Message boards :
Number crunching :
Public beta for nVidia AstroPulse, rev 521
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 30 · Next
Author | Message |
---|---|
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Hmmmm, wonder why this one decided to go inconclusive? http://setiathome.berkeley.edu/workunit.php?wuid=781560387 Thats probably because best pulses may varies. Should validate with third host. With each crime and every kindness we birth our future. |
Frizz Send message Joined: 17 May 99 Posts: 271 Credit: 5,852,934 RAC: 0 |
Not looking too good here in my side. I get lots of overflows: http://setiathome.berkeley.edu/results.php?hostid=6108600&offset=0&show_names=0&state=2&appid=5 |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Not as bad as you might think Frizz. Your wingmates are also showing as 30/30 overflows on those. Why they are going out to another is another question though. PROUD MEMBER OF Team Starfire World BOINC |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Not looking too good here in my side. As said in first post then you need to reduce unroll factor. |
perryjay Send message Joined: 20 Aug 02 Posts: 3377 Credit: 20,676,751 RAC: 0 |
Raistmer, he's down to an unroll of 4 and he has five different wing men also giving 30/30 overflows. It's just a bad string of work units. PROUD MEMBER OF Team Starfire World BOINC |
halfempty Send message Joined: 2 Jun 99 Posts: 97 Credit: 35,236,901 RAC: 114 |
Here's some data on my first batch of units, all run in low priority. GTS 450 with 266.58 drivers running one task at a time, AMD Phenom II running CPU Seti on all 4 cores. Data is in the format of Task, Run Time, CPU Time, Blanking. -unroll 12 -ffa_block 8192 -ffa_block_fetch 2048 2000647746 5,959.56 1,272.56 2.39 2000647748 5,735.61 1,059.23 0 -unroll 12 -ffa_block 6144 -ffa_block_fetch 1536 2000767200 5,185.02 824.29 0 2000767203 6,008.74 1,790.99 11.33 2001090114 5,217.46 1,024.21 3.03 2001090124 4,934.34 751 0 2001090144 4,940.01 750.88 0 -unroll 8 -ffa_block 6144 -ffa_block_fetch 1536 2001192123 5,261.58 1,044.43 2.88 2001242946 5,230.54 1,005.05 2.39 2001556444 12,680.55 8,374.49 82.61 2003220580 5,016.48 789.69 0 2003220651 5,220.56 1,000.50 2.34 2003220653 5,358.10 1,007.10 2.39 2003220657 13,440.94 9,143.48 90.82 2004967204 5,011.65 792.3 0 Block/Fetch of 8192/2048 produced too much screen lag. 6144/1536 only produces a minor occasional lag I can live with. Hope this is useful to somebody. |
Slavac Send message Joined: 27 Apr 11 Posts: 1932 Credit: 17,952,639 RAC: 0 |
http://setiathome.berkeley.edu/result.php?resultid=2001666989 http://setiathome.berkeley.edu/result.php?resultid=2001666977 560ti. Both are showing up as CPU which is strange, but they both validated. Using stock settings from Claggy of: <app> <name>astropulse_v505</name> </app> <file_info> <name>ap_5.06_win_x86_SSE3_OpenCL_NV_r521.exe</name> <executable/> </file_info> <file_info> <name>AstroPulse_Kernels_r521.cl</name> <executable/> </file_info> <app_version> <app_name>astropulse_v505</app_name> <version_num>505</version_num> <platform>windows_intelx86</platform> <avg_ncpus>0.04</avg_ncpus> <max_ncpus>0.20</max_ncpus> <plan_class>cuda</plan_class> <cmdline>-ffa_block 6144 -ffa_block_fetch 1536 -hp -unroll 6 -instances_per_device 2</cmdline> <file_ref> <file_name>ap_5.06_win_x86_SSE3_OpenCL_NV_r521.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>AstroPulse_Kernels_r521.cl</file_name> <copy_file/> </file_ref> <coproc> <type>CUDA</type> <count>.5</count> </coproc> </app_version> Any thoughts on how I should tailor those settings? Executive Director GPU Users Group Inc. - brad@gpuug.org |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
http://setiathome.berkeley.edu/result.php?resultid=2001666989 Its not strange at all. Version_num has to be 506. Your 560s have 8 compute units so try unroll -8, -10, -12 I use <avg_ncpus>0.05</avg_ncpus> <max_ncpus>0.05</max_ncpus> Try <cmdline>-ffa_block 8192 -ffa_block_fetch 2048 -unroll 12 -instances_per_device 2</cmdline> With each crime and every kindness we birth our future. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
