I wasn't getting any new work from E@H on my GPU computer

capnrob97
capnrob97
Joined: 9 Feb 05
Posts: 15
Credit: 53,788,270
RAC: 653,401
Topic 204123

I run several projects on it.

Finally, I suspended all the projects on it and I got a ton of E@H workunits.

Is it normal for a project to not send any WUs if it is busy crunching other projects?

I deleted seti@home for now and re-enabled the rest, we will see if I get more on this computer when these are gone.

archae86
archae86
Joined: 6 Dec 05
Posts: 2,556
Credit: 1,849,747,456
RAC: 2,560,823

capnrob97_2 wrote:Is it

capnrob97_2 wrote:
Is it normal for a project to not send any WUs if it is busy crunching other projects?

BOINC on your computer decides what requests to make of what projects.  The Einstein project does not make judgments on your balance with other projects.  If you got "a ton" of work you probably have an excessively large queue depth setting.  That is an easy way to get projects out of balance with each other if there are intermittent availability issues.  If BOINC on your computer thinks them out of balance, it may make no request to a particular one for quite a while.

Holmis
Joined: 4 Jan 05
Posts: 933
Credit: 541,339,546
RAC: 777,953

capnrob97 wrote:I run several

capnrob97 wrote:
I run several projects on it.
Finally, I suspended all the projects on it and I got a ton of E@H workunits.

When running more than one project one should not expect there to always be work for all projects in the cache at all times. It all depends on how you have your "Resource share" setup between projects and if your projects have been able to provide work when Boinc on your computer requested it.

If you open Boinc Manager on your computer and look in the "Projects" tab of the "Advanced view" you should have a column named "Resource share". Having big differences in resource share makes it harder for Boinc to try and live up to your wishes and if projects can't deliver work when requested that also makes things worse.

Best advice is to set a fairly low work cache when running multiple projects and then just let it run, it should even out somewhere in the region of your set resource share, but it will take time and sometimes that can be in the order of weeks or months.

capnrob97 wrote:
Is it normal for a project to not send any WUs if it is busy crunching other projects?
I deleted seti@home for now and re-enabled the rest, we will see if I get more on this computer when these are gone.

Einstein is fairly stable and most of the time you get all the work you request compared to Seti that has been more unstable and less prone to fulfill work request, I would guess that your computer had done more Einstein work than your resource share was set to and so was trying to catch up with Seti work. This might also be true for other projects you are running.
When the newly downloaded Einstein work is done I suspect that Boinc on your computer will try to catch up on other projects for a while, so you should expect there to not be any Einstein work in your cache for a while.
If you change you resource share things will be different and if you start to suspend projects things will probably be worse than just letting it run and sort itself out.

capnrob97
capnrob97
Joined: 9 Feb 05
Posts: 15
Credit: 53,788,270
RAC: 653,401

I have resource share set to

I have resource share set to 100 for all the projects.

 

They all get 16.67%

 

Holmis
Joined: 4 Jan 05
Posts: 933
Credit: 541,339,546
RAC: 777,953

capnrob97_2 wrote:I have

capnrob97_2 wrote:
I have resource share set to 100 for all the projects.
They all get 16.67%

So you run 6 projects. Then I would advice you to have a very small work cache setting, something like 0.5 days total. Boinc should always ask for work from the project that has done the least amount of work in recent time first. The more often Boinc asks for work the greater the chance of that request going through. That will help Boinc to honor your set resource share.

With so many projects active the chance for all of them being unable to provide work and you running out of work to do seems quite remote, unless your Internet connection is unreliable.

capnrob97
capnrob97
Joined: 9 Feb 05
Posts: 15
Credit: 53,788,270
RAC: 653,401

I have my work cache setting

I have my work cache setting to .1, maybe that is it?

I also have my 'Use at most % of CPus' set to 87.5% (7 out of 8 cpus) because I thought I read somewhere to keep a cpu free to feed GPU intensive tasks.

Maybe I should bump 87.5 up to 95%?

 

Holmis
Joined: 4 Jan 05
Posts: 933
Credit: 541,339,546
RAC: 777,953

capnrob97 wrote:I have my

capnrob97 wrote:
I have my work cache setting to .1, maybe that is it?


That's good, keep it at 0.1, increasing the cache size will only make it harder for Boinc to balance the resource share. 

Boinc probably thinks you've done to much Einstein work and are trying to catch up on other projects. I suspect that one or more of the other projects have had problems delivering work a while ago so Boinc ran work from Einstein instead, that debt is now being repaid. If you leave Boinc to do it's thing for a few days/weeks it should go back to normal and alternate between your different projects.

capnrob97 wrote:

I also have my 'Use at most % of CPus' set to 87.5% (7 out of 8 cpus) because I thought I read somewhere to keep a cpu free to feed GPU intensive tasks.

Maybe I should bump 87.5 up to 95%?


This shouldn't have any effect on what projects Boinc thinks should run due to resource share.
But mixing projects with both GPU and CPU apps with CPU only projects will probably lead to the CPU only projects dominating the CPU work on the host while the GPU/CPU project will only use the GPU.

capnrob97
capnrob97
Joined: 9 Feb 05
Posts: 15
Credit: 53,788,270
RAC: 653,401

Thank you Holmis for the

Thank you Holmis for the info, I appreciate it

Holmis
Joined: 4 Jan 05
Posts: 933
Credit: 541,339,546
RAC: 777,953

Happy to help!

Happy to help! Wink

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.