Been burning-in a new i7/win7-64 unit with just Seti for about a month, now want to share it with rah.
Connected, and shortly after noticed that BOINC was requesting work for not only CPU, but GPU too. In fact requesting GPU every 4 minutes. The machine does have a CUDA capable card, but near as I can tell, rah does not have GPU crunchers, and I have SETI set to not use the GPU (not ever, so far, maybe later....).

will the GPU requests go away as it gets satisfied with more downloads of CPU work, or will it continue to bother the servers for work they don't have?
I think I have 6.10.18 BOINC. Or maybe is there a CC directive that says "don't request GPU from rah?"

Dave

Been burning-in a new i7/win7-64 unit with just Seti for about a month, now want to share it with rah.
Connected, and shortly after noticed that BOINC was requesting work for not only CPU, but GPU too. In fact requesting GPU every 4 minutes. The machine does have a CUDA capable card, but near as I can tell, rah does not have GPU crunchers, and I have SETI set to not use the GPU (not ever, so far, maybe later....).

will the GPU requests go away as it gets satisfied with more downloads of CPU work, or will it continue to bother the servers for work they don't have?
I think I have 6.10.18 BOINC. Or maybe is there a CC directive that says "don't request GPU from rah?"

Dave

It is a Boinc thing, it may never go away, however on projects that do support your gpu there is a setting to not get gpu work. You can safely ignore it, it can't get what a project does not have to offer.

Not a problem for Crunchers (just ingore this messages) but is problem for project servers: Such clients activity are creating a lot of unnecessary queries, and accordingly a lot of unnecessary / excessive load on the project servers.

P.S.
My BOINC 6.10.18 is also constantly asks for GPU tasks from RAH, but at least not every 4 minutes, but only together with reports on completed tasks and when ask new tasks for the CPU

I apologize for the thread necromancy... however I was wondering if there was any update here. I only run rosetta which has no GPU work and was wondering if there was any way to stop these GPU scheduler requests. Even with "Use GPU Never" selected it still sends these requests. Lately, with the low amount of work available, it's wasting several scheduler requests for GPU work when the CPUs are idle... i.e. Requesting work for GPU 3-4 times before it even attempts to request work for the CPUs.

---

I apologize for the thread necromancy... however I was wondering if there was any update here. I only run rosetta which has no GPU work and was wondering if there was any way to stop these GPU scheduler requests. Even with "Use GPU Never" selected it still sends these requests. Lately, with the low amount of work available, it's wasting several scheduler requests for GPU work when the CPUs are idle... i.e. Requesting work for GPU 3-4 times before it even attempts to request work for the CPUs.

Either create, or edit cc_config.xml (mine was in C:Documents and SettingsAll UsersApplication DataBOINC) and put:

<cc_config>
<log_flags>
</log_flags>
<options>
<no_gpus>1</no_gpus>
</options>
</cc_config>

1 = no gpu, 0 = default gpu. You can either restart BOINC or go to Advanced|Read Config File and push it through yourself. You should read "Config: Don't use Coprocessors" if it worked right. Of course I'm not getting any work right now and my comm is deferred for some time, but I'm 90% sure it's the proper way to stop GPU req's from the server.

+1.

Perfect! Worked great. Thanks very much!

---

I was wondering if there was any update here. I only run rosetta which has no GPU work and was wondering if there was any way to stop these GPU scheduler requests. Even with "Use GPU Never" selected it still sends these requests. Lately, with the low amount of work available, it's wasting several scheduler requests for GPU work when the CPUs are idle... i.e. Requesting work for GPU 3-4 times before it even attempts to request work for the CPUs.

Either create, or edit cc_config.xml (mine was in C:Documents and SettingsAll UsersApplication DataBOINC) and put:

<cc_config>
<log_flags>
</log_flags>
<options>
<no_gpus>1</no_gpus>
</options>
</cc_config>

1 = no gpu, 0 = default gpu. You can either restart BOINC or go to Advanced|Read Config File and push it through yourself. You should read "Config: Don't use Coprocessors" if it worked right. Of course I'm not getting any work right now and my comm is deferred for some time, but I'm 90% sure it's the proper way to stop GPU req's from the server.

Thanks. I got the proper response you described too. I'll confirm later if it's had the desired effect. Anything to reduce server hits must be a good thing.

---

Thanks. I got the proper response you described too. I'll confirm later if it's had the desired effect. Anything to reduce server hits must be a good thing.

