| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: Estimated runtime on Laptop GPUs wayyyyy off |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 5
|
|
| Author |
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges: 
|
Hi,
----------------------------------------I've notice that the estimate runtime on Laptop GPUs is way off. I get an estimates ~15-25minutes / WU, when in fact it takes 1,5-3hour to complete the WU (depending on laptop). In any case the estimate is very very very optimistic (read: unrealistic). This causes scheduler to request 5-10x more work than I can run. I see the same behavior on both intel and AMD gpus (dont have nvidia on laptops), but it seems that Intel is much worse. |
||
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges:
|
And since I have much work I cant crunch in time, should I abort those jobs, or leave it to scheduler?
---------------------------------------- |
||
|
|
goben_2003
Advanced Cruncher Joined: Jun 16, 2006 Post Count: 146 Status: Offline Project Badges: 
|
I left it to the scheduler. It aborts them if you do not start them in time. Then I reduced the requested cache so I would not have so many intel gpu units at a time and not need the scheduler to abort any in the future.
---------------------------------------- |
||
|
|
kwolff88
Cruncher Joined: Dec 31, 2004 Post Count: 19 Status: Offline Project Badges: 
|
While I can't comment on laptop GPUs directly, I've observed that BOINC estimates tend to be incorrect at the start, but as you complete more work units, it gradually learns what your system can do and they become more accurate.
|
||
|
|
biini
Senior Cruncher Finland Joined: Jan 25, 2007 Post Count: 334 Status: Offline Project Badges:
|
It's been like this for three weeks / hundreds of WUs
---------------------------------------- |
||
|
|
|