| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: Planning ahead (GPU Discussion.) |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 3
|
|
| Author |
|
|
OldChap
Veteran Cruncher UK Joined: Jun 5, 2009 Post Count: 978 Status: Offline Project Badges: 
|
I may have this all wrong and if so I apologise but if my understanding is correct then shortly before this hiatus each gpu running in a rig was allowed to hold a maximum of 400 WU's regardless of type and speed.
----------------------------------------Further, there is a maximum of 2500 wu's a day allowed. Is this per rig or per card? (I have been trying to find where I read this) 1. Would any of you kind and informed folk care to talk a little about how we can make GPU follow the normal cache rules going forward? It would be nice to be able to set things normally but I can see, especially when one is running app_info's, that this may be technically challenging and currently setting more than one instance to run GPU messes with the projected completion times a lot. 2. Those of you with a greater understanding of both WCG and BOINC I ask this: How much effort would it take to implement that which is achieved by using app_info's but running in the normal environment? Do you think this is even achievable?  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Never understand the difficulty of finding most anything:
----------------------------------------http://www.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=397016 http://www.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=399616 Used to be per-core/thread under server 601 up to 80 per day, but that could have changed with server 700. If 40 "in progress" per host, then the total on my octo does not tally as it has some 100, so guess it's per thread/card. The 15 per call assigned... well I've caught my octo and quad getting 1 at a time in a longer series when alternating the buffer between 0 and 1 day, meaning it will then get the buffer filled in 20-30 serial requests. Can't see that as helping the scheduler, but if that's the method to ensure that every host that calls for work get's at least one of any active type it asks for, that's fine by me [not my problem so to speak]. edit: spell [Edit 1 times, last edit by Former Member at Nov 22, 2012 9:08:31 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Of course, seen this 1 at the time repeatedly and then testing just now got this... maybe it's now 15 maximum per thread :D... as said, [not my problem... however the WCG techs deem things run best].
8349 World Community Grid 22-11-2012 21:52:36 [sched_op] Starting scheduler request 8350 World Community Grid 22-11-2012 21:52:36 Sending scheduler request: To fetch work. 8351 World Community Grid 22-11-2012 21:52:36 Requesting new tasks for CPU 8352 World Community Grid 22-11-2012 21:52:36 [sched_op] CPU work request: 1971313.67 seconds; 0.00 devices 8353 World Community Grid 22-11-2012 21:52:43 Scheduler request completed: got 86 new tasks 8354 World Community Grid 22-11-2012 21:52:43 [sched_op] Server version 701 8355 World Community Grid 22-11-2012 21:52:43 Project requested delay of 11 seconds 8356 World Community Grid 22-11-2012 21:52:43 [sched_op] estimated total CPU task duration: 1981897 seconds |
||
|
|
|