Benchmark package for Asteroids@home

Message boards : Number crunching : Benchmark package for Asteroids@home

1 · 2 · Next

Author	Message
BilBg Send message Joined: 19 Jun 12 Posts: 221 Credit: 623,640 RAC: 0	Message 3080 - Posted: 21 May 2014, 19:39:37 UTC Benchmark package for Asteroids@home (for Windows, 'composed' by me) - contains all the current stock applications + FMA4 - contains a few real WUs + the same WUs shortened by me (compute for ~10x less time) Get it from this link (Bench_Asteroids_BilBg.7z - 1.78 MB): http://www38.zippyshare.com/v/98596921/file.html Short instructions: - uncompress the package - run MBbench210_Asteroids.cmd - at the end of test look at 'Quick timetable' Long instructions, notes, warnings: - WARNING: Never get anything from CNET (download.cnet.com) ! - if you need 7-Zip get it from http://www.7-zip.org/ - read MBbench_README.txt - BOINC will be auto-paused during testing - do not do any 'heavy' computing (7-Zip compressing, video encoding, gaming, ...) during the test to not skew the results - test log can be found (is saved) in ...\MBbench_Asteroids\Testdatas\ directory - don't mind the lines that say "Result : Strongly similar, Q= ..." - this info was intended for SETI@home and may/will be wrong for Asteroids@home * Before starting the test: -- move apps (.exe .dll files) in/out ...\Science_apps\Reserve\ directory (to choose which of them to run/benchmark) - you can have many apps in Science_apps\ directory -- move WUs in/out ...\TestWUs\Reserve\ directory (to choose which of them to use for benchmark) - you can have many WUs in TestWUs\ directory --- you can copy any of your Asteroids@home real WUs to TestWUs\ directory, rename them (the copies) with .wu extension and they will be used in the test - every app in Science_apps\ will be run/tested with every WU in TestWUs\ directory - by default (if you just uncompress the package and do not move files) only two apps will be tested with one short WU - I'm not sure if I used proper method to make short WUs - project admins have to check/say -- my method: using hex editor (HxD) changed the second number on the first line from 0.5 to 5.0 This Benchmark package is based on MBbench 2.10 (made by 'Lunatics' for SETI@home) http://lunatics.kwsn.net/index.php?module=Downloads;catd=5 http://lunatics.kwsn.net/index.php?module=Downloads;sa=dlview;id=387 My log: MBbench210_Asteroids.cmd ====================================== 1 testWU(s) found (input_22147_73_short.wu) 1 reference science app(s) found (period_search_10210_windows_intelx86__sse2.exe -verb -nog) 1 science app(s) found (period_search_10210_windows_intelx86__sse3.exe -verb -nog) ====================================== period_search_10210_windows_intelx86__sse2.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86__sse2.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 19:46:54.751 Ended at : 19:53:59.876 Result : stored as ref for validations. 424.281 secs Elapsed 417.078 secs CPU time [ stderr ] 19:46:55 (4184): Can't open init data file - running in standalone mode 19:53:57 (4184): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_intelx86__sse3.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86__sse3.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 19:54:03.345 Ended at : 20:01:05.548 421.906 secs Elapsed 414.891 secs CPU time Speedup : 0.52% Ratio : 1.01x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 19:54:04 (4040): Can't open init data file - running in standalone mode 20:01:03 (4040): called boinc_finish [ /stderr ] ------------ Quick timetable WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 424.281 secs CPU 417.078 secs period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 421.906 secs, speedup: 0.56% ratio: 1.01x CPU 414.891 secs, speedup: 0.52% ratio: 1.01x ------------ - ALF - "Find out what you don't do well ..... then don't do it!" :) ID: 3080 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3083 - Posted: 22 May 2014, 2:03:26 UTC - in response to Message 3080. Hello BilBg, thanks a lot for the bench package. Should help a lot to get some more reliable comparisons. I was just looking for something like that. :) The usage is fairly straightforward IMHO. However, have you tried to run the bench with more than two science apps (reference and another) in one run? I tried to run all apps with your default short workunit on a Win7 64bit system with AMD FX-8350 & GTX Titan and I guess something went wrong with the second GPU bench and all CPU benches after the reference and the second app. I have a hunch that some leftover from the previous run interfered here, could that be? The script MBbench210_Asteroids.cmd is fairly complex to me, although I should be able to understand it, so it might take me some time to find the reason here. So help is appreciated. :) Here's the quick timetable. The results with an elapsed time of 2.something secs appear to be bogus. WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 281.144 secs CPU 278.337 secs period_search_10112_windows_intelx86__cuda55.exe -verb -nog : Elapsed 126.292 secs, speedup: 55.08% ratio: 2.23x CPU 2.465 secs, speedup: 99.11% ratio: 112.92x period_search_10112_windows_x86_64__cuda55.exe -verb -nog : Elapsed 2.278 secs, speedup: 99.19% ratio: 123.42x CPU 0.234 secs, speedup: 99.92% ratio: 1189.47x period_search_10210_windows_intelx86.exe -verb -nog : Elapsed 578.662 secs, speedup: -105.82% ratio: 0.49x CPU 575.800 secs, speedup: -106.87% ratio: 0.48x period_search_10210_windows_intelx86__avx.exe -verb -nog : Elapsed 2.044 secs, speedup: 99.27% ratio: 137.55x CPU 0.031 secs, speedup: 99.99% ratio: 8978.61x period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64__avx.