New GPU3/Fermi Projects P6800+P6811

[stoanee]

I seem to have an underachiever of a 460 here. I was using the advmethods flag which dropped the ppd to this:


Any thoughts?

[lowfat]

That still is a 912 point WU so you should be getting much higher than that. Maybe your GPU flaked out and is running @ stock clocks.

[stoanee]

A reboot seems to have solved it....

[Sagath]

Update with some information on these WU's: I'm pulling in 18kPPD with my brand new 580

Downside to the additional PPD? CPU usage has gone back up and is now sucking about 2-2.5% CPU time.

Also, I havnt seen an additional unit besides the 6800's since the first day these came out. I guess they pulled the other tester WU's? (72xx wernt they?).

[10e]

I see a constant 3% core usage by each GTX 4xx/fermi card with my X6 1055T.

I decided to just give 'em a core and go to -smp5 and use WinAFC to lock them to the last core.

I don't know about the 72xx units, but the new units require -advmethods flag to be used for now. That will likely change soon knowing them.

Originally Posted by Sagath

Update with some information on these WU's: I'm pulling in 18kPPD with my brand new 580

Downside to the additional PPD? CPU usage has gone back up and is now sucking about 2-2.5% CPU time.

Also, I havnt seen an additional unit besides the 6800's since the first day these came out. I guess they pulled the other tester WU's? (72xx wernt they?).

[sswilson]

If the 72XX units were the 611 pointers, I doubt you'll see them unless the coders come up with something else they want (need) to test. AFAIK, those units were test units for Fermi architecture that weren't actually doing any reportable science.

While I'm in here... somebody over on the FAH support forums posted this link with Mr. (or is it doctor?) Pande speaking about GPU computing.....

Folding@home: Petaflops on the Cheap Today; Exaflops Soon? - 2010 GPU Technology Conference, San Jose Convention Center San Jose, CA

[Soultribunal]

Originally Posted by sswilson

If the 72XX units were the 611 pointers, I doubt you'll see them unless the coders come up with something else they want (