Comparing CPU core productivities

Forums › Cruncher's Corner

Orgil
Orgil
Joined: 29 Oct 05
Posts: 55
Credit: 91242
RAC: 0
12 Dec 2010 2:34:00 UTC
Topic 195506
(moderation:
    )

    I suddenly noticed that single core output of mobile core i5 430 is likely more faster than single core of i7 920 cpu.

    I mean both cpu's processing 2 wu's in 2 threads of a core but mobile core i5 is finishing a wu in 22 000 sek and i7 920 is finishing a wu in 24 000 sek. I checked samples from 2 different members' i7's.

    Does this mean 32nm cpu technology advancement is this much over 45nm tech of desktop level?!

    Here are outputs from mobile core i5 430:

    210411537 89222414 3671338 8 Dec 2010 4:40:11 UTC 10 Dec 2010 5:13:18 UTC Completed and validated 22,945.92 22,636.13 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210353257 89197411 3671338 7 Dec 2010 21:24:52 UTC 9 Dec 2010 6:29:06 UTC Completed and validated 25,622.96 24,539.00 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210353253 89197409 3671338 7 Dec 2010 21:23:47 UTC 9 Dec 2010 6:11:54 UTC Completed and validated 25,426.39 24,357.03 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210326619 89185526 3671338 7 Dec 2010 15:15:29 UTC 8 Dec 2010 21:00:42 UTC Completed and validated 23,620.40 23,023.39 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210326615 89185524 3671338 7 Dec 2010 15:02:23 UTC 8 Dec 2010 20:54:42 UTC Completed and validated 23,612.74 23,017.70 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210178155 89120281 3671338 6 Dec 2010 19:43:14 UTC 8 Dec 2010 14:27:07 UTC Completed and validated 22,471.68 22,184.49 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210178152 89120279 3671338 6 Dec 2010 19:41:02 UTC 8 Dec 2010 14:24:08 UTC Completed and validated 22,439.02 22,148.81 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)
    210000025 89042988 3671338 5 Dec 2010 17:10:13 UTC 7 Dec 2010 22:18:01 UTC Completed and validated 22,370.61 22,064.44 74.95 251.07 Global Correlations S5 HF search #1 v3.04 (S5GCESSE2)

    DanNeely
    DanNeely
    Joined: 4 Sep 05
    Posts: 1364
    Credit: 3562358667
    RAC: 0

    Comparing CPU core productivities

    12 Dec 2010 3:10:51 UTC
    Message 101110
    (moderation:
      )

      AFAIK there aren't any functional changes made during the 45/32nm switch that would affect performance. Looking at your task history I see you appear to be only doing 2 WU's not the 4 that your CPU is capable of. The difference between running hyperthreaded and non-hyperthreaded is probably the explanation.

      Orgil
      Orgil
      Joined: 29 Oct 05
      Posts: 55
      Credit: 91242
      RAC: 0

      I also checked some Xeon

      12 Dec 2010 3:45:46 UTC
      Message 101111
      (moderation:
        )

        I also checked some Xeon E5520 output that its crunch time is 32000sek and that is also 45nm cpu. This Xeon has even same turbo technology: http://ark.intel.com/Product.aspx?id=40200

        I guess all types of cpu's receive only one size/type of wu from e@h?!

        archae86
        archae86
        Joined: 6 Dec 05
        Posts: 3157
        Credit: 7214814931
        RAC: 978164

        As already mentioned, the

        12 Dec 2010 5:16:41 UTC
        Message 101112
        (moderation:
          )

          As already mentioned, the question of whether a particular host ran the work hyperthreaded or not is rather crucial to performance conclusions.

          There is also the question of whether someone has overclocked their system.

          My Westmere system, which has a Xeon E5620 CPU, currently runs hyperthreaded, so eight virtual core, but overclocked to about 3.4 GHz, not the 2.4 GHz reported to BOINC. It is reporting about 22,500 CPU seconds for Global Correlations S5 HF search #1 v3.04 (S5GCESSE2) work. This is a Nehalem type CPU logically, but on 32 nm process, and desktop.

