SETI applications for NVIDIA GPU improvement

Author	Message
Zalster Volunteer tester Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242	Message 1814870 - Posted: 3 Sep 2016, 20:46:39 UTC - in response to Message 1814841. Not seen a lot of improvement with both in the commandline. Few seconds improvement Normal AR http://setiathome.berkeley.edu/result.php?resultid=5134899382 WU true angle range is : 0.445407 Run time 4 min 1 sec CPU time 3 min 55 sec http://setiathome.berkeley.edu/result.php?resultid=5135025942 WU true angle range is : 0.432571 Run time 4 min 6 sec CPU time 3 min 39 sec VLARs http://setiathome.berkeley.edu/result.php?resultid=5134892915 WU true angle range is : 0.010503 Run time 7 min 7 sec CPU time 7 min 4 sec http://setiathome.berkeley.edu/result.php?resultid=5134893156 WU true angle range is : 0.011400 Run time 7 min 20 sec CPU time 7 min 18 sec ID: 1814870 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1814876 - Posted: 3 Sep 2016, 21:03:42 UTC - in response to Message 1814871. any noticeable improvement could be only on VLAR cause it's PulseFind area and VLAR is where PulseFind dominates. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1814876 · Reply Quote

robertmiles Volunteer tester Send message Joined: 16 Jan 12 Posts: 213 Credit: 4,117,756 RAC: 6	Message 1814892 - Posted: 3 Sep 2016, 22:39:15 UTC Raistmer, suggestions for future improvements: Check if Nvidia OpenCL offers a way for an application to check the temperature of its GPU. If so, occasionally check it, and print the maximum value somewhere in stderr. This should make overheating as a cause of problems more obvious. Also, when the Task postponed: Suspicious spike results notice is issued, mention it in stderr to help identify which task it was from. ID: 1814892 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1815405 - Posted: 6 Sep 2016, 20:47:36 UTC We have 8.18 deployed on beta, test please. Hopefully they are new RC for main. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1815405 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1815406 - Posted: 6 Sep 2016, 20:49:12 UTC - in response to Message 1814892. Raistmer, suggestions for future improvements: Check if Nvidia OpenCL offers a way for an application to check the temperature of its GPU. If so, occasionally check it, and print the maximum value somewhere in stderr. This should make overheating as a cause of problems more obvious. Also, when the Task postponed: Suspicious spike results notice is issued, mention it in stderr to help identify which task it was from. There is no temperature reading interface in OpenCL. Also, no need to link to particular task in general, What is needed is the knowledge that smth wrong with computational device. One could look into tasks list and see what task is postponed. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1815406 · Reply Quote

J. Mileski Volunteer tester Send message Joined: 9 Jun 02 Posts: 632 Credit: 172,116,532 RAC: 572	Message 1815462 - Posted: 6 Sep 2016, 23:11:37 UTC - in response to Message 1815405. Last modified: 6 Sep 2016, 23:14:48 UTC We have 8.18 deployed on beta, test please. Hopefully they are new RC for main. I have switched 2 of my computers to beta, the first is a xeon E5-2670 8 core 16 thread with 2 GTX 750ti's the other is an AMD 1090T 6 core with 3 GTX 550ti's I've got 2 gt 240, here on an E8500 core2duo, but the newest driver available from Nvidia is 241.96, I do not believe it is new enough for your open CL apps. ID: 1815462 · Reply Quote

Kiska Volunteer tester Send message Joined: 31 Mar 12 Posts: 302 Credit: 3,067,762 RAC: 0	Message 1815476 - Posted: 7 Sep 2016, 1:05:16 UTC - in response to Message 1815462. Last modified: 7 Sep 2016, 1:13:48 UTC We have 8.18 deployed on beta, test please. Hopefully they are new RC for main. I have switched 2 of my computers to beta, the first is a xeon E5-2670 8 core 16 thread with 2 GTX 750ti's the other is an AMD 1090T 6 core with 3 GTX 550ti's I've got 2 gt 240, here on an E8500 core2duo, but the newest driver available from Nvidia is 241.96, I do not believe it is new enough for your open CL apps. Latest version for GTS 240(which I am assuming, couldn't find GT 240) is 337.88, which is located here for windows Which does not result in invalids ID: 1815476 · Reply Quote

robertmiles Volunteer tester Send message Joined: 16 Jan 12 Posts: 213 Credit: 4,117,756 RAC: 6	Message 1815482 - Posted: 7 Sep 2016, 1:28:24 UTC - in response to Message 1815406. Last modified: 7 Sep 2016, 1:28:48 UTC Raistmer, suggestions for future improvements: Check if Nvidia OpenCL offers a way for an application to check the temperature of its GPU. If so, occasionally check it, and print the maximum value somewhere in stderr. This should make overheating as a cause of problems more obvious. Also, when the Task postponed: Suspicious spike results notice is issued, mention it in stderr to help identify which task it was from. There is no temperature reading interface in OpenCL. Also, no need to link to particular task in general, What is needed is the knowledge that smth wrong with computational device. One could look into tasks list and see what task is postponed. Writing that something is wrong with the GPU, with no checkable hint on what, does not help me find the problem. I am NOT going to get another 560 to try instead. As for postponed tasks, that seldom lasts long enough for me to find time for looking at the tasks list. ID: 1815482 · Reply Quote

