Message boards :
Number crunching :
Set of new CPU AstroPulse builds
Message board moderation
Author | Message |
---|---|
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
For those who likes to test on own hardware I decided to make available this set of AstroPulse builds for CPU: https://dl.dropboxusercontent.com/u/60381958/AstroPulse_CPU_different_builds.7z Only few of them will be shipped in new installer but maybe something else would be the best for your particular setup. Binaries are provided as is, for those who familiar with offline benchmarking and app_info.xml creation in hope they could be useful. SETI apps news We're not gonna fight them. We're gonna transcend them. |
cov_route Send message Joined: 13 Sep 12 Posts: 342 Credit: 10,270,618 RAC: 0 |
Nice. What are they built with these days? gcc? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Nice. What are they built with these days? gcc? MSVC 2008 SETI apps news We're not gonna fight them. We're gonna transcend them. |
cov_route Send message Joined: 13 Sep 12 Posts: 342 Credit: 10,270,618 RAC: 0 |
Results for a Phenom II 945. The first result is the 0.41 installer app. |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
Raistmer, thanks for these new apps. Could someone explain how to make a bench test? With which tool & tasks? Thanks. |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
On Core2Duo running Windows XP AP6_win_x86_SSE3_CPU_r2148.exe wins by a very small margin. @Dirk: Get the benchmark package and test WUs from lunatics: http://lunatics.kwsn.net/index.php?module=Downloads;catd=5 Inside the benchmark package is a readme file explaining the usage. Aloha, Uli |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
|
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
Thanks. I never made a bench test run with apps which make a ".wisdom" file. AFAIK, with the 1st & 2nd run (different tasks) of such app this file will be done. 3rd+ run (other tasks) the calculation time is shorter. How should I take this into account in the bench test run case? 1st & 2nd run (different tasks?) just for creation of the ".wisdom" file? I could use this ".wisdom" file from one app also with an other app? Thanks. |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
Best task for comparison is Clean01.wu What is the estimated time to completion for that wu? Aloha, Uli |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Best task for comparison is Clean01.wu Depends on the CPU of course. On my FX its around 300 seconds. With each crime and every kindness we birth our future. |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
Depends on the CPU of course. Thanks, then i'll try again with that wu! :) Aloha, Uli |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Depends on the CPU of course. You should use r_1797 as reference app to make sure what the difference is to previous app. With each crime and every kindness we birth our future. |
Ulrich Metzner Send message Joined: 3 Jul 02 Posts: 1256 Credit: 13,565,513 RAC: 13 |
You should use r_1797 as reference app to make sure what the difference is to previous app. I did exactly that, here is the result: WU : Clean_01LC.wu AP6_win_x86_SSE_CPU_r1797.exe -verbose : Elapsed 397.906 secs CPU 377.453 secs AP6_win_x86_SSE_CPU_r2122_all_opt_archSSE.exe -verbose : Elapsed 391.578 secs, speedup: 1.59% ratio: 1.02x CPU 370.813 secs, speedup: 1.76% ratio: 1.02x AP6_win_x86_SSE2_CPU_r2137_all_opt_archSSE2.exe -verbose : Elapsed 398.625 secs, speedup: -0.18% ratio: 1.00x CPU 379.891 secs, speedup: -0.65% ratio: 0.99x AP6_win_x86_SSE2_CPU_r2137_no_opt_noarch_USE_SSE.exe -verbose : Elapsed 393.906 secs, speedup: 1.01% ratio: 1.01x CPU 374.328 secs, speedup: 0.83% ratio: 1.01x AP6_win_x86_SSE3_CPU_AMD_r2148.exe -verbose : Elapsed 388.875 secs, speedup: 2.27% ratio: 1.02x CPU 369.203 secs, speedup: 2.19% ratio: 1.02x AP6_win_x86_SSE3_CPU_r2137_all_opt_archSSE2.exe -verbose : Elapsed 386.875 secs, speedup: 2.77% ratio: 1.03x CPU 367.547 secs, speedup: 2.62% ratio: 1.03x <- This one now wins. AP6_win_x86_SSE3_CPU_r2148.exe -verbose : Elapsed 387.828 secs, speedup: 2.53% ratio: 1.03x CPU 369.188 secs, speedup: 2.19% ratio: 1.02x Aloha, Uli |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
[quote]Results for a Phenom II 945. The first result is the 0.41 installer app. Would be nice to mention what on Y-axis and on what task bench was done. SETI apps news We're not gonna fight them. We're gonna transcend them. |
cov_route Send message Joined: 13 Sep 12 Posts: 342 Credit: 10,270,618 RAC: 0 |
[quote]Results for a Phenom II 945. The first result is the 0.41 installer app. I used ap_18se08aa_B6_P1_00046_1LC25.wu that I got from here. 2.34% blanked. Default command line switches. All the apps validated strongly similar to the baseline. Y-axis is cpu time in seconds. Might as well mention also the time for the baseline 1797 is wrong. It's an old cached result. I may have tweaked my northbridge since then. The current cpu time for 1797 on this machine is 454.469s. So the slowest of the test apps are about the same as 1797. Here is the full output: https://dl.dropbox.com/s/5xwrpj51x58ew62/13-12_20140214-0013-benchAP_raistmer.txt?dl=1 |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
The current cpu time for 1797 on this machine is 454.469s. So the slowest of the test apps are about the same as 1797. This concurs with our results of inner testing. SETI apps news We're not gonna fight them. We're gonna transcend them. |
MarkJ Send message Joined: 17 Feb 08 Posts: 1139 Credit: 80,854,192 RAC: 5 |
Can someone explain the file names. The version number, SSE, etc I get, but what are the no_opt and the noarch? I am guessing no_opt means no optimisations. Also was there an all_opt for x64? It looks like the only x64 was a vanilla version. Does avx yield anything? BOINC blog |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Still no improvement over r557 on my C2Ds. Both machines are now using AP6_win_x86_SSE3_CPU_r2148.exe and the run-times are slightly longer. Both machines are clocked at 3 GHz... |
cov_route Send message Joined: 13 Sep 12 Posts: 342 Credit: 10,270,618 RAC: 0 |
Still no improvement over r557 on my C2Ds. Both machines are now using AP6_win_x86_SSE3_CPU_r2148.exe and the run-times are slightly longer. Both machines are clocked at 3 GHz... There's a good chance a c2d could be memory bound. |
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
So far I am seeing shorter run times. Not by much but definite shorter run times. I was running SSE r2083 CPU and it was averaging 40000 sec on a 0% blanked task Now with r2148 AMD SSE3 CPU I am seeing 36000 sec on a comparable 0% blanked task. Will keep testing to make sure these observations are correct. Boinc 7.3.2 Pre Release x64 Phenom 2 1100T @ 3.8 NB 2.6 HT 2.6 Michael Miles Team New Mexico |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.