http://setiathome.berkeley.edu/result.php?resultid=2001666989 Version number is irrelevant - it is ignored when using anonymous_platform. The processor shown on the website is the one to which the task was originally issued. If that task was issued in response to a CPU work request, the database will continue to show it as a CPU task, even if you subsequently reschedule it to run on a GPU. If you have other tasks which have been rescheduled from CPU to GPU, do NOT change the version number in app_info - the mis-match will cause all cached tasks to be abandoned. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
My bad sorry. With each crime and every kindness we birth our future. |
JohnDK Send message Joined: 28 May 00 Posts: 1222 Credit: 451,243,443 RAC: 1,127 |
Your 560s have 8 compute units so try unroll -8, -10, -12 My 460 have 7 compute units, what numbers should I use? |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Your 560s have 8 compute units so try unroll -8, -10, -12 -unroll 12 should be fine. With each crime and every kindness we birth our future. |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
http://setiathome.berkeley.edu/result.php?resultid=2001666989 Those command parameter settings aren't the Stock settings, just settings for Hopefully Maximum Compatibility, and Minimum Lag, there is a wider range of Nvidia OpenCL capible GPU's than ATI OpenCL GPU's, from 8400 GS up to GTX590, the Stock Settings would be -ffa_block 8192 -ffa_block_fetch 2048 -unroll 10 (same as ATI variant) Claggy |
CryptokiD Send message Joined: 2 Dec 00 Posts: 150 Credit: 3,216,632 RAC: 0 |
i gave it a shot on the one rig of mine which actually works right now. when i crunch mb with cuda the cpu stays around 8-15% usage with this old athlon64 3600+ (2.25 real ghz) when i crunch ap with cuda using the 5.06 my cpu stays pegged at 100% i dont like that. but i cant complain since my cpu is so slow i dont even use it for crunching. only to keep mr cuda fed. |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
http://setiathome.berkeley.edu/result.php?resultid=2001666989 This is looking like the ol' proverbial can o' worms to the kittyman. Question.... Are these settings mostly GPU specific, or does the system CPU capability play into it too? Perhaps somebody could start to compile a chart with recommended starting points for these settings based on the GPU. There is going to be an onslaught of questions regarding this once the app leaves beta and is added to the Lunatics installer. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
It is GPU related Mark but you can always start with basic settings. Each modell has different amount of compute units. Basic settings -ffa_block 6144 -ffa_block_fetch 1536 -unroll 6 should work on each discrete GPU. Rest is optimizing. With each crime and every kindness we birth our future. |
JohnDK Send message Joined: 28 May 00 Posts: 1222 Credit: 451,243,443 RAC: 1,127 |
i gave it a shot on the one rig of mine which actually works right now. The 100% CPU is properly due to the drivers you have installed, this is what Raistmer wrote in his post: 2 hosts already reported greatly increased CPU consumption when running with 27x.xx drivers. Rolling back to 26x.xx ones solve this issue in both cases. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
i gave it a shot on the one rig of mine which actually works right now. It could be also that those units are heavily blanked. Blankings are processed by CPU. Just finnish some tasks that we can look at it. With each crime and every kindness we birth our future. |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
It is GPU related Mark but you can always start with basic settings. ffa around the block and do the dosey-doe, fetch ur padna, do a spin and try then to unroll.....LOL. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
CryptokiD Send message Joined: 2 Dec 00 Posts: 150 Credit: 3,216,632 RAC: 0 |
it takes about 40 hours to crunch an ap with this athlon64 cpu. it takes less then 1.5 hours with the 550ti. thats why i dont even use the cpu for cr4unching anymore. not worth the extra heat and power when cuda literally is 30 times faster. i might try rolling back to earlier driver, but things are going so good on this rig, i might just leave it alone and wait for the final build of ap_nv app to come out. why are any of us even crunching astropulse when the ap portion of seti is apparently broken?. i remember reading about that somwhere. |
©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.