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64__sse2.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 2.044 secs, speedup: 99.27% ratio: 137.55x CPU 0.031 secs, speedup: 99.99% ratio: 8978.61x The result file from the first run is included below. EDIT: Just finished a second run with only one GPU app and one CPU app and this time it worked. So there is definitely a problem with running two GPU apps or two CPU apps (not including the reference run) after each other, at least on my system. And it is not caused be the apps themselves, they seem to be ok. Here's the quick timetable from the second run. WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 281.144 secs CPU 278.337 secs period_search_10112_windows_x86_64__cuda55.exe -verb -nog : Elapsed 156.780 secs, speedup: 44.23% ratio: 1.79x CPU 2.777 secs, speedup: 99.00% ratio: 100.23x period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 216.466 secs, speedup: 23.01% ratio: 1.30x CPU 213.721 secs, speedup: 23.22% ratio: 1.30x I have no idea why the 64bit CUDA app is slower here than the 32bit app in the first run. In both bench runs, the CPU was running with 4.0 GHz (no turbo or throttling) during the cpu tests, the GPU with 980 / 1500 MHz (not overclocked, not in DP mode) during the CUDA tests. I should do some more testing here... - don't mind the lines that say "Result : Strongly similar, Q= ..." - this info was intended for SETI@home and may/will be wrong for Asteroids@home To avoid those, it seems sufficient to delete the line "if exist .\testDatas\ref\ref*!wunbr%%w!.res call .\tools\mb_validate.cmd" from the MBbench210_Asteroids.cmd file (I commented it out instead). In that case, I suppose, you could also delete mb_validate.cmd, rescmpv5.exe and the rescmp subdirectory from the .\Tools directory. The result file from the second run is also included below. ID: 3083 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3084 - Posted: 22 May 2014, 2:07:31 UTC MBbench210_Asteroids.cmd ====================================== 1 testWU(s) found (input_22147_73_short.wu) 1 reference science app(s) found (period_search_10210_windows_intelx86__sse2.exe -verb -nog) 9 science app(s) found (period_search_10112_windows_intelx86__cuda55.exe -verb -nog) (period_search_10112_windows_x86_64__cuda55.exe -verb -nog) (period_search_10210_windows_intelx86.exe -verb -nog) (period_search_10210_windows_intelx86__avx.exe -verb -nog) (period_search_10210_windows_intelx86__sse3.exe -verb -nog) (period_search_10210_windows_x86_64__avx.exe -verb -nog) (period_search_10210_windows_x86_64__sse2.exe -verb -nog) (period_search_10210_windows_x86_64__sse3.exe -verb -nog) (period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog) ====================================== period_search_10210_windows_intelx86__sse2.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86__sse2.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:30:59.385 Ended at : 02:35:40.576 Result : stored as ref for validations. 281.144 secs Elapsed 278.337 secs CPU time [ stderr ] 02:30:59 (6000): Can't open init data file - running in standalone mode 02:35:38 (6000): called boinc_finish [ /stderr ] ------------ period_search_10112_windows_intelx86__cuda55.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10112_windows_intelx86__cuda55.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:35:43.711 Ended at : 02:37:50.050 126.292 secs Elapsed 2.465 secs CPU time Speedup : 99.11% Ratio : 112.92x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:35:43 (4480): Can't open init data file - running in standalone mode CUDA RC12!!!!!!!!!! CUDA Device number: 0 CUDA Device: GeForce GTX TITAN Compute capability: 3.5 Multiprocessors: 14 Grid dim: 224 = 14*16 Block dim: 128 02:37:47 (4480): called boinc_finish [ /stderr ] ------------ period_search_10112_windows_x86_64__cuda55.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10112_windows_x86_64__cuda55.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:37:53.295 Ended at : 02:37:55.619 2.278 secs Elapsed 0.234 secs CPU time Speedup : 99.92% Ratio : 1189.47x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:37:53 (6032): Can't open init data file - running in standalone mode 02:37:53 (6032): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_intelx86.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:37:58.848 Ended at : 02:47:37.573 578.662 secs Elapsed 575.800 secs CPU time Speedup : -106.87% Ratio : 0.48x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:37:58 (1956): Can't open init data file - running in standalone mode 02:47:35 (1956): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_intelx86__avx.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86__avx.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:47:40.818 Ended at : 02:47:42.940 2.044 secs Elapsed 0.031 secs CPU time Speedup : 99.99% Ratio : 8978.61x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:47:40 (4980): Can't open init data file - running in standalone mode 02:47:40 (4980): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_intelx86__sse3.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86__sse3.