          Just to complicate matters further, of the three RAM channels, one is currently not operative (unsure whether a failed stick of RAM, a bad socket, a failure of the CPU RAM controller, or just a dust particle in the RAM socket). This means it is reporting longer times than it would if working properly.

          Orgil
          Orgil
          Joined: 29 Oct 05
          Posts: 55
          Credit: 91242
          RAC: 0

          Here in top pc category check

          12 Dec 2010 5:46:02 UTC
          Message 101113
          (moderation:
            )

            Here in top pc category check all those Xeons most of them have slower single core/thread speed than mobile core i5: http://einstein.phys.uwm.edu/stats.php - amazing; including that top opteron time.

            transient
            transient
            Joined: 3 Jun 05
            Posts: 62
            Credit: 115835369
            RAC: 0

            If the Xeons run

            12 Dec 2010 9:07:56 UTC
            Message 101114
            (moderation:
              )

              If the Xeons run hyper-threaded and you don't than it is not actually amazing. :D

              Richard Haselgrove
              Richard Haselgrove
              Joined: 10 Dec 05
              Posts: 2143
              Credit: 2951540261
              RAC: 690030

              There is also the question of

              12 Dec 2010 9:47:03 UTC
              Message 101115
              (moderation:
                )

                There is also the question of which version of the application is in use. I recently reduced the processing time of a X3430 running S5GC1HF from ~28,000 seconds to ~19,000 seconds. Nothing to do with the hardware, just a software versioning problem.

                Orgil
                Orgil
                Joined: 29 Oct 05
                Posts: 55
                Credit: 91242
                RAC: 0

                Well until yesterday my

                12 Dec 2010 14:44:48 UTC
                Message 101116
                (moderation:
                  )

                  Well until yesterday my understanding was mostly mobile cpu productivity is a lot lower than desktop 45nm cpu but in per core productivity wise the real comparisions prove that mobile 32nm cpu is shockingly powerful over many Xeons and core quads (because they are 45nm... maybe). I hope you verified with this link: http://einstein.phys.uwm.edu/top_hosts.php

                  (Well I believe that in e@h all cpu get one size/type of wu's. if not then that is different question.)

                  ExtraTerrestrial Apes
                  ExtraTerrestria...
                  Joined: 10 Nov 04
                  Posts: 770
                  Credit: 576513560
                  RAC: 192520

                  The average of 7 of the WUs

                  12 Dec 2010 21:31:05 UTC
                  Message 101117
                  (moderation:
                    )

                    The average of 7 of the WUs you linked to (I missed one) is 23.5 ks, not 22 ks. And you should only compare runtimes of WUs giving the same amount of credits - I can't check this for your host.

                    And then, as others said, the actual clock speed the CPU is running at is vey important. If you don't compare at similar clocks or at least know the clocks you're not comparing architectures but mere clock speeds. This is especially true given the existence of Turbo and the fact that BOINC doesn't report the real CPU clock speed.

                    It's also crucial which apps were run: for Einstein it's probably best to compare only "running Einstein on all logical cores", as anything else (not using HT, running a project mix) should probably make the run times faster (not neccessarily throughput, though).

                    MrS

                    Scanning for our furry friends since Jan 2002

                    Orgil
                    Orgil
                    Joined: 29 Oct 05
                    Posts: 55
                    Credit: 91242
                    RAC: 0

                    In very first post I showed

                    12 Dec 2010 21:45:49 UTC
                    Message 101119
                    (moderation:
                      )

                      In very first post I showed you cpu time of mobile core i5 430 now please compare with top host pc's from http://einstein.phys.uwm.edu/top_hosts.php :

