PrimeGrid

[Deleted member 88227]

Any special settings? My production seems low for the number of GPUs I have on this. 2x 1080 and 6x 980 Ti GPUs and I'm not even making 1M PPD. Something seems off.

 

[Gilthanis]

What does your utilization look like on those GPU's?

 

[Deleted member 88227]

99%+

 

[pututu]

A few days ago, I ran PPS Sieve (this is the best PPD project) with one 1080 at 55% power limit (~108W) and I'm getting about 1M PPD. BTW below is running two tasks per GPU.

I looked at some of your stats and you are also running other GPU projects (GFN, AP27) which has much lower PPD.

 

[MN Scout]

Skillz, It is all about the sub projects you run. For possible prime finding the highest PPD will be the Genefer projects -- specifically the highest number projects. If you want to sieve out possible primes for future people to crunch and see if they really are primes, then the PPS-Sieve project is the ultimate point producer, and you will want to run two tasks at once to fully load your GPU.

So max points, only enable PPS-Sieve gpu tasks and change to 2 tasks by adding this to the app_config and read the note below:
<app>
<name>pps_sr2sieve</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>

PPS-Sieve truly needs a full core to itself, if the program is trying to share a fully loaded CPU then your points will suffer greatly. I'd recommend watching the total task time with a full core load and without, so you know what the shortest possible time is.

Here is a screenshot of a spreadsheet I've shared before for my 1080TI:

Link to the primegrid spreadsheet that I made. It also has some run time numbers for CPU subprojects. It has gotten a little busy looking lately, but should give you an idea.

 

[MN Scout]

PrimeGrid has sieved really far for the PPS projects, and they won't get to the PPS LLR prime checking for each possible Proth Prime in this range for a handful of years from now. I'm considering changing one GPU to the Genefer 21 and 22 projects and the second GPU to manual sieving for the Genefer projects, since there isn't a real need to keep running the PPS-Sieve application right now.

 

[pututu]

You have the same idea as me of setting power limit on the GPU. I normally run at 55% power limit as it appears to give the best PPD/W. I just ran GFN22 and completed this task morning on my 1080 and it took about 58.2 hours @108W (GPU only, not system). So, one GFN22 task consumes about 6.3kWhr of electricity. If you don't set the power limit on GFN22, probably it will run at 100% and consumes roughly 200W and much more than 6.3kWhr.

 

[ChristianVirtual]

I run two GPU (980Ti, 1080ti) with PPS sieve and occasional GPUgrid with no tweaking setting and run 2.2M PPD; Linux

 

[Deleted member 88227]

Made some changes to my computing preferences. Gonna see if that does anything to my PPD.

 

[Deleted member 88227]

MN Scout I changed the app_config to reflect your config, but it's still running 1 task w. .27 CPU. I even tried to restart the client all together and still nothing. 1 GPU .27 CPU

 

[MN Scout]

MN Scout I changed the app_config to reflect your config, but it's still running 1 task w. .27 CPU. I even tried to restart the client all together and still nothing. 1 GPU .27 CPU

I wrote the below, but then looked at your computers. I wonder if you are running into a support issue because of running a really old client 7.2.47? 7.6.33 is the one I'm using. Just within the last week they released a new one, but there are complaints that it isn't stable. Maybe try and find a link to the 7.6.33 on the site. Edit:It looks like multiple GPU tasks per projects has been supported since 7.0.64, so maybe an old manager is not your issue...

You are running PPS-Sieve tasks then? That app_config entry won't affect other PG sub projects.

There are so many subprojects with confusing names, so I'll just include a photo. This is the one you want selected:

Restarting the Boinc Manager should have pulled in the changes.

Here is my app_config.xml that is in my C:\ProgramData\BOINC\projects\www.primegrid.com folder:

Spoiler

<app_config>
<app>
<name>llrGCW</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrGCW</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>pps_sr2sieve</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
<app>
<name>ap26</name>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>0.304</cpu_usage>
</gpu_versions>
</app>
<app>
<name>llr321</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llr321</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrTPS</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrTPS</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrWOO</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrWOO</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrCUL</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrCUL</app_name>
<cmdline>-t 2</cmdline>
<avg_ncpus>2</avg_ncpus>
</app_version>
<app>
<name>llrPSP</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrPSP</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrPPS</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrPPS</app_name>
<cmdline>-t 2</cmdline>
<avg_ncpus>2</avg_ncpus>
</app_version>
<app>
<name>llrSOB</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrSOB</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrTRP</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrTRP</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrPPSE</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrPPSE</app_name>
<cmdline>-t 1</cmdline>
<avg_ncpus>1</avg_ncpus>
</app_version>
<app>
<name>llrSR5</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrSR5</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrESP</name>
<fraction_done_exact/>
</app>
<app_version>
<app_name>llrESP</app_name>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>
<app>
<name>llrMEGA</name>
<fraction_done_exact/>
<max_concurrent>3</max_concurrent>
</app>
<app_version>
<app_name>llrMEGA</app_name>
<cmdline>-t 3</cmdline>
<avg_ncpus>3</avg_ncpus>
</app_version>
</app_config>

Try going Options-> read config files. I useally use that, but there is a bug in the 7.6,33 client I'm using that it will load up and run the second tasks, but it won't change the description on them, untill the manager is restarted.

 

[Deleted member 88227]

Oh that's what I was doing wrong. I was placing the app_config file in the same location as the cc_config file (the root above /projects).

Works great now. Thanks man.

 

[MN Scout]

Oh that's what I was doing wrong. I was placing the app_config file in the same location as the cc_config file (the root above /projects).

Works great now. Thanks man.

Awesome!

Should get a big boost in points now.

 

[phoenicis]

It could be my imagination, but driver version seem to make more of difference on PrimeGrid than other GPU projects. I noticed an improvement on the Ti gpus when I updated to 384.76.