devlin85

bcavnaugh
 
To Run Long Tasks Only :
 
Location: C:\ProgramData\BOINC\projects\www.gpugrid.net
 
File Name: app_config.xml
 
<app_config>
   <app>
   <name>acemdlong</name>
   <max_concurrent>0</max_concurrent>
   <gpu_versions>
       <gpu_usage>1.0</gpu_usage>
       <cpu_usage>1.0</cpu_usage>
   </gpu_versions>
</app>
</app_config>
 

does that work for you? everytime I try to restrict a project like that ( with 0 max concurrent) it just completely ignores it. 
 
GPUGrid's website lets you pick which projects though. Wish I could get it to work, I would like to run specific tasks on certain GPU's..

Sorry what Orange_1050 Said.
Each Project as a settings page like the one Orange_1050 shows in his post.

That's what I thought, there really needs to be a way to do it though.. without global settings, Need some way to restrict on a machine basis. I.E. In GPUGrid I would like my 780 to get all long tasks, but my 760's to get all short..
 
And that damn High Priority Panic Mode Nonsense! :P

I use BAM too, but even when setting up different work preferences there's no way to nail down which sub-projects to which pc's. Say I wanna run prime grid on all my gpus, but each one does better with another. Like I don't want my 650ti doing world record genefer (at 30 days) when my 780 will do it in 3. But as posted b4 the 650ti is a killer at PPS Sieve chomping em up at 15-16 minutes. (760 will do one at 10:15 on the nose everytime, and the 780 only 9 minutes and some change) so definately not worth running on the 780, it just goes on and on.

If I'm not mistaken there is no way to do this. I used to be able to change the settings on the site, hit update on the machine, then do the same for the others and so on, but now the clients are auto updating themselves and just taking the newest settings constantly.  

Khalifrio

yodap

Khalifrio
If your running a older and/or slower GPU then I recommend you stick to short tasks. By older/slower I am referring to anything under a GTX 670 or a GTX 770.
 
For long runs you need to finish them within 24 hours to get a bonus point reward. So if you have a 670, 680, 690, 770, 780, 780 Ti, Titan, 970, or a 980 then go for the long runs. Other cards might be able to make the 24 hour dead line but I know the ones listed will do so.
 
I will be joining in with my 680 and 780 Ti. My CPU cores will be crunching away on SIMAP of course.

This weekend I'll test one on a 660ti, since it 's right on or just below your Mendoza line.

The 660 Ti is the one that is right on the line. It all depends on the specific 660 Ti. Some do better than others. If your lucky and got one that is a tad faster than most then you should be OK.

I've started the great 660ti experiment. I have 3 such cards and the one I believe to be my best card picked up a short-runs before I could change my settings. The other 2 are running great on long-runs. Looks to be 9-10 hours each after 15% complete. I'm not sure if I'm doing the correct wu though because the deadline is like 5 days not 1.
I'll keep you posted.

I will test my GTX 660Ti today as well, it is installed on my 2P Rig.   FAILED

Orange_1050
I'm happy to see you all prepare to this event.
Curious on how we will do.
Big names on the participant list.
Let's make it a 5 day Blitzkrieg

Indeed a "Lightning War"

My GTX 660Ti today as well, it is installed on my 2P Rig Failed to run Outcome Computation error and Client state Compute error.
http://www.gpugrid.net/result.php?resultid=13308999
http://www.gpugrid.net/workunit.php?wuid=10242255
 
name unfoldx337-NOELIA_UNFOLD-14-72-RND4878
application Long runs (8-12 hours on fastest card)
created 1 Nov 2014 | 17:10:10 UTC
I blame this on Ubuntu 14.04 Desktop and not the GTX 660Ti. Over 25 failed today.
I never did get a single GPUGRID to work, I did forget that I have tested this before.
 
<core_client_version>7.2.42</core_client_version>
<![CDATA[
<message>
process exited with code 199 (0xc7, -57)
</message>
<stderr_txt>
SWAN : FATAL : Cuda driver error 3 in file 'swanlibnv2.cpp' in line 446.
# SWAN swan_assert -57

</stderr_txt>
]]>
 
FYI  Collatz Conjecture and MilkyWay@home run fine on my 4P and 2P. 650Ti and 660Ti

So then, I noticed with the default config there is this cpu based task "Test app for CPU md blah blah". Exactly what are these and do they contribute to the contest?
  

devlin85
I use BAM too, but even when setting up different work preferences there's no way to nail down which sub-projects to which pc's. Say I wanna run prime grid on all my gpus, but each one does better with another. Like I don't want my 650ti doing world record genefer (at 30 days) when my 780 will do it in 3. But as posted b4 the 650ti is a killer at PPS Sieve chomping em up at 15-16 minutes. (760 will do one at 10:15 on the nose everytime, and the 780 only 9 minutes and some change) so definately not worth running on the 780, it just goes on and on.