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:47:46.184 Ended at : 02:47:48.259 2.028 secs Elapsed 0.016 secs CPU time Speedup : 99.99% Ratio : 17396.06x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:47:46 (3572): Can't open init data file - running in standalone mode 02:47:46 (3572): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_x86_64__avx.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_x86_64__avx.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:47:51.504 Ended at : 02:47:53.595 2.028 secs Elapsed 0.016 secs CPU time Speedup : 99.99% Ratio : 17396.06x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:47:51 (5224): Can't open init data file - running in standalone mode 02:47:51 (5224): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_x86_64__sse2.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_x86_64__sse2.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:47:56.855 Ended at : 02:47:58.961 2.028 secs Elapsed 0.016 secs CPU time Speedup : 99.99% Ratio : 17396.06x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:47:56 (5784): Can't open init data file - running in standalone mode 02:47:56 (5784): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_x86_64__sse3.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_x86_64__sse3.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:48:02.206 Ended at : 02:48:04.281 2.028 secs Elapsed 0.016 secs CPU time Speedup : 99.99% Ratio : 17396.06x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:48:02 (6844): Can't open init data file - running in standalone mode 02:48:02 (6844): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_x86_64_bd_fma4_gcc.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 02:48:07.525 Ended at : 02:48:09.647 2.044 secs Elapsed 0.031 secs CPU time Speedup : 99.99% Ratio : 8978.61x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_short.wu.res Result : Strongly similar, Q= 1.010e+004% [ stderr ] 02:48:07 (1964): Can't open init data file - running in standalone mode Using: FMA4 02:48:07 (1964): called boinc_finish [ /stderr ] ------------ Quick timetable WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 281.144 secs CPU 278.337 secs period_search_10112_windows_intelx86__cuda55.exe -verb -nog : Elapsed 126.292 secs, speedup: 55.08% ratio: 2.23x CPU 2.465 secs, speedup: 99.11% ratio: 112.92x period_search_10112_windows_x86_64__cuda55.exe -verb -nog : Elapsed 2.278 secs, speedup: 99.19% ratio: 123.42x CPU 0.234 secs, speedup: 99.92% ratio: 1189.47x period_search_10210_windows_intelx86.exe -verb -nog : Elapsed 578.662 secs, speedup: -105.82% ratio: 0.49x CPU 575.800 secs, speedup: -106.87% ratio: 0.48x period_search_10210_windows_intelx86__avx.exe -verb -nog : Elapsed 2.044 secs, speedup: 99.27% ratio: 137.55x CPU 0.031 secs, speedup: 99.99% ratio: 8978.61x period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64__avx.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64__sse2.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 2.028 secs, speedup: 99.28% ratio: 138.63x CPU 0.016 secs, speedup: 99.99% ratio: 17396.06x period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 2.044 secs, speedup: 99.27% ratio: 137.55x CPU 0.031 secs, speedup: 99.99% ratio: 8978.61x ------------ ID: 3084 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3085 - Posted: 22 May 2014, 2:11:17 UTC MBbench210_Asteroids.cmd ====================================== 1 testWU(s) found (input_22147_73_short.wu) 1 reference science app(s) found (period_search_10210_windows_intelx86__sse2.exe -verb -nog) 2 science app(s) found (period_search_10112_windows_x86_64__cuda55.exe -verb -nog) (period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog) ====================================== period_search_10210_windows_intelx86__sse2.exe -verb -nog / input_22147_73_short.wu : Result cached, skipping execution 281.144 secs Elapsed 278.337 secs CPU time Stderr.txt : not found ------------ period_search_10112_windows_x86_64__cuda55.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10112_windows_x86_64__cuda55.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 03:16:20.165 Ended at : 03:18:57.023 156.780 secs Elapsed 2.777 secs CPU time Speedup : 99.00% Ratio : 100.23x [ stderr ] 03:16:20 (5812): Can't open init data file - running in standalone mode CUDA RC12!!!!!!!!!! CUDA Device number: 0 CUDA Device: GeForce GTX TITAN Compute capability: 3.5 Multiprocessors: 14 Grid dim: 224 = 14*16 Block dim: 128 03:18:54 (5812): called boinc_finish [ /stderr ] ------------ period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_x86_64_bd_fma4_gcc.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 03:19:00.221 Ended at : 03:22:36.734 216.466 secs Elapsed 213.721 secs CPU time Speedup : 23.22% Ratio : 1.30x [ stderr ] 03:19:00 (6564): Can't open init data file - running in standalone mode Using: FMA4 03:22:34 (6564): called boinc_finish [ /stderr ] ------------ Quick timetable WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 281.144 secs CPU 278.337 secs period_search_10112_windows_x86_64__cuda55.exe -verb -nog : Elapsed 156.780 secs, speedup: 44.23% ratio: 1.79x CPU 2.777 secs, speedup: 99.00% ratio: 100.23x period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 216.466 secs, speedup: 23.01% ratio: 1.30x CPU 213.721 secs, speedup: 23.22% ratio: 1.30x ------------ ID: 3085 · Rating: 0 · rate: / Reply Quote

