Message boards :
Number crunching :
Astropulse time estimation
Message board moderation
Author | Message |
---|---|
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
Hello there I have been doing Astropulse with my GPU under the anonymous platform and my allotment is 100 tasks for my cache. The estimations are 53 minutes for each task. I have just recently started doing CPU Astropulse tasks 601 CPU sse r2083 and the cache will only fill up to 20 tasks because the time estimation is 161 hours for each task. Normally these estimations come down dramatically after a few tasks have been crunched thus telling the server I can take far more. Now each new task coming in is 161 hours even though they only take 10-11 / task Is there a way to solve this problem. Wisdom file has generated for 2083 tasks. Would deleting the wisdom file help? The estimations are only off for CPU tasks as my 604 SSE2 r2058 GPU tasks are okay and 100 is my limit. Brings me to ask is there a CPU app for Astropulse that will use sse2 compared to just sse planned? Thank You for any help at all. Michael Miles Team New Mexico Amd Phenom II 1100T Boinc 7.2.39 x64 AP6_win_x86_SSE2_OpenCL_NV_r2058 AP6_win_x86_SSE_CPU_r2083 |
betreger Send message Joined: 29 Jun 99 Posts: 11362 Credit: 29,581,041 RAC: 66 |
IIRC you have to get 10 tasks validated before the time estimates are corrected. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
actualy is after 11... |
Wiggo Send message Joined: 24 Jan 00 Posts: 34945 Credit: 261,360,520 RAC: 489 |
And arn't they suppose to be tasks with no blanking? Cheers. |
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
Ten tasks before time goes to normal. Will do and we shall see if it does come down. Thanks for the replies I just hope AP stays up for a while so I can get a full cache for the week Michael Miles |
Mike Send message Joined: 17 Feb 01 Posts: 34261 Credit: 79,922,639 RAC: 80 |
And arn't they suppose to be tasks with no blanking? No, why should it be different as on GPU ? 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 |
And arn't they suppose to be tasks with no blanking? Yes, with less than 5 or 10% Blanking (I forget which) Claggy |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Ten tasks before time goes to normal. Validated task only! Until you reach that number just increase the cache size to 10 days, that will give you some more WU, not to many but it´s what you could do to get some more. Just not forget to back it to your normal number when the times stabilizes. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34945 Credit: 261,360,520 RAC: 489 |
And arn't they suppose to be tasks with no blanking? Thanks Claggy, I knew that I read somewhere that there were certain parameters involved there. Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
You can quickly fix it by adding a FLOPS entry into the app_info file. You must be careful though, too high a value will result in too short an estimated runtime. This will cause any work in progress to exit with a time exceeded error. An approximate value would be around 40000000000, you would enter it as such; <app_version> <app_name>astropulse_v6</app_name> <version_num>602</version_num> <flops>40000000000</flops> <file_ref> <file_name>ap_6.02r561_OSX64</file_name> <main_program/> It would probably be a good idea to remove one of the zeros on the first attempt if you have any work in progress. You might want to fine tune the number based on what is displayed in the Remaining (estimated) column after restart. A higher number will give a lower estimated time. |
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
You can quickly fix it by adding a FLOPS entry into the app_info file. I thought that is what the wisdom file did. Michael Miles |
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
Until you reach that number just increase the cache size to 10 days, that will give you some more WU, not to many but it´s what you could do to get some more. Just not forget to back it to your normal number when the times stabilizes It is already 10 and 10 additional. I'll wait until I get 5 more done that should give me the 11 I need to have the server give more work in cache I'm amazed at the difference from CPU to GPU. 53 minutes by GPU 460 768 meg or 11 hours by CPU Phenom 2 1100T Curious as to why there are only SSE Astropulse apps and not an SSE2 Would not SSE2 be faster than just SSE? Michael Miles |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
I thought you wanted a quick fix...nevermind. |
Mike Send message Joined: 17 Feb 01 Posts: 34261 Credit: 79,922,639 RAC: 80 |
Until you reach that number just increase the cache size to 10 days, that will give you some more WU, not to many but it´s what you could do to get some more. Just not forget to back it to your normal number when the times stabilizes GPU is much more efficient than CPU. Especially AMD. We are working on new apps. Stay tuned. With each crime and every kindness we birth our future. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
If you don´t do, try to run 2 WU at a time. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
Estimates will just about instantly fix themselves after you have completed 10 tasks. This means 10 tasks that have less than 10% blanking, and do not early exit with 30/30. Any NEW tasks that are assigned after 10 tasks have completed will have reasonable estimates, but any that you already have in your cache will not magically change once that happens (I think..its been a while). DCF itself will bring the estimates down, but very slowly at first. Since they are completing in much less time than the estimate, the DCF estimate only adjusts downward by I think a 10th of a percent (0.1%) or something like that. As you start getting closer, the estimate will start shifting in larger intervals. I remember when I started a new machine. It was 180 hours to do what only takes 12 hours. At first, the adjustment was only a few seconds, and after about 10 tasks, it started becoming several minutes at a time, until eventually it started taking hours off at a time off. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
If you don´t do, try to run 2 WU at a time My GPU is showing very good load. I get 90% load with one task running. If I run 2 at a time I get problems with watching video at the same time. I do not own a TV so downloaded content it what I watch. This machine runs 24/7 I have no problems with GPU tasks. The only problem with CPU is the estimates. Will wait till it stabilizes itself Ap should be up for another day hopefully Thank you for all the replies Michael Miles |
Batter Up Send message Joined: 5 May 99 Posts: 1946 Credit: 24,860,347 RAC: 0 |
OP, 600 cobblestones credit for 2 seconds work; I would stick with it. http://setiathome.berkeley.edu/workunit.php?wuid=1423371188 |
Mike Send message Joined: 17 Feb 01 Posts: 34261 Credit: 79,922,639 RAC: 80 |
OP, 600 cobblestones credit for 2 seconds work; I would stick with it. Yep, credit new works perfect. With each crime and every kindness we birth our future. |
Michael W.F. Miles Send message Joined: 24 Mar 07 Posts: 268 Credit: 34,410,870 RAC: 0 |
OP, 600 cobblestones credit for 2 seconds work; I would stick with it. That is wild eh! That is the reason I am switching to AP cpu from V7 The time per task and credit given for v7 makes no sense doing them AP is double credit for computation time in compared to v7 The new credit system is just a laugh and I feel it should be changed Michael |
©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.