                      From top host list this is output of 3-d pc: Intel(R) Xeon(R) CPU X5650 ; Darwin

                      210947396 88366059 11 Dec 2010 20:43:02 UTC 12 Dec 2010 15:10:07 UTC Completed and validated 29,450.98 29,329.26 74.96 251.13 Global Correlations S5 HF search #1 v5.06
                      210937133 88337388 11 Dec 2010 23:16:02 UTC 12 Dec 2010 21:12:49 UTC Completed and validated 29,450.32 29,342.61 74.96 251.13 Global Correlations S5 HF search #1 v5.06
                      210928007 88346177 11 Dec 2010 20:41:58 UTC 12 Dec 2010 15:10:07 UTC Completed and validated 29,436.97 29,330.44 74.96 251.13 Global Correlations S5 HF search #1 v5.06
                      210924028 89384773 11 Dec 2010 23:14:57 UTC 12 Dec 2010 21:12:49 UTC Completed, waiting for validation 29,392.10 29,275.44 74.96 pending Global Correlations S5 HF search #1 v5.06
                      210923513 89384849 11 Dec 2010 23:13:40 UTC 12 Dec 2010 21:12:49 UTC Completed, waiting for validation 29,385.05 29,266.10 74.96 pending Global Correlations S5 HF search #1 v5.06
                      210923238 89443683 11 Dec 2010 22:58:51 UTC 12 Dec 2010 15:26:32 UTC Completed and validated 29,398.52 29,281.26 74.96 251.13 Global Correlations S5 HF search #1 v5.06
                      210910245 89438123 11 Dec 2010 15:03:03 UTC 12 Dec 2010 13:00:57 UTC Completed and validated 29,352.38 29,244.70 74.96 251.13 Global Correlations S5 HF search #1 v5.06
                      210909740 89437891 11 Dec 2010 14:58:25 UTC 12 Dec 2010 7:22:22 UTC Completed, waiting for validation 29,339.75 29,219.30 74.96 pending Global Correlations S5 HF search #1 v5.06

                      This is 4th xeon with Linux : Intel(R) Xeon(R) CPU E5645

                      210999581 89477018 12 Dec 2010 4:24:43 UTC 12 Dec 2010 20:54:19 UTC Completed, waiting for validation 30,028.99 29,937.17 75.00 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210999573 89477014 12 Dec 2010 4:07:21 UTC 12 Dec 2010 20:44:57 UTC Completed and validated 29,878.49 29,786.40 75.00 251.24 Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210999570 89477012 12 Dec 2010 4:21:35 UTC 12 Dec 2010 20:47:05 UTC Completed and validated 29,997.07 29,903.85 75.00 251.24 Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210999562 89477008 12 Dec 2010 4:06:16 UTC 12 Dec 2010 20:44:57 UTC Completed and validated 30,043.68 29,950.02 75.00 251.24 Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210996802 89475783 12 Dec 2010 4:02:06 UTC 12 Dec 2010 20:44:57 UTC Completed, waiting for validation 30,227.85 30,135.62 75.00 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210996784 89475774 12 Dec 2010 3:28:22 UTC 12 Dec 2010 20:05:04 UTC Completed, waiting for validation 29,946.06 29,846.64 75.00 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210995470 89475182 12 Dec 2010 3:09:26 UTC 12 Dec 2010 19:40:52 UTC Completed, waiting for validation 29,869.68 29,775.88 75.00 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)

                      This is 9th pc core i7 980: win7

                      210718583 88248548 10 Dec 2010 6:30:05 UTC 11 Dec 2010 1:47:33 UTC Completed and validated 19,703.36 17,610.47 74.96 251.10 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210717452 89354020 10 Dec 2010 6:29:59 UTC 11 Dec 2010 7:06:54 UTC Completed and validated 19,318.32 17,481.38 74.96 251.10 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210716567 89353618 10 Dec 2010 6:30:35 UTC 10 Dec 2010 11:30:31 UTC Completed, waiting for validation 17,905.68 16,548.51 74.96 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210716563 89353616 10 Dec 2010 6:30:35 UTC 10 Dec 2010 11:32:52 UTC Completed, waiting for validation 18,029.35 16,530.35 74.96 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210716560 89353614 10 Dec 2010 6:30:35 UTC 10 Dec 2010 11:32:52 UTC Completed, waiting for validation 17,980.55 16,554.16 74.87 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210716557 89353613 10 Dec 2010 6:30:33 UTC 10 Dec 2010 11:31:43 UTC Completed, waiting for validation 17,956.21 16,558.55 74.96 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)

