| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: Double-size GPU work units (BOINC 7.0.38 problem?) |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 10
|
|
| Author |
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
In the past hour or so, the HCC1 GPU work units have been using 13% CPU on my Core i7-3770 during the CPU-phase, as measured by Task Manager. This is a full virtual core, rather than the 5% to 6% I usually get. There was some mention of double-work units recently, and so maybe this is it. But I don't see any change on another of my PCs, and so the roll-out may take some time. And I don't have any validated results yet to see if the points per work unit is higher, but that is the next thing to check.
----------------------------------------For people having to hassle with app_info files trying to get more work out of your cards, this may make it not worth the trouble. [Edit 1 times, last edit by Jim1348 at Nov 4, 2012 10:13:10 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You know the expression "Trust Me" and that applies to this not having happened without the pre-cursor Beta that knreed mentioned would happen [Both CPU + GPU jobs will be doubled, and not seen them even the freshest downloads doing this]. No sinecure to package 2 in one job and get them out right again.
|
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
Then they must have just gotten more difficult. My power supply is now drawing 10 watts more than before (during the GPU phase), which is what I would normally see when running two work units on this card (HD 7770). Whether there will be any more points is another matter. But for my purposes, I don't really care whether they call it "double" or not, as long as it is getting more work done.
----------------------------------------[Edit 1 times, last edit by Jim1348 at Nov 4, 2012 2:57:49 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There's always been sets of harder targets in the mix. To make it easy, check the logs on the Result Status page. Surely they will give away if there were 2.
----------------------------------------Twice as long, drawing 10 watts more... something gave way [No, not my PC, that's *always* been rocksolid]. I understand these 2 in 1 tasks will run serial. If they would not, there'd be problems for those that have specified 9 to run concurrent on their 7970 [Another reason why the Tech would not let these loose without some form of pre-warning to "stretch the comp" crunchers. My Doolittle quad is being assignments refused because 175MB is demanded per task when there's only 128MB on offer. [Edit 1 times, last edit by Former Member at Nov 4, 2012 3:19:34 PM] |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
I understand these 2 in 1 tasks will run serial. If they would not, there'd be problems for those that have specified 9 to run concurrent on their 7970 Yes, that is what I had understood too. But I haven't seen these "harder" ones before. And as noted above, they are too new to see any results, but they should be coming in shortly. If anything gave-way on my PC, it doesn't seem to be hurting the work units; they are completing in their usual time of 2 minutes 10 seconds; they just use twice as much CPU and GPU power. Unfortunately, I can't check the GPU usage directly, since the latest version of GPU-Z does not work with the Catalyst 12-10 drivers, so that is why I am relying on the power usage, but that is also a reliable indicator. But I have seen cases (on other projects) where the same work unit will run twice at the same time, thereby consuming twice as many resources, so that is possible and I will have to check it out. So something is up, whatever it is. [Edit 3 times, last edit by Jim1348 at Nov 4, 2012 3:50:02 PM] |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: 
|
For people having to hassle with app_info files trying to get more work out of your cards, this may make it not worth the trouble. I don't think it will be an issue. Just long crunch time. Might have to back down a few but until they come out, won't be able to tell how well it works. Will be interesting how its coded. Will it pause in between, use the cpu, then back to gpu or straight crunching once the gpu is engaged.
Crunching for humanity since 2007!
 |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
I am getting my usual 30 points per work unit, so no doubling-up of credit. But there are no errors either. I think it is pretty clear that BOINC is running two of the same work units at once. That is probably related to the fact that I am running the latest alpha (7.0.38), which has changes related to scheduling of GPU work units from multiple projects. Very likely after I removed the app_info file (for running two HCC WUs at once), something got stuck in BOINC, even though I rebooted.
So after the cache is empty, I will uninstall BOINC (even the data folder), and re-install 7.0.36, which works fine for WCG. Problem solved, I think. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
If what you're saying, 2 separate jobs run concurrent on your GPU, that could be the doing of app_info. BOINC itself through 7.0.38 [runs fine here], nor WCG are pushing this to run more than 1 per card. What I'm not sure about, but suspect is, that some elements of app_info are pulled into the client_state.xml [a file you don't wanna mess with]. Just removing app_info and restarting the client wont make the change until the loaded tasks have all cleared.
|
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
That is probably it, though I never had a problem before after getting rid of the app_info. But I have cleared out all the GPU tasks, and am finishing the CPU tasks now.
I think I will just allow more in rather than reinstall everything, and see if it works. If not, I will post back. Thanks for the input. |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
Just removing app_info and restarting the client wont make the change until the loaded tasks have all cleared. Removing the app_info before emptying your cache will cause you to lose any remaining tasks in your cache.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
|