Message boards :
Number crunching :
7850/7870
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 Nov 10 Posts: 10 Credit: 212,710,651 RAC: 0 |
Can the 7800 series run MW@H? otherwise should i stick with 5850 and 6950? |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
Can the 7800 series run MW@H? Yes they can, but they only have 1/16th double precision as opposed to 1/4 for the 79xx. My 7750 takes 10 minutes per, so I would guess around 5 minutes per for the 7850. My 4830 takes 3:45 and my 5830 took 2:15. |
Send message Joined: 13 Nov 10 Posts: 10 Credit: 212,710,651 RAC: 0 |
What about some of the 4 and 5 series nvidia cards? how do they compare time wise? |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
I think my 560 was around 8 minutes and the 460 was at 10. |
Send message Joined: 19 Jul 10 Posts: 601 Credit: 19,056,911 RAC: 6,600 |
My 7750 takes 10 minutes per, so I would guess around 5 minutes per for the 7850. And I thought my old HD3850 would be slow... but apparently with less than 7 minutes per WU it's faster than a current 100€-GPU... at least for Milkyway crunching. |
Send message Joined: 23 Nov 09 Posts: 29 Credit: 17,119,258 RAC: 0 |
My HD 7950 takes 41 seconds per task without an app_info.xml. Can't find a sample app_info.xml though, would love to stop the droops in utilization between completed and starting tasks. |
Send message Joined: 22 Mar 09 Posts: 99 Credit: 503,422,495 RAC: 0 |
You can use this app_info.xml to run two tasks parallel. Kick them out of phase and you will have lesser utilization-drops. <app_info> <app> <name>milkyway</name> </app> <file_info> <name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</name> <executable/> </file_info> <app_version> <app_name>milkyway</app_name> <version_num>102</version_num> <flops>1.0e13</flops> <avg_ncpus>0.05</avg_ncpus> <max_ncpus>1</max_ncpus> <plan_class>opencl_amd</plan_class> <coproc> <type>ATI</type> <count>0.5</count> </coproc> <cmdline></cmdline> <file_ref> <file_name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</file_name> <main_program/> </file_ref> </app_version> </app_info> Cheers Nowi |
Send message Joined: 9 May 12 Posts: 12 Credit: 10,339,447 RAC: 0 |
Thank you [boinc.at] Nowi, I have been looking for an app_info formated correctly for Milkyway. You are right about putting them out of phase. I paused the second one till about 40% thru the first one and now they don't both try to use the CPU at the same time. Sweet ! We had been using Boinc manager to reserve a CPU core but for some reason using the app_info and putting the CPU setting in Boinc back to 100% seems to better the run time. Was 43 sec. running 1 , and now it is 76 sec. running 2. 76/2 = 38 sec. each. Should increase our RAC a good bit. Now if we could only afford a second 7970 we could really pump up the RAC... ;-) |
Send message Joined: 26 Feb 11 Posts: 170 Credit: 205,557,553 RAC: 0 |
My 7750 takes 10 minutes per, so I would guess around 5 minutes per for the 7850. Lol i was thinking the same right now ^^ DSKAG Austria Research Team: http://www.research.dskag.at |
Send message Joined: 8 Apr 09 Posts: 70 Credit: 11,027,167,827 RAC: 0 |
You can compare the theoretical douple precision performance here in Wikipedia: http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units GPU SP DP 3870 496 99.2 4770 960 192 7750 819 51.2 7870 2560 180 a 4770 gets about 70k a day. i won't expect much more from a 7870. I guess 2.5 to 3 min for a WU on the 7870. Tomast, is you 7970 a standard card 925 core and 1375 mem or a overclocked card? It could also be a "2 cards in one box" problem. With 2 cards i need 45 to 47 sec for a WU. Thank you very much [boinc.at] Nowi for the app_info.xml file :) PS: I had the oportunity to test MW on a Tesla (GTX 480 version of Tesla) a WU took 4 min there. That was roughly a year ago. Does anyone knows how long a WU on a normal GTX 480 runs? The Tesla cards should have 1/2 DP of the SP power. |
Send message Joined: 9 May 12 Posts: 12 Credit: 10,339,447 RAC: 0 |