                      This is 30th pc: Intel(R) Xeon(R) CPU X5670 ; Linux ; maybe OC'd?!

                      210782561 89355473 12 Dec 2010 6:48:04 UTC 12 Dec 2010 16:52:41 UTC Completed and validated 15,896.89 15,888.40 74.99 251.23 Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210782426 89382275 12 Dec 2010 6:20:34 UTC 12 Dec 2010 16:25:06 UTC Completed, waiting for validation 15,832.87 15,825.12 74.99 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210782425 89382274 12 Dec 2010 5:11:32 UTC 12 Dec 2010 15:33:44 UTC Completed, waiting for validation 15,842.80 15,835.52 74.99 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210782419 89382242 12 Dec 2010 4:27:20 UTC 12 Dec 2010 15:00:41 UTC Completed, waiting for validation 16,039.21 16,029.22 74.99 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210782346 89351020 12 Dec 2010 3:56:47 UTC 12 Dec 2010 15:00:41 UTC Completed, waiting for validation 15,950.16 15,943.12 74.99 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)
                      210782259 89351018 12 Dec 2010 3:46:24 UTC 12 Dec 2010 13:50:07 UTC Completed, waiting for validation 16,281.29 16,271.66 74.99 pending Global Correlations S5 HF search #1 v1.06 (S5GCESSE2)

                      43rd pc core i7 930: Linux

                      208526910 88395510 28 Nov 2010 3:31:23 UTC 12 Dec 2010 7:26:01 UTC Completed and validated 20,469.00 20,432.36 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208526905 88395509 28 Nov 2010 3:30:18 UTC 12 Dec 2010 4:34:20 UTC Completed and validated 20,503.46 20,465.78 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208526824 88395473 28 Nov 2010 3:29:13 UTC 12 Dec 2010 4:18:22 UTC Completed and validated 20,431.34 20,395.94 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208526655 88395404 28 Nov 2010 3:28:07 UTC 12 Dec 2010 3:35:40 UTC Completed and validated 20,123.93 20,090.84 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208526651 88395402 28 Nov 2010 3:27:04 UTC 12 Dec 2010 3:22:27 UTC Completed and validated 20,432.46 20,401.78 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208482779 88210030 28 Nov 2010 3:25:54 UTC 12 Dec 2010 3:09:59 UTC Completed and validated 20,473.54 20,445.06 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208466189 88370484 28 Nov 2010 3:24:50 UTC 12 Dec 2010 2:52:05 UTC Completed and validated 20,131.76 20,095.79 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)
                      208466187 88370483 28 Nov 2010 3:23:47 UTC 12 Dec 2010 2:35:25 UTC Completed and validated 20,448.31 20,398.93 75.00 251.24 Global Correlations S5 HF search #1 v1.04 (S5GCESSE2)

                      What I understand is this is clear comparison of unit WU processing capabilities of selected cpu's. I found that there is even core i5 430 cpu with linux is doing 16 000 average.