Dutch Send message Joined: 30 Jan 14 Posts: 9 Credit: 2,780,160 RAC: 0	Message 3086 - Posted: 22 May 2014, 2:38:20 UTC My old Xeon X5570's : = MB Knabench 2.10 W32-W64 2012-02-18 by Kna + Simon + Joe = mods: quick timetable, stderr, speedup/ratio, AppTimes = /ref/ by Raistmer = BOINC install detection by Richard Haselgrove 1 testWU(s) found └─(input_22147_73_short.wu) 1 reference science app(s) found └─(period_search_10210_windows_intelx86__sse2.exe -verb -nog) 2 science app(s) found └─(period_search_10210_windows_intelx86__sse3.exe -verb -nog) └─(period_search_10210_windows_x86_64__sse3.exe -verb -nog) ====================================== Suspending BOINC crunching... Previous BOINC state will be restored when test finishes. ------------ Running app : period_search_10210_windows_intelx86__sse2.exe -verb -nog with WU : input_22147_73_short.wu Started at : 03:46:19.955 Ended at : 03:50:49.513 Result : stored as ref for validations. 269.514 secs Elapsed 266.372 secs CPU time ------------ Running app : period_search_10210_windows_intelx86__sse3.exe -verb -nog with WU : input_22147_73_short.wu Started at : 03:50:52.595 Ended at : 03:55:15.707 263.076 secs Elapsed 259.851 secs CPU time Speedup : 2.45% Ratio : 1.03x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_shor t.wu.res Result : Strongly similar, Q= 1.010e+004% ------------ Running app : period_search_10210_windows_x86_64__sse3.exe -verb -nog with WU : input_22147_73_short.wu Started at : 03:55:18.917 Ended at : 03:55:20.971 2.018 secs Elapsed 0.016 secs CPU time Speedup : 99.99% Ratio : 16648.25x R2: .\ref\ref-period_search_10210_windows_intelx86__sse2.exe-input_22147_73_shor t.wu.res Result : Strongly similar, Q= 1.010e+004% ------------ Bench results file XEONX2-20140522-0355-benchMB.txt stored in .\Testdatas\ directory. Quick timetable WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 269.514 secs CPU 266.372 secs period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 263.076 secs, speedup: 2.39% ratio: 1.02x CPU 259.851 secs, speedup: 2.45% ratio: 1.03x period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 2.018 secs, speedup: 99.25% ratio: 133.56x CPU 0.016 secs, speedup: 99.99% ratio: 16648.25x ====================================== Restoring BOINC to pretest state... ========== That's all folks ========== ID: 3086 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3088 - Posted: 22 May 2014, 3:29:01 UTC - in response to Message 3086. Good mornin' period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 2.018 secs, speedup: 99.25% ratio: 133.56x CPU 0.016 secs, speedup: 99.99% ratio: 16648.25x I see, you're having the same problem with the second run (not including the ref). ID: 3088 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3089 - Posted: 22 May 2014, 3:55:12 UTC BilBg, it might be a good idea to include a copy of the default test wu "input_22147_73_short.wu" also in the subdirectory .\TestWUs\Reserve\Short of your bench package. Otherwise it could get lost accidentally if one changes the WUs to bench. ;) ID: 3089 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3090 - Posted: 22 May 2014, 7:46:02 UTC ... btw, same is true for the period_search_10210_windows_intelx86__sse3.exe app, which should also be included in .\Science_apps\Reserve IMHO. ID: 3090 · Rating: 0 · rate: / Reply Quote