That did work well, thanks. I'll copy that over to my other machines now.

So after checking back in on this thread, I see that setting this in the cc config does indeed make it not request GPU tasks. But what if, in one machine, I wanted s@h to use the GPU, and wanted r@h to not waste time requesting GPU work?

Is there a way to differentiate based on project within one host?

thx

Dave

So after checking back in on this thread, I see that setting this in the cc config does indeed make it not request GPU tasks. But what if, in one machine, I wanted s@h to use the GPU, and wanted r@h to not waste time requesting GPU work?

Is there a way to differentiate based on project within one host?

thx

Dave

yes you can do it thru the website by having different settings for work, home or school or the default and then setting each pc where you want it. You go under Your Account, Computing Preferences and then at the bottom you can see where you can add separate preferences for each place. The place name is immaterial but the preferences can be set so you can do gpu only crunching under home and cpu only crunching under school and then crunch both under work, for example. After you attach a pc you can then go into view your computers and at the bottom change it to the different places.

So after checking back in on this thread, I see that setting this in the cc config does indeed make it not request GPU tasks. But what if, in one machine, I wanted s@h to use the GPU, and wanted r@h to not waste time requesting GPU work?

Is there a way to differentiate based on project within one host?

thx

Dave

Yes you can do it thru the website by having different settings for work, home or school or the default and then setting each pc where you want it. You go under Your Account, Computing Preferences and then at the bottom you can see where you can add separate preferences for each place. The place name is immaterial but the preferences can be set so you can do gpu only crunching under home and cpu only crunching under school and then crunch both under work, for example. After you attach a pc you can then go into view your computers and at the bottom change it to the different places.

You could still see the status message 'not requesting any gpu work' even if you change your settings though. This is JUST a status message and the programmers think it is a good thing.

So after checking back in on this thread, I see that setting this in the cc config does indeed make it not request GPU tasks. But what if, in one machine, I wanted s@h to use the GPU, and wanted r@h to not waste time requesting GPU work?

Is there a way to differentiate based on project within one host?

thx

Dave

I wanted the same thing for my single host setup as well.

I tried to run the no-gpu command with my SETI account preferences set to USE GPU = Yes and USE CPU = no. It failed as it pulled "missing coproc" for all my SETI@HOME tasks. That was my hope, that profiles could override the client config but in lieu of that, I've got my s@h setup only running GPU tasks and not fetching CPU tasks, and when running it side by side with r@h I don't catch it only asking for GPU tasks for Rosetta, but it still tags it on, as all the requests are fetching CPU and GPU tasks. It still gives you your cpu tasks, but it doesn't stop it from sending out the gpu request as well. So none of this is actually help, but at least this link can be, it's off the BOINC wiki and gives the complete list of flags and options you can set for the different versions of BOINC (maybe there's something that can be done I clearly missed):

http://boinc.berkeley.edu/wiki/Client_configuration

I wanted the same thing for my single host setup as well.

It still gives you your cpu tasks, but it doesn't stop it from sending out the gpu request as well.

What is the actual message it is giving you in the Messages tab of Boinc Manager?

I wanted the same thing for my single host setup as well.

It still gives you your cpu tasks, but it doesn't stop it from sending out the gpu request as well.

What is the actual message it is giving you in the Messages tab of Boinc Manager?

Requesting new tasks for CPU and GPU.

It's a crapshoot though. I've seen it ask for just CPU half of the time, and the combo the other half, rarely if ever (not that I've seen so far) have I gotten caught in one of those GPU only request cycles when running GPU work before fetching new Rosetta work.

I wanted the same thing for my single host setup as well.

It still gives you your cpu tasks, but it doesn't stop it from sending out the gpu request as well.

What is the actual message it is giving you in the Messages tab of Boinc Manager?

Requesting new tasks for CPU and GPU.

It's a crapshoot though. I've seen it ask for just CPU half of the time, and the combo the other half, rarely if ever (not that I've seen so far) have I gotten caught in one of those GPU only request cycles when running GPU work before fetching new Rosetta work.

I don't know then, sorry!
I do know that ever since 25 Dec you have been doing units in almost 1/3 the time you used to:
388720965 355387081 25 Dec 2010 9:16:44 UTC Over 6,817.61 25.00 29.55
388690167 355358899 25 Dec 2010 9:16:44 UTC Over 14,396.00 52.79 56.67. You used to be in the 14000 second per unit range but are now in the 6800 second per unit range. And some of your latest units are down to 5400 seconds per unit!