                      NOW CORE i7 920:

                      pc 2248121 rank 541th: win2008

                      210969873 89463994 11 Dec 2010 23:50:07 UTC 12 Dec 2010 15:04:08 UTC Completed, waiting for validation 24,448.89 24,421.52 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210943889 89452753 12 Dec 2010 0:02:37 UTC 12 Dec 2010 20:14:39 UTC Completed and validated 24,486.57 24,448.21 74.99 251.21 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210943882 89452750 11 Dec 2010 23:58:29 UTC 12 Dec 2010 20:14:39 UTC Completed, waiting for validation 24,478.08 24,455.11 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210937233 89449823 11 Dec 2010 18:49:53 UTC 12 Dec 2010 15:04:08 UTC Completed, waiting for validation 24,463.09 24,443.39 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210923288 89443708 11 Dec 2010 17:17:33 UTC 12 Dec 2010 13:22:40 UTC Completed and validated 24,454.55 24,423.70 74.99 251.21 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210923284 89443706 11 Dec 2010 17:16:27 UTC 12 Dec 2010 13:22:40 UTC Completed, waiting for validation 24,458.45 24,433.19 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210915513 89436426 11 Dec 2010 17:15:21 UTC 12 Dec 2010 13:22:40 UTC Completed and validated 24,474.32 24,448.10 74.99 251.21 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)

                      pc 3189962 rank 554th, win7; :

                      210971050 89464542 11 Dec 2010 23:56:36 UTC 12 Dec 2010 13:17:51 UTC Completed, waiting for validation 25,377.84 24,925.09 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210970990 89464518 11 Dec 2010 23:55:32 UTC 12 Dec 2010 13:17:51 UTC Completed, waiting for validation 25,344.13 24,893.53 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210959790 89459585 11 Dec 2010 23:20:27 UTC 12 Dec 2010 13:17:51 UTC Completed, waiting for validation 25,348.77 24,906.46 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210956203 89457971 11 Dec 2010 21:35:01 UTC 12 Dec 2010 10:55:31 UTC Completed, waiting for validation 25,372.60 24,941.49 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210951967 89456092 11 Dec 2010 20:57:57 UTC 12 Dec 2010 10:42:45 UTC Completed and validated 25,252.66 24,909.18 74.99 251.22 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210943799 89452713 11 Dec 2010 20:39:49 UTC 12 Dec 2010 10:42:45 UTC Completed, waiting for validation 25,369.28 24,912.25 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210927269 89445496 11 Dec 2010 19:19:48 UTC 12 Dec 2010 9:13:36 UTC Completed, waiting for validation 25,469.49 24,854.19 74.99 pending Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210514276 89266687 8 Dec 2010 20:07:07 UTC 12 Dec 2010 9:13:36 UTC Completed and validated 25,436.27 24,831.49 74.99 251.22 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)
                      210509526 89264600 8 Dec 2010 18:47:55 UTC 12 Dec 2010 6:20:52 UTC Completed and validated 25,520.64 24,863.33 74.99 251.22 Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)

                      ExtraTerrestrial Apes
                      ExtraTerrestria...
                      Joined: 10 Nov 04
                      Posts: 770
                      Credit: 576513560
                      RAC: 192520

                      RE: I mean both cpu's

                      13 Dec 2010 9:21:11 UTC
                      Message 101120
                      (moderation:
                        )

                        Quote:
                        I mean both cpu's processing 2 wu's in 2 threads of a core but mobile core i5 is finishing a wu in 22 000 sek and i7 920 is finishing a wu in 24 000 sek. I checked samples from 2 different members' i7's.

                        Apparently you didn't understand what we've been saying. Did your i5 run 4 Einstein WUs with HT active? If not then you're comparing apples and oranges.

                        And at which clock speed was your CPU running (including modifications by Turbo)? And the problem with selecting comparison numbers from the top hosts list is that we don't know under which circumstances these have been generated (i.e. clock speed and which projects were run, HT on/off).

                        24 ks looks about right for a stock i7 920 with Turbo and HT on, though. Maybe a bit less if running purely GC WUs.

                        MrS

                        Scanning for our furry friends since Jan 2002

                        Comment viewing options

                        Select your preferred way to display the comments and click "Save settings" to activate your changes.

                        Forums › Cruncher's Corner