BilBg Send message Joined: 19 Jun 12 Posts: 221 Credit: 623,640 RAC: 0	Message 3094 - Posted: 22 May 2014, 12:46:54 UTC Uh oh, sorry - I forget to change one of the filenames in the .cmd file (state.sah --> period_search_state) This may prevent running several tests in a row The previous package have BUG, get the new: http://www68.zippyshare.com/v/79720833/file.html You can have only one app in Reference\ You can have many apps in Science_apps\ If you accidentally delete some file or any other bad change - just uncompress the .7z package again Now I'll wait for the next bug report ;) - ALF - "Find out what you don't do well ..... then don't do it!" :) ID: 3094 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3098 - Posted: 23 May 2014, 7:49:57 UTC - in response to Message 3094. Great, works now as intended. Thanks a lot! If you accidentally delete some file or any other bad change - just uncompress the .7z package again Yes, sure. But both \Reference subdirectories "are intended to store what you plan to keep within reach" (quote from the readme :) ), so it makes sense to put all apps and test WUs in it IMHO. This time I ran it concurrently with 3 Asteroids tasks within BOINC by using the 'mode noBS' switch in the provided BenchCfg.txt (and an app_config.xml for BOINC). Works great, too. MBbench210_Asteroids.cmd ====================================== 1 testWU(s) found (input_22147_73_short.wu) 1 reference science app(s) found (period_search_10210_windows_intelx86__sse2.exe -verb -nog) 10 science app(s) found (period_search_10112_windows_intelx86__cuda55.exe -verb -nog) (period_search_10112_windows_x86_64__cuda55.exe -verb -nog) (period_search_10210_windows_intelx86.exe -verb -nog) (period_search_10210_windows_intelx86__avx.exe -verb -nog) (period_search_10210_windows_intelx86__sse2.exe -verb -nog) (period_search_10210_windows_intelx86__sse3.exe -verb -nog) (period_search_10210_windows_x86_64__avx.exe -verb -nog) (period_search_10210_windows_x86_64__sse2.exe -verb -nog) (period_search_10210_windows_x86_64__sse3.exe -verb -nog) (period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog) ====================================== period_search_10210_windows_intelx86__sse2.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10210_windows_intelx86__sse2.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 08:45:07.825 Ended at : 08:50:06.829 Result : stored as ref for validations. 298.973 secs Elapsed 296.542 secs CPU time [ stderr ] 08:45:07 (4928): Can't open init data file - running in standalone mode 08:50:04 (4928): called boinc_finish [ /stderr ] ------------ period_search_10112_windows_intelx86__cuda55.exe -verb -nog / input_22147_73_short.wu : AppName: period_search_10112_windows_intelx86__cuda55.exe AppArgs: -verb -nog TaskName: input_22147_73_short.wu Started at : 08:50:09.949 Ended at : 08:52:14.313 124.317 secs Elapsed 0.998 secs CPU time Speedup : 99.66% Ratio : 297.14x [ stderr ] 08:50:09 (3572): Can't open init data file - running in standalone mode CUDA RC12!!!!!!!!!! CUDA Device number: 0 CUDA Device: GeForce GTX TITAN Compute capability: 3.5 Multiprocessors: 14 Grid dim: 224 = 14*16 Block dim: 128 08:52:12 (3572): called boinc_finish [ /stderr ] ------------ (...) ------------ Quick timetable WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 298.973 secs CPU 296.542 secs period_search_10112_windows_intelx86__cuda55.exe -verb -nog : Elapsed 124.317 secs, speedup: 58.42% ratio: 2.40x CPU 0.998 secs, speedup: 99.66% ratio: 297.14x period_search_10112_windows_x86_64__cuda55.exe -verb -nog : Elapsed 156.671 secs, speedup: 47.60% ratio: 1.91x CPU 1.435 secs, speedup: 99.52% ratio: 206.65x period_search_10210_windows_intelx86.exe -verb -nog : Elapsed 610.497 secs, speedup: -104.20% ratio: 0.49x CPU 607.717 secs, speedup: -104.93% ratio: 0.49x period_search_10210_windows_intelx86__avx.exe -verb -nog : Elapsed 610.913 secs, speedup: -104.34% ratio: 0.49x CPU 608.092 secs, speedup: -105.06% ratio: 0.49x period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 308.726 secs, speedup: -3.26% ratio: 0.97x CPU 306.277 secs, speedup: -3.28% ratio: 0.97x period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 284.498 secs, speedup: 4.84% ratio: 1.05x CPU 281.987 secs, speedup: 4.91% ratio: 1.05x period_search_10210_windows_x86_64__avx.exe -verb -nog : Elapsed 356.507 secs, speedup: -19.24% ratio: 0.84x CPU 353.951 secs, speedup: -19.36% ratio: 0.84x period_search_10210_windows_x86_64__sse2.exe -verb -nog : Elapsed 262.440 secs, speedup: 12.22% ratio: 1.14x CPU 259.960 secs, speedup: 12.34% ratio: 1.14x period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 266.152 secs, speedup: 10.98% ratio: 1.12x CPU 263.735 secs, speedup: 11.06% ratio: 1.12x period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 238.649 secs, speedup: 20.18% ratio: 1.25x CPU 236.014 secs, speedup: 20.41% ratio: 1.26x ------------ ID: 3098 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3099 - Posted: 23 May 2014, 10:42:36 UTC Some more figures from my two intels. Both are notebooks and they will throttle under full load, so below figures will not be achieved then. i7-3720QM (Ivy Bridge), avg. clock during tests 3.3 GHz WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 189.790 secs CPU 187.357 secs period_search_10210_windows_intelx86.exe -verb -nog : Elapsed 386.662 secs, speedup: -103.73% ratio: 0.49x CPU 384.433 secs, speedup: -105.19% ratio: 0.49x period_search_10210_windows_intelx86__avx.exe -verb -nog : Elapsed 200.648 secs, speedup: -5.72% ratio: 0.95x CPU 198.418 secs, speedup: -5.90% ratio: 0.94x period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 189.634 secs, speedup: 0.08% ratio: 1.00x CPU 187.373 secs, speedup: -0.01% ratio: 1.00x period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 181.210 secs, speedup: 4.52% ratio: 1.05x CPU 179.011 secs, speedup: 4.45% ratio: 1.05x period_search_10210_windows_x86_64__avx.exe -verb -nog : Elapsed 182.052 secs, speedup: 4.08% ratio: 1.04x CPU 179.932 secs, speedup: 3.96% ratio: 1.04x period_search_10210_windows_x86_64__sse2.exe -verb -nog : Elapsed 168.028 secs, speedup: 11.47% ratio: 1.13x CPU 165.735 secs, speedup: 11.54% ratio: 1.13x period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 172.244 secs, speedup: 9.24% ratio: 1.10x CPU 170.103 secs, speedup: 9.21% ratio: 1.10x i7-3632QM (Ivy Bridge), avg. clock during CPU tests 2.9 GHz, nVIDIA GeForce GT 650M WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 216.072 secs CPU 212.349 secs period_search_10112_windows_intelx86__cuda55.exe -verb -nog : Elapsed 410.356 secs, speedup: -89.92% ratio: 0.53x CPU 2.886 secs, speedup: 98.64% ratio: 73.58x period_search_10112_windows_x86_64__cuda55.exe -verb -nog : Elapsed 477.314 secs, speedup: -120.91% ratio: 0.45x CPU 2.808 secs, speedup: 98.68% ratio: 75.62x period_search_10210_windows_intelx86.exe -verb -nog : Elapsed 433.743 secs, speedup: -100.74% ratio: 0.50x CPU 429.892 secs, speedup: -102.45% ratio: 0.49x period_search_10210_windows_intelx86__avx.exe -verb -nog : Elapsed 227.043 secs, speedup: -5.08% ratio: 0.95x CPU 223.066 secs, speedup: -5.05% ratio: 0.95x period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 213.876 secs, speedup: 1.02% ratio: 1.01x CPU 210.102 secs, speedup: 1.06% ratio: 1.01x period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 201.677 secs, speedup: 6.66% ratio: 1.07x CPU 198.542 secs, speedup: 6.50% ratio: 1.07x period_search_10210_windows_x86_64__avx.exe -verb -nog : Elapsed 205.078 secs, speedup: 5.09% ratio: 1.05x CPU 201.475 secs, speedup: 5.12% ratio: 1.05x period_search_10210_windows_x86_64__sse2.exe -verb -nog : Elapsed 197.574 secs, speedup: 8.56% ratio: 1.09x CPU 190.368 secs, speedup: 10.35% ratio: 1.12x period_search_10210_windows_x86_64__sse3.exe -verb -nog : Elapsed 195.328 secs, speedup: 9.60% ratio: 1.11x CPU 191.273 secs, speedup: 9.93% ratio: 1.11x Surprisingly the 64bit SSE2 app is fastest in the first run while the 64bit SSE3 app comes out on top in the second but only by a small margin. The AVX apps, especially the 32bit flavor, are disappointing (are those maybe a bit overoptimized for Haswell ;) ). And again, btw, is the 64bit CUDA app slower than the 32bit one, strange. ID: 3099 · Rating: 0 · rate: / Reply Quote

