| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: Did the maximum number of work units per device increase? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 5
|
|
| Author |
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges: 
|
In the past, even if I increased the maximum allowable days of work, my threadripper could download around 500 tasks before it would stop accepting new work with the message:
Not requesting tasks: too many runnable tasks. If you're at 499, you could exceed the cap by asking for more - the servers will give one final set of tasks. Today I noticed my computer was assigned over a thousand tasks in progress! I'm not sure when the cap was lifted, but it should allow everyone to continue crunching during WCG downtimes and decrease the pressure on the Jurisica lab to match IBM's server stability. So overall, I really like the change. Is it possible to separate the cap by CPU and GPU work? Now that my computer is full of CPU tasks, it's not requesting GPU work either. So my maximum throughput has decreased. |
||
|
|
Aperture_Science_Innovators
Advanced Cruncher United States Joined: Jul 6, 2009 Post Count: 139 Status: Offline Project Badges: 
|
This is actually a BOINC thing , not a WCG thing -- there's a built-in limit of 1000 tasks maximum. Most of my dual-CPU Xeon systems regularly have somewhere between 700 and 1000 tasks each (not lately with the work shortage, but otherwise).
----------------------------------------I've actually considered building a custom BOINC version to bump this up to 2k or 3k, because my EPYC system can still blast through its 1000 tasks in about a day, making it too low of a cache for comfort. But haven't had the time to experiment with it yet.  |
||
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges:
|
Projects have their own limits.
GPU Grid won't let you download more than a few projects at once. |
||
|
|
Aperture_Science_Innovators
Advanced Cruncher United States Joined: Jul 6, 2009 Post Count: 139 Status: Offline Project Badges:
|
I believe the message you were seeing though, "Not requesting tasks: too many runnable" is BOINC only. The servers for the projects may have a separate limit, but I'm nearly certain you'd get a different error message if so (I've not spent a lot of time poking around in the source code, but the only place I've seen the "too many runnable" seems to be when BOINC hits its limit of 1000).
---------------------------------------- |
||
|
|
Mad_Max
Cruncher Russia Joined: Nov 26, 2012 Post Count: 22 Status: Offline Project Badges: 
|
Yes, for a project side limits there are another messages
Like "No work sent. Reached daily quota of ั ั ั tasks" - this one for the per day limit set by project. It can be dynamic like decreasing with every returned task error and increasing after successful validation on returned WUs (up to some static maximum) And "No tasks sent This computer has reached a limit on tasks in progress" for number of WUs downloaded and stored on user host limited by project (sum of all WUs on client side: unprocessed just sitting in work queue , currently in progress, and completed but not returned to server yet) For example right now OPNG has such limit set to 50 WUs in progress per each GPU So host with 1 GPU can get up to 50 WUs in local BOINC cache, with 2 GPUs up to 100 WUs, etc. |
||
|
|
|