6x256 kB L2, 12 MB L3 (2 MB per core)) does. Correct me if my reasoning is wrong here.

                                Comment

                                Working...
                                X