HA-SOFT, s.r.o. Project developer Project tester Send message Joined: 21 Dec 12 Posts: 176 Credit: 136,462,135 RAC: 1	Message 3100 - Posted: 23 May 2014, 14:05:11 UTC - in response to Message 3099. AVX is optimized for Haswell. On Ivy and Sandy Bridge SSEx apps are better. Instruction set is not the only difference between apps. AVX has slightly different code than SSE app. 32bit apps are a little out in SSE and AVX world... ID: 3100 · Rating: 0 · rate: / Reply Quote

alexander Send message Joined: 28 Apr 13 Posts: 87 Credit: 26,736,520 RAC: 1,304	Message 3101 - Posted: 23 May 2014, 18:21:05 UTC Bench results file RAJ-20140523-1931-benchMB.txt stored in .\Testdatas\ directory. Quick timetable WU : input_22147_73_short.wu period_search_10210_windows_intelx86__sse2.exe -verb -nog : Elapsed 286.073 secs CPU 283.033 secs period_search_10210_windows_intelx86__sse3.exe -verb -nog : Elapsed 278.273 secs, speedup: 2.73% ratio: 1.03x CPU 274.718 secs, speedup: 2.94% ratio: 1.03x period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 220.928 secs, speedup: 22.77% ratio: 1.29x CPU 217.356 secs, speedup: 23.20% ratio: 1.30x Thanks to all that were engaged in creating these apps! CPU: AMD-A10-7700K, std clock 3.4GHz, DDR3-1866MHz, hostid=88984 Alexander ID: 3101 · Rating: 0 · rate: / Reply Quote

