Message boards :
Number crunching :
Estimated remaining times a bit out of track
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I know Matt knows this, but I would like to share my observations. When starting with the test CPU WU's, that took way longer then firstly anticipated, the estimate time of the GPU WU's went up too. Now i have no longer CPU WU's in the queue (waiting for the improved win app), the remaining estimation of the GPU WU's is more then 180 hours. That is a bit to long for a GTX780Ti, Greetings from TJ |
MJHSend message Joined: 12 Nov 07 Posts: 696 Credit: 27,266,655 RAC: 0 Level ![]() Scientific publications ![]()
|
Were you running with CPU allocated to the GPU tasks, or did you let the CPUMD tasks take all your CPU? Matt |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Were you running with CPU allocated to the GPU tasks, or did you let the CPUMD tasks take all your CPU? Yes 5 threads where allowed to run CPU and 2 for the GPU's and 1 "spare". Greetings from TJ |
|
Send message Joined: 25 Sep 13 Posts: 293 Credit: 1,897,601,978 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
http://www.gpugrid.net/workunit.php?wuid=10159833 TJ- this task is at step 4.5million with little time left until the deadline. It won't complete before deadline- as we know about GPUGRID tasks validated after deadline- GPUGRID grants zero credit. Five days worth of compute time will be down the drain. The re-made deadline to seven days would of been sufficient. I already had 24 hr invested before the 7 day deadline change for CPUMD tasks. |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I know that, but am talking about the estimates of the GPU WU's. I have one that is 45% ready in 2 hours and estimated time is still 153 hours to go. Well within 3 hours it is finished. Greetings from TJ |
|
Send message Joined: 11 Jul 09 Posts: 1639 Credit: 10,159,968,649 RAC: 318 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
I'm seeing the same with the v8.47 applications released this morning. That's nothing to worry about - they seem to be running at the normal speed, so will finish at the normal time - it's just BOINC's notorious failure to get its estimates right for the first 101 tasks (overall) or 11 tasks (per host) after a new version is deployed. |
|
Send message Joined: 25 Sep 13 Posts: 293 Credit: 1,897,601,978 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Task is running at normal speed with estimates all over the place- I can confirm- a GPU task at 91% complete--- estimated remaining time is 100hr-- a GPU task at 68% complete- estimated time is 200hr. I downloaded a Primegrid CPU Genefer task to see if affected and estimated time is triple of what task normally finishes. Before CPUMD tasks- all tasks had proper estimates for BOINC 7.4.21. Starting with an empty cache and having no tasks for BOINC. Estimates are still over the place. |
|
Send message Joined: 26 Jun 09 Posts: 815 Credit: 1,470,385,294 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
Strange issue is that I have run the CPU WU's on two rigs that also run the GPU WU's. With one rig the estimate times are normal again, at another rig it is still very long, more then 180 hours when starting. So that should indicate that a new version (8.47) is not the cause of it. Both rigs run with 8.47 since yesterday. Greetings from TJ |
©2025 Universitat Pompeu Fabra