If I'm not mistaken there is no way to do this. I used to be able to change the settings on the site, hit update on the machine, then do the same for the others and so on, but now the clients are auto updating themselves and just taking the newest settings constantly.  

Here is a snippit from one of my cc_config files. The system has a 650Ti and 3x 8400 GS video cards in it. So, since the 8400 GS is not capable of the dual precision work, I clearly don't want them running certain apps. However I do want them running others within the project. So, what I had to do is the following:
 
<cc_config>
<options>
<max_tasks_reported>25</max_tasks_reported>
<use_all_gpus>1</use_all_gpus>
<report_results_immediately>1</report_results_immediately>
<http_1_0>1</http_1_0>
<exclude_gpu>
<url>http://www.gpugrid.net/</url>
<device_num>1</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>1</device_num>
<type>NVIDIA</type>
<app>genefer</app>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>1</device_num>
<type>NVIDIA</type>
<app>genefer_wr</app>
</exclude_gpu>
<exclude_gpu>
<url>http://milkyway.cs.rpi.edu/milkyway/</url>
<device_num>1</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://www.gpugrid.net/</url>
<device_num>2</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>2</device_num>
<type>NVIDIA</type>
<app>genefer</app>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>2</device_num>
<type>NVIDIA</type>
<app>genefer_wr</app>
</exclude_gpu>
<exclude_gpu>
<url>http://milkyway.cs.rpi.edu/milkyway/</url>
<device_num>2</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://www.gpugrid.net/</url>
<device_num>3</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>3</device_num>
<type>NVIDIA</type>
<app>genefer</app>
</exclude_gpu>
<exclude_gpu>
<url>http://www.primegrid.com/</url>
<device_num>3</device_num>
<type>NVIDIA</type>
<app>genefer_wr</app>
</exclude_gpu>
<exclude_gpu>
<url>http://milkyway.cs.rpi.edu/milkyway/</url>
<device_num>3</device_num>
</exclude_gpu>
</options>
</cc_config>
 
So to answer your question, yes  you can exclude running specific apps on specific cards. Notice the PrimeGrid exclusion where I tell the cards not to run the Genefer work units. I have both types in the file. And I had to exclude them from GPUGrid altogether because the 650Ti will pull work and then BOINC will try to run it on any available card. They will fail on the 8400 GS cards so, that is a bad thing.

bcavnaugh

dave851
So then, I noticed with the default config there is this cpu based task "Test app for CPU md blah blah". Exactly what are these and do they contribute to the contest?

Can't say for sure but then that is part of the learning curve and it will take awhile to learn each BOINC Project.

Im interested in how those CPU WU's will credit, especially now just before the Cruncathlon.
I'm most certain it will Count in the contest.
Didn't get any of those yet, but I'll keep an eye on it and report back.
 
 
 
 
Edit: Yay, lookie lookie
12cpu's

After Testing this week this is what I will be running:
Short Run Tasks: 3 GTX 690
Long Run Tasks: 3 GTX 680 & 3 GTX 780
Day Runs only Short Run Tasks: 2 GTX Titans.

devlin85
I have concluded the cpu tasks based on credit to be a waste of time.. I have run through 4 of the "Test application for CPU MD", using 8 cores for 1 process (i7-4820k @ 3.7-3.9Ghz) it took just over ten hours for 500,000 Gflops to process.. netting a less than measly 3k in points. Sticking with the GPU!

Why would you say they aren't worth running? Do you have a more important project to run on your CPU's? Just curious is all. Also, keep in mind that CPU work isn't always the same science research as the GPU app. So, if your goal is to push the teams stance on this project, why wouldn't you squeeze the extra 3k here? Or do you justify it earning more points elsewhere. Again, just curious.

Well...I figured you would say POGS...kinda a sore point for me. I feel they are wasting resources to begin with and dragging their feet terribly. I like the project for the most part, but they really need to move things along. To me, adding more work there is less smart than GPUGrid. 1.) Because they (POGS) should have a CPU optimized app long before now that is crazy better than the one they are using. 2.) They are also working on a GPU app that should have received a lot more attention by now as well. However, these go back to before April and just haven't gotten but a fraction of the attention that it should have by now. However, they may have me help test their original project they are porting to BOINC, so there is some kind of progress happening....

SourceFinder will more than likely be CPU only. At least at first.

27 hours? yikes! was it still just 3k credit 500,000 gflops project or was this some massive wu due to the large amount of cores?