Kyong Project administrator Project developer Project tester Project scientist Send message Joined: 9 Jun 12 Posts: 584 Credit: 52,667,664 RAC: 0	Message 3102 - Posted: 24 May 2014, 8:50:28 UTC - in response to Message 3080. - I'm not sure if I used proper method to make short WUs - project admins have to check/say -- my method: using hex editor (HxD) changed the second number on the first line from 0.5 to 5.0 Hi, I am sorry for replying so late. You are almost right. You have to change first and third number. The first is where the app should start and the third is where the app should end. So this is the way how to make a short wu. You just change the range of searching with changing first and third number. The second number is how high step the app should do but if it is more than 0.5 than the accuracy of finding period is decreased. ID: 3102 · Rating: 0 · rate: / Reply Quote

BilBg Send message Joined: 19 Jun 12 Posts: 221 Credit: 623,640 RAC: 0	Message 3103 - Posted: 24 May 2014, 13:47:04 UTC - in response to Message 3098. This time I ran it concurrently with 3 Asteroids tasks within BOINC by using the 'mode noBS' switch in the provided BenchCfg.txt (and an app_config.xml for BOINC). Works great, too. Yes, that's the way to do more real-life-like test. Usually the test runs on a free from other load machine. On machine under high load the relative performance of the apps may change because of: - CPU cache misses/pollution/dirty - Hyper-Threading in Intel CPUs - "module" in the new AMD CPUs (different apps may be influenced by different ways/degree) In this case it is better to divide the test in two runs: 1) for CPU apps only - free one (more) CPU core in BOINC (and keep GPU usage as is) 2) for GPU apps only - run one less GPU task (if usually only one GPU task runs - just suspend GPU from Activity menu)(and keep CPU tasks in BOINC at the usual number) - ALF - "Find out what you don't do well ..... then don't do it!" :) ID: 3103 · Rating: 0 · rate: / Reply Quote

BilBg Send message Joined: 19 Jun 12 Posts: 221 Credit: 623,640 RAC: 0	Message 3104 - Posted: 24 May 2014, 14:18:09 UTC - in response to Message 3099. And again, btw, is the 64bit CUDA app slower than the 32bit one, strange. Not strange at all ;) The CUDA developer for SETI@home (jason_gee) tested this many times (and do not produce/offer 64bit CUDA apps at all) "As 64 bit addresses use twice as much register space as 32 bit ones, and GPUs use thousands of threads (each with many registers), 64 bit GPU code tends to be slower as well, chewing up precious gpu registers quickly. That can shift with driver latency being dominant at the moment (at least on Windows), and improved compiler technologies..." "For GPUs, 64 bit does not mean faster or better in any way, only can access more memory" http://setiathome.berkeley.edu/forum_thread.php?id=74083&postid=1483194#1483194 - ALF - "Find out what you don't do well ..... then don't do it!" :) ID: 3104 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3129 - Posted: 31 May 2014, 17:03:10 UTC - in response to Message 3101. WU : input_22147_73_short.wu (...) period_search_10210_windows_x86_64_bd_fma4_gcc.exe -verb -nog : Elapsed 220.928 secs, speedup: 22.77% ratio: 1.29x CPU 217.356 secs, speedup: 23.20% ratio: 1.30x (...) CPU: AMD-A10-7700K, std clock 3.4GHz, DDR3-1866MHz, hostid=88984 I'm surprised how fast this Steamroller-Kaveri is compared to my 4.0 GHz Piledriver-Vishera. 221 vs. 216 secs elapsed time with FMA4. Are you sure it ran at 3.4 GHz and not with turbo frequency 3.8 GHz for example? Otherwise AMD must have done a hell of a job with Steamroller improvements here. ;) ID: 3129 · Rating: 0 · rate: / Reply Quote

