GPU Processing

[tparin]

Are there any tweeks to enhance GPU processing other than OC'ing. Have GTX 570 pumping around 500-600K/day

[Crazybob]

Welcome aboard tparin! Nice card! Mainly you'll get the biggest improvement with OC'ing. There are some tweaks you can do depending on the project you're running. With Milkyway, you can add an app_info file to the projects folder so that MW will run 2-3 work units at a time. This works well for that project, but doesn't really do much for the others.

[Maxwell]

Since you're on PG with that card, there's nothing else I'd recommend other than OCing it.

CB is right - there are tweaks you can make with other projects, but PG is coded well and not one of them.

[DrPop]

Hi, welcome to the team! Sure wish I had a 570 about now. I think with PG, all you do is O/C that puppy to the hilt...their code is pretty optimized and the credits are sweet as candy!

[Crazybob's Son]

Welcome to the team tparin! Glad to see you here and also already crunching on the GPU.

I took a quick look at your cpid and noticed you only joined the team in primegrid. One unfortunate thing about running multiple projects is if you decide to join/switch teams you have to update each individual project.

Are you running anything else alongside your GPU crunching? If you aren't set on one particular project, right now primaboinca/AQUA give the best credits depending on your OS (as I recall one gets better with 32-bit OS and the other better with 64-bit OS?). Also, make sure to add WuProp to your machines as this runs alongside any projects you are crunching and is basically free credit on any/all machines

- Jason aka Crazybob's Son

[tparin]

Ran into Comp errors yesterday....anyone else see that? I can oc my sys any more. Old sys E6750 @ 3.4 fsb 1800 timing is at 4 4 4. Sandy bridge is in my future.

[Shadow]

What's your GPU temp running at?

[-GER-]

Are there any tweeks to enhance GPU processing other than OC'ing. Have GTX 570 pumping around 500-600K/day

Thanks to help from Maxwell, Bryan and PG's boards, I run this app_info on 570. It allows 2 wu's to run at once. A small but noticeable gain over 1 wu at a time.

<app_info>
<app>
<name>pps_sr2sieve</name>
<user_friendly_name>Proth Prime Search (Sieve)</user_friendly_name>
</app>
<file_info>
<name>primegrid_tpsieve_1.38_windows_intelx86__cud a23.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart.dll</name>
</file_info>
<app_version>
<app_name>pps_sr2sieve</app_name>
<version_num>138</version_num>
<plan_class>cuda23</plan_class>
<avg_ncpus>0.5</avg_ncpus>
<max_ncpus>1</max_ncpus>
<flops>1.0e11</flops>
<coproc>
<type>CUDA</type>
<count>.5</count>
</coproc>
<cmdline></cmdline>
<file_ref>
<file_name>primegrid_tpsieve_1.38_windows_intelx86 __cuda23.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
</app_version>
</app_info>