J. Mileski Volunteer tester Send message Joined: 9 Jun 02 Posts: 632 Credit: 172,116,532 RAC: 572	Message 1815628 - Posted: 7 Sep 2016, 22:39:44 UTC - in response to Message 1815476. We have 8.18 deployed on beta, test please. Hopefully they are new RC for main. I have switched 2 of my computers to beta, the first is a xeon E5-2670 8 core 16 thread with 2 GTX 750ti's the other is an AMD 1090T 6 core with 3 GTX 550ti's I've got 2 gt 240, here on an E8500 core2duo, but the newest driver available from Nvidia is 241.96, I do not believe it is new enough for your open CL apps. Latest version for GTS 240(which I am assuming, couldn't find GT 240) is 337.88, which is located here for windows Which does not result in invalids No, it is not a GTS 240, it really is a GT240 ID: 1815628 · Reply Quote

Wiggo Send message Joined: 24 Jan 00 Posts: 34744 Credit: 261,360,520 RAC: 489	Message 1815639 - Posted: 7 Sep 2016, 23:13:49 UTC I've got 2 gt 240, here on an E8500 core2duo, but the newest driver available from Nvidia is 241.96, I do not believe it is new enough for your open CL apps. Those GT 240's are using the latest (341.96, I noticed the typo) and very likely the last driver you'll be able to use with them as all pre Fermi GPU's are now regarded by Nvidia as legacy products so they will not be releasing anymore driver updates for them. Cheers. ID: 1815639 · Reply Quote

J. Mileski Volunteer tester Send message Joined: 9 Jun 02 Posts: 632 Credit: 172,116,532 RAC: 572	Message 1815641 - Posted: 7 Sep 2016, 23:31:14 UTC - in response to Message 1815639. Sorry, yes it was a typo. ID: 1815641 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1815692 - Posted: 8 Sep 2016, 5:05:09 UTC Here https://cloud.mail.ru/public/5Kth/LJGfW7s5i is new builds, r3528. Please test if both recently found issues (CL file compilation, high prec timer omission in low-perth path) are fixed. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1815692 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1815891 - Posted: 9 Sep 2016, 11:35:42 UTC Finally, cards kindly donated by Mr. Kevvy and those shipment was organized and carried by Mike received and tested. Both in fine working condition. I would like to thank you both, Mr. Kevvy and Mike, for this action. Hope it will help to make some new discoveries on NV-specific optimization path. Thank you! SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1815891 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1815896 - Posted: 9 Sep 2016, 12:12:09 UTC And first results: both GTX460SE and GTX570 confirmed compilation issue on PTX level for 8.18 NV build (errors wording is different but both failed to build CL file). Also confirmed that r3528 is free from this issue. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1815896 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1817033 - Posted: 14 Sep 2016, 3:56:17 UTC Last modified: 14 Sep 2016, 3:56:43 UTC 8.19 (r3528) is deployed on beta. Please test. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1817033 · Reply Quote

robertmiles Volunteer tester Send message Joined: 16 Jan 12 Posts: 213 Credit: 4,117,756 RAC: 6	Message 1817054 - Posted: 14 Sep 2016, 5:18:54 UTC - in response to Message 1815896. And first results: both GTX460SE and GTX570 confirmed compilation issue on PTX level for 8.18 NV build (errors wording is different but both failed to build CL file). Also confirmed that r3528 is free from this issue. Does r3500 share this problem or is it with some change since r3500? Does it affect the beta4 now under test? ID: 1817054 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1817056 - Posted: 14 Sep 2016, 5:45:38 UTC - in response to Message 1817054. And first results: both GTX460SE and GTX570 confirmed compilation issue on PTX level for 8.18 NV build (errors wording is different but both failed to build CL file). Also confirmed that r3528 is free from this issue. Does r3500 share this problem or is it with some change since r3500? Does it affect the beta4 now under test? r3500 doesn't have it. r3528 doesn't have it too (8.19) SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1817056 · Reply Quote

Raistmer Volunteer developer Volunteer tester Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121	Message 1817146 - Posted: 14 Sep 2016, 16:53:28 UTC Last modified: 14 Sep 2016, 16:53:45 UTC Just remind: the more different hosts will participate in beta currently the earlier we will be able to get main updated. I would recommend to switch most computational resourses to beta for 1-2 next weeks. SETI apps news We're not gonna fight them. We're gonna transcend them. ID: 1817146 · Reply Quote

AMDave Volunteer tester Send message Joined: 9 Mar 01 Posts: 234 Credit: 11,671,730 RAC: 0	Message 1817167 - Posted: 14 Sep 2016, 19:33:29 UTC - in response to Message 1817033. Last modified: 14 Sep 2016, 19:33:51 UTC Are there any default settings for switches (exÂ Â -sbs N where N=128) of which we should be aware? ID: 1817167 · Reply Quote

Jimbocous Volunteer tester Send message Joined: 1 Apr 13 Posts: 1853 Credit: 268,616,081 RAC: 1,349	Message 1817170 - Posted: 14 Sep 2016, 19:45:10 UTC - in response to Message 1817146. I would recommend to switch most computational resourses to beta for 1-2 next weeks. There goes my RAC :) You got it ... ID: 1817170 · Reply Quote

©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.

SETI applications for NVIDIA GPU improvement - how you can help