Is there any why to tell the BOINC Manager to use one GPU for SETI and the other for GPUGRID?
 
OK, so here is my setup...
 
X58 with a 920
Two 680's
All watercooled.
 
I am running BOINC manager.
 
I have WCG and SETI and GPUGRID all running on this ones system and allowing 100% usage 24/7.
 
What I noticed today, if I have any GPIGRID running, it stops SETI. All the GPUGRID's "Claim" to be high priorities.
 
From where I sit these 680's eat anything and everything that either project can toss at it. I slam through SETI's in 4-7 minutes and so far the two GPUGRID's I have done are complete in just over (4) hours each... This makes me want to run both!

I haven't found anything, I haven't really looked into it that much as I'm sure it would result in more work than I was willing to put into it for the time I was gonna use it.  I wanted to have PrimeGrid and GPUGrid use my 460 and not my 8800 and let everything else use only the 8800.  I think there is a way through editing the project files to only use a specific gpu, but I would have had to edit it for every gpu-capable project I crunch and I wasn't going to keep the 460 adn the 8800 in the same system for long enough to make it worth it.  linux might have more knowledge on this than me....
 
Edit:  Or you could "No more work" every project but GPUGrid and set the work buffer for a few days and update then no more work it and allow more work everything else after changing the buffer again and letting the manager do it's thing....  Or maybe lower the amount for GPUGrid so it gives Seti a higher priority, eventually you'll get GPUGrid tasks as it has to balance out project work, but by giving Seti more resources and work time, it should ensure more Seti tasks that don't get bumped.  Or...I haven't looked too much into batch files, but I'm sure you could have a batch file manage suspending and resuming the two on a schedule to "manually" balance the workload.

Simba123

Hmmm.  perhaps perhaps  ........

try this in your cc_config file.  I can't guarantee that this will work, but it might.  Remember to open your cc_config file in wordpad (not notepad) and make sure to save it as .XML .  make sure to save your original file somewhere as well in case you have to go back to it.


add this to your existing file.  If I've done it right, it 'may' exclude the first GPU from getting GPUGrid tasks. See how you go.


<options>

<exclude_gpu>
<url>http://www.gpugrid.net/</url>
<device_num>0</device_num>
<type>nvidia</type>
<app>acemd</app>
<app>acemdlong</app>
</exclude_gpu>

</options>

 
You know, I'm pretty sure I saw that somewhere....  I thought it was more an app_info thing than a cc_config.  But maybe it was because I was looking more for PrimeGrid and specifically excluding the 8800 from the Genefer tasks as it can't run those....
 
Edit:  http://boinc.berkeley.edu/dev/forum_thread.php?id=7486  if I'm understanding this right, that's correct and I don't know why I couldn't find this before.  My PrimeGrid and Genefer scores would be much higher....

I am not getting the option to use wordpad, only notepad is popping up... weird... 
 
This is how it looks in the "Open" feature to explorer and it is saved this way...
 

I'd go with Option only once.... And unless they changed it with the newest version, and even if they have or you're running an older one, it wouldn't hurt to suspend everything, exit boinc, start it back up and have it read the cc-config file to ensure everything is set....  But, yeah, it might be good to wait for the current gpu tasks to end incase the settings cause it to toss the unit out instead of finishing it.  I don't think a gpu task changes cards if it is suspended and resumed, and if it can, I don't think it likes it too much....  Or it could be they don't like going from a 460 to an 8800....

OK... Did not work. Started new job on "0"... Shutting down, editing to remove the option  tags in the middle...
 
*Update*
 
Fired it all back up, BOINC Manger did not abort any of the jobs (Thank goodness), still running as if I did nothing.
 
I will let the jobs run through their course. So it will be a few hours before another update...
 
 

OK... I keep trying.. Even took out the "/" at the end of GPUGRID.net... And get this message (Still running BOINC 6.10)
 

 
 
 
 
**Edit**
 
I found this info on boinc.berkley about half way down the page...
 
http://boinc.berkeley.edu/wiki/Client_configuration
 
Is that saying this GPU script is for 6.13 and up? Or that this works for any boinc, just happened to be added with 6.13?

Simba123

As I said, no promises, it was midnight when I posted it and I was tired and that was something I had a brainwave about and quickly threw together.
Give me a bit and I'll see if can refine/get it working.

to open the file in wordpad, right-click, select 'open with', then choose wordpad.  (notepad doesn't save xml files properly, that's why you need to use wordpad, which does)

Your assist was and is fantastic. This was my issue...
 
You see... I am normally a set it and forget it fella. But these 680's and all the other GPU production we can do got my blood flowing again. I have not updated my BOINC in over two years it looks like. 
 
Good news!!! Updating "from" WCG's 6.10 to BOINC's newer 7.0.28 allowed the script you wrote, less the options text... 
 
Now the bad news...
 
All the work that was completed, and in process deleted when I upgraded. SO I deleted all of the BOINC and started fresh just now.
 
P.S. I was using the right click, failed to use the open with . 
 
**Update**
 
This script is recognized... Only now it is reading the second "App" line and shutting off "0" to all. This means SETI is still sitting doing nothing as GPUGRID takes over on the device #1... Sad panda
 
Deleting script and running both GPU's on GPUGRID for now...

Nope.. Looks like this (Without added GPU script)
 
[link=file:///C:/ProgramData/BOINC/cc_config.xml#]-[/link] <cc_config>

  <log_flags />

[link=file:///C:/ProgramData/BOINC/cc_config.xml#]-[/link] <options>
  <client_version_check_url>http://www.worldcommunity..._version_check_url>

  <client_download_url>http://www.worldcommunity...lient_download_url>

  <network_test_url>http://www.ibm.com/</network_test_url>

  <start_delay>120</start_delay>

  </options>


  </cc_config>

Well... I was all excited... Did "Exactly" what we want... Finished two SETI's while running one GPUGRID and bam... Second GPUDRID took over and SETI has stopped again...
 
That said... Not to get all excited to quickly... But it just did two SETI's in just over 2 minutes each! AND both GPUGRIDs are showing 2-3 hours to completion. "Crosses fingers"
 
BOINC Text 7.0

Correct. And GPUGRID was only on "1".. Once SETI completed the second job "0" became the active as well for GPUGRID...
 
And not to worry man. I am learning as I go. The spacing seems to be something I should know 

On it 

also, go to your GPUGrid profile at GPUGrid.com and make sure the short tasks are NOT selected.
 
 

Simba123

also, go to your GPUGrid profile at GPUGrid.com and make sure the short tasks are NOT selected.

I looked and looked. I do not see anything called profile or anything I clicked saying short anything. 

^^^ Ah.. That I have. All three are selected by default, and it looks like I ran one... Oh well =]... No longer selected thank you!!