As in this years "EVGA POTM August 2014 SETI@Home Wow!-Event 2014 - 15th August Thurogh 29th August 2014" we found the correct code needed for the app_config.xml settings.
 
TURN OFF SLI
 
SETI app_config.xml

<app_config>
   <app>
     <name>astropulse_v6</name>
     <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>0.3</cpu_usage>
     </gpu_versions>
   </app>
   <app>
     <name>setiathome_v7</name>
     <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>0.3</cpu_usage>
     </gpu_versions>
   </app>
</app_config>

 
Today I found the settings for MilkyWay app_config.xml File Settings.
Turned 3 Titans into running 6 GPU Projects running and from what I have seen today only 1 to 1.5 minutes added.
 

<app_config>
   <app>
     <name>milkyway</name>
     <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>0.3</cpu_usage>
     </gpu_versions>
   </app>
   <app>
     <name>milkyway_separation__modified_fit</name>
     <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>0.3</cpu_usage>
     </gpu_versions>
   </app>
</app_config>

 
Please add to this Thread for any other GPU Projects that you may have found the correct app_config.xml file settings.
For me the Hard Part was finding the correct app name like for MilkyWay it was milkyway and milkyway_separation__modified_fit note the two __ underscores in the second name. MilkyWay Projects can take up to 240 MB so you will need more memory if you have 3 or 4 cards installed.
 
Special Thanks to planetclown http://forums.evga.com/FindPost/2206961 for the SETI app_config.xml File Settings.
 
I am sure that their are other settings we can use to boost our Production.

For Collatz Conjecture note the <max_concurrent>6</max_concurrent> setting.
 
app_config.xml File Settings.
 

<app_config>
<app>
<name>solo_collatz</name>
<max_concurrent>6</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.5</cpu_usage>
</gpu_versions>
</app>
</app_config>

 

planetclown
This should work for any BOINC GPU applications, and you can find the correct app name in the client_state.xml file.  This is found in "program data\boinc" folder on windows, then search for <app> in the xml file after you've already started crunching at least one task for the application.
 
Keep in mind the 'optimal' GPU and CPU numbers will differ depending on the GPU card and application.

Thanks planetclown that will help a lot. To bad we cannot do this with Folding.

Tested this on my GTX 690 Rig.
Power usage went from 500-530 Watts to 710-740 Watts but the are running. Temps are 48C-52C
Cannot really use the computer due to display delays. I did change to 1 CPU from .5 CPU.

<app_config>
   <app>
     <name>acemdshort</name>
     <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>1.0</cpu_usage>
     </gpu_versions>
   </app>
</app_config>

Not really sure why it shows no CPU or Kernel Usage.

Here's my settings for primegrid, including all projects, what a pain that was! :P
I have 2 gpu's in this computer so max concurrent of 6 gives access to both gpu's and half my cores leaving room for other projects.
 

<app_config>
<project_max_concurrent>6</project_max_concurrent>
 
<app>
 <name>pps_sr2sieve</name>
 <max_concurrent>2</max_concurrent>
 <gpu_versions>
 <gpu_usage>1</gpu_usage>
 <cpu_usage>.25</cpu_usage>
 </gpu_versions>
 </app>
 
<app>
 <name>genefer</name>
 <max_concurrent>2</max_concurrent>
 <gpu_versions>
 <gpu_usage>1</gpu_usage>
 <cpu_usage>.5</cpu_usage>
 </gpu_versions>
 </app>
 
<app>
 <name>genefer_wr</name>
 <max_concurrent>2</max_concurrent>
 <gpu_versions>
 <gpu_usage>1</gpu_usage>
 <cpu_usage>.5</cpu_usage>
 </gpu_versions>
 </app>
 
<app>
 <name>llrMEGA</name>
 <max_concurrent>2</max_concurrent>
 </app>
 
<app>
 <name>llrCUL</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrWOO</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrPPS</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrSOB</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llr321</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>trp_sr2sieve</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrTRP</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrSR5</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrPPSE</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>llrTPS</name>
 <max_concurrent>1</max_concurrent>
 </app>
 
<app>
 <name>psp_sr2sieve</name>
 <max_concurrent>1</max_concurrent>
 </app>
</app_config>

GPUGrid
 
Not really sure how relevant <cpu_usage> is for running GPU tasks. 0.125 to 0.5 seems to have little to no effect on the amount of work my computer crunches, so I just have it set to 0.125 for now to leave enough CPU threads available for other tasks and programs I use.
 
Presently, I'm running 2 GPUGrid tasks per GPU, of which I have 2, so 4 GPU tasks total, and 8 CPU tasks for World Community Grid. In the general UI BOINC settings, I have it set to use 67% (two thirds) of my CPU's threads, so that's 8 of 12 threads, 8 tasks. So to clarify, that's 8 CPU World Community Grid tasks and 4 GPU GPUGrid tasks I'm running simultaneously.
 
I have SLI enabled and prefer maximum performance in the Nvidia control panel. (SLI is enabled for some video games, like Watch Dogs and Battlefield 4.) I also have the PCIe gen. 3 force enabled with the Nvidia patch for the X79 systems, so each of my SC Titan Blacks are on 16 gen. 3 PCIe lanes.
 