(retired account) Send message Joined: 3 Jan 13 Posts: 30 Credit: 1,705,200 RAC: 0	Message 3130 - Posted: 31 May 2014, 17:06:21 UTC - in response to Message 3104. "For GPUs, 64 bit does not mean faster or better in any way, only can access more memory" Then it makes no sense to provide a true 64bit app, if you don't need the larger memory space, right? The project could still provide a renamed 32bit app for the 64bit Windows flavors. ID: 3130 · Rating: 0 · rate: / Reply Quote

BilBg Send message Joined: 19 Jun 12 Posts: 221 Credit: 623,640 RAC: 0	Message 3305 - Posted: 6 Jul 2014, 8:49:03 UTC Last modified: 6 Jul 2014, 9:13:44 UTC Time to introduce another package. This one will be as 'Do it yourself (DIY)' To test for the optimal number of GPU tasks get this package: SetiPerformance 1.8 + Added apps from Lunatics v0.41 http://setiathome.berkeley.edu/forum_thread.php?id=73524 (I collected it originally for SETI@home (the SetiPerformance program itself is not mine)) You may add GPU apps/tasks from any project this way: - Make a new folder (in 'zip' folder) (do not use spaces and special symbols) - Put a GPU app in it (.exe .dll ...) - Put one task/WU file in it - Rename the task/WU file to the standard general name used by the app For Asteroids@home Rename it to period_search_in For another project look in 'slots' (BOINC Data dir) to see the name (copy only the name, not the dummy file - it is not real WU) (E.g. for SETI@home v7 use work_unit.sah , for AstroPulse - in.dat , ...) - Edit config.xml (with Notepad) - Copy one <data> section and Edit the copy to set: <name> (any of your choice (do not use spaces and special symbols)) <folder> and <exe> to what you did in above first 2 steps Example: <data> <name>MB7_win_x86_SSE_OpenCL_ATi_r1843</name> <folder>OpenCL</folder> <exe>MB7_win_x86_SSE_OpenCL_ATi_r1843.exe</exe> <parameter>%20--device%20%device%20</parameter> </data> - Stop BOINC (you may if you wish stop only GPU computing) - you are ready to run SetiPerformance To use many WUs: - go to 'workunits' folder - Move all old files (from other projects) to workunits_OFF - Put new task/WU files in 'workunits' folder - Rename them with .wu extension - in SetiPerformance check 'Use all XX workunits' ) - ALF - "Find out what you don't do well ..... then don't do it!" :) ID: 3305 · Rating: 0 · rate: / Reply Quote

BilBg Send message Joined: 19 Jun 12 Posts: 221 Credit: 623,640 RAC: 0	Message 3308 - Posted: 7 Jul 2014, 11:17:29 UTC - in response to Message 3305. Last modified: 7 Jul 2014, 11:34:47 UTC Sorry - I gave some wrong info: 1) In the new folder/directory you need to put also a (dummy) file named work_unit.sah (just because SetiPerformance checks if such file exist) 2) It's not possible to automatically 'use many WUs' (because SetiPerformance renames them to work_unit.sah) (It may be possible if: - you hack the app .exe file (e.g. change string in it from period_search_in to work_unit.sah) - or (unlikely) if the application have command line switch to specify the input file (e.g. --infile) Then you may change the <parameter> to: <parameter>%20--device%20%device%20--infile%20work_unit.sah%20</parameter> ) Note: The progress indicator of SetiPerformance will not work since it is designed to analyze only SETI@home state files. It will be hard to generalize this indicator/analyze since state files are very different even at one project (app - file - tag): SETI@home v7 - state.sah - <prog>0.69267820</prog> AstroPulse - pulse.out* - <fraction_done>0.225225</fraction_done> - ALF - "Find out what you don't do well ..... then don't do it!" :) ID: 3308 · Rating: 0 · rate: / Reply Quote

1 · 2 · Next

Message boards : Number crunching : Benchmark package for Asteroids@home