Sebastian* , It is 950 MHz GPU clock and 1425 MHz Mem clock. (Factory Overclocked) |
Send message Joined: 17 Nov 10 Posts: 12 Credit: 482,269,492 RAC: 4,154 |
I am running 2 ati 6950 cards+they take 1:08 each to run.I would like to try running 2 wu's per gpu.I tried to use the app_info file from the above post,but I keep getting that boinc cannot find any compatible files+it aborts all the other wu's.I am using boinc 7.0.25 with windows xp 32 bit.What am I doing wrong?Thanks,Kurt |
Send message Joined: 18 Jul 09 Posts: 300 Credit: 303,562,776 RAC: 0 |
I am running 2 ati 6950 cards+they take 1:08 each to run.I would like to try running 2 wu's per gpu.I tried to use the app_info file from the above post,but I keep getting that boinc cannot find any compatible files+it aborts all the other wu's.I am using boinc 7.0.25 with windows xp 32 bit.What am I doing wrong?Thanks,Kurt This one works for me: <app_info> <app> <name>milkyway</name> </app> <file_info> <name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</name> <executable/> </file_info> <app_version> <app_name>milkyway</app_name> <version_num>102</version_num> <flops>1.0e11</flops> <avg_ncpus>0.05</avg_ncpus> <max_ncpus>1</max_ncpus> <plan_class>ati14ati</plan_class> <coproc> <type>ATI</type> <count>0.5</count> </coproc> <cmdline>--gpu-target-frequency 10 --gpu-disable-checkpointing</cmdline> <file_ref> <file_name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</file_name> <main_program/> </file_ref> </app_version> </app_info> [edit] Oops. Didn't notice you're using 32 bit [/edit] |
Send message Joined: 14 Jun 12 Posts: 13 Credit: 2,295,346 RAC: 0 |
I can confirm that running a 7850 at 900MHz completes the workunit in 5 minutes. |
Send message Joined: 29 May 10 Posts: 19 Credit: 5,917,319 RAC: 0 |
If you want to have only MT WU's use your CPU and have the separations only crunch on your GPU's since it's way faster, the following is the app_info that I made and that I use. I can now crunch 3 WU's of the multi-threaded work units(4 cores per unit) and 4 of the GPU work units (2 units per card). I'm on 2x 7970's and I can pump out 4 GPU work units in total every 77 seconds. As per http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2841 <app_info> <app> <name>milkyway</name> </app> <file_info> <name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</name> <executable/> </file_info> <app_version> <app_name>milkyway</app_name> <version_num>102</version_num> <flops>1.0e11</flops> <avg_ncpus>0.05</avg_ncpus> <max_ncpus>1</max_ncpus> <plan_class>ati14ati</plan_class> <coproc> <type>ATI</type> <count>0.5</count> </coproc> <cmdline>--gpu-target-frequency 10 --gpu-disable-checkpointing</cmdline> <file_ref> <file_name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</file_name> <main_program/> </file_ref> </app_version> <app> <name>milkyway_nbody</name> <user_friendly_name>MilkyWay@Home nbody Simulation</user_friendly_name> </app> <file_info> <name>milkyway_nbody_0.84_windows_x86_64__mt.exe</name> <executable/> </file_info> <file_info> <name>libgomp_64-1.dll</name> <executable/> </file_info> <file_info> <name>pthreadGC2_64.dll</name> <executable/> </file_info> <app_version> <app_name>milkyway_nbody</app_name> <version_num>84</version_num> <plan_class>mt</plan_class> <avg_ncpus>4</avg_ncpus> <max_ncpus>4</max_ncpus> <cmdline>--nthreads=4</cmdline> <file_ref> <file_name>milkyway_nbody_0.84_windows_x86_64__mt.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>libgomp_64-1.dll</file_name> </file_ref> <file_ref> <file_name>pthreadGC2_64.dll</file_name> </file_ref> </app_version> </app_info> |
Send message Joined: 8 Sep 09 Posts: 1 Credit: 10,261,180 RAC: 0 |
HI! Have recently installed a HD 7850 in my system. Just wondering if it's natural for MW@Home 1.02 WU completion time to vary much, as fastest so far has been 5m04s and slowest 8m15s? Previous card (HD 5850 OC (765/4500)) completed every WU in 1m23s which was also the estimated time to completion. With the HD 7850 the estimated completion time is consistently 3m01s. Thanks in advance. |
©2024 Astroinformatics Group