With SLI being enabled, VRAM usage is mirrored between my cards. So in memory usage, that's 4 tasks per card. Even so, my BOINC VRAM memory load is only about 30% 40%. (It's nice having 6GB VRAM per card.) GPU load is about 80% to 90% on average, which is up about 20% from only running one GPUGrid task per GPU. Because of this and the lack of down time between tasks, I expect a significant increase in the total amount of work per time my computer will do.
 
If anyone has more insight about what <cpu_usage> should be set to and why, I would appreciate hearing it.
 
Generally speaking, and provided your video cards can handle it, the more GPU tasks you can run, the better. So far as I can tell, the GPUGrid project is included.
 
C:\ProgramData\BOINC\cc_config.xml (general BOINC setup for running multiple GPU tasks, one per GPU):
 

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

 
C:\ProgramData\BOINC\projects\www.gpugrid.net\app_config.xml (with the various GPUGrid task names - also, with a <max_concurrent> of 0, the number of tasks aren't limited here):
 

<app_config>
<!-- Short runs (2-3 hours on fastest card) -->
<app>
<name>acemdshort</name>
<max_concurrent>0</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.125</cpu_usage>
</gpu_versions>
</app>
<!-- Long runs (8-12 hours on fastest card) -->
<app>
<name>acemdlong</name>
<max_concurrent>0</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.125</cpu_usage>
</gpu_versions>
</app>
<!-- ACEMD beta version -->
<app>
<name>acemdbeta</name>
<max_concurrent>0</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.125</cpu_usage>
</gpu_versions>
</app>
</app_config>

bcavnaugh
When <max_concurrent>0</max_concurrent> is set to 0 is this the same as not have this line in in the config file?
 

I think so (not positive though). In effect it's the same as setting it to 9999, or what have you. Basically, so the number of BOINC CPU and GPU tasks that can be run simultaneously aren't limited by this value. I'm using this 0 value as a setting, but also as a place holder for this setting so I can easily change it later, if I want to.
 
Recently here I've set the general BOINC CPU usage setting to 100% and the <cpu_usage> value in the app_config.xml file for GPUGrid to 1.0. Since I'm running 4 GPU tasks, this reserves 4 CPU threads for them from BOINC (leaving the threads mostly open for other programs to use as needed), and so the remanding 8 of 12 threads are used for CPU tasks.

Bumping this post to help anyone for the PPS Sieve challenge.
I just added a Fermi card to a Kepler rig and so far the project defaults to the Kepler card. I'll post back with specific questions if I can't get it set up.

Orange_1050

yodap
Bumping this post to help anyone for the PPS Sieve challenge.
I just added a Fermi card to a Kepler rig and so far the project defaults to the Kepler card. I'll post back with specific questions if I can't get it set up.

If the problem is that BOINC use one GPU only, this cc_config should do the trick.

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
  </options>
< /cc_config>

And to add to what Orange_1050 said, BOINC by default will try to use what it deems as the better of the two cards. So, if you want all cards to run, use that cc_config flag he posted. If you actually only want one card to run, then you will want to add an exclude GPU flag instead.

yodap, I am referring to a cc_config.xml. You could do it on an app_config.xml level but then that would only apply to the apps you list in it. By doing it via cc_config you are applying it to all projects.
 
http://hardforum.com/showpost.php?p=1039993069&postcount=8
 

yodap
Thanks guys. I can't seem to make it happen. Looks like the 560ti will be watching from the sidelines. Good news is I'll have 6 other cards working it.

You could use the 560Ti as your Main GPU that way all the other GPU's would only run Projects or Tasks.

Thanks guys. Worked like a charm
Always wondered how to get the other gpu to crunch.

Thanks for the always useful knowledge bcav. 

Testing Now, the default is 0.2 CPU and it run one task on all your GPUs so it run 0.2CPUx2GPU 1 Task.
 
This route takes about 25 minuets
moowrap.net
<app_config>
<app>
<name>dnetc</name>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
Or
<app_config>
<app>
<name>dnetc</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config> 
OR
 
(Retesting) This route takes about 46 minuets
<app_config>
<app>
<name>dnetc</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.5</cpu_usage>
</gpu_versions>
</app>
</app_config> 
 
This route takes about 15 minuets (Retesting)
moowrap.net
<app_config>
<app>
<name>dnetc</name>
<gpu_versions>
<gpu_usage>2.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
 
 
This route takes about 15 minuets
moowrap.net
<app_config>
<app>
<name>dnetc</name>
<gpu_versions>
<gpu_usage>2.0</gpu_usage>
<cpu_usage>2.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
 
dnetc_r72_1461261822_12_768_0 15 Min
dnetc_r72_1461256203_12_768_0 25 Min
dnetc_r72_1461267146_12_768_0 46 Min

bcavnaugh

robbysites

yodap
Thanks for the always useful knowledge bcav. 

Awesome work Bcavnaugh!!
I run 2 900 cards on GPUGRID, but my 780s will only run 1. Any thoughts, ideas help??
Thanks man!
robbysites

I thought you had this fixed!
     <max_concurrent>Put the Number of Cards Here</max_concurrent>
 
<app_config>
  <app>
    <name>acemdlong</name>
     <max_concurrent>2</max_concurrent>
    <gpu_versions>
        <gpu_usage>1.0</gpu_usage>
        <cpu_usage>1.0</cpu_usage>
    </gpu_versions>
  </app>
</app_config>

I did on my rig with 2 900s, not my rig with 2 780s. They run 2 WUs folding, but only 1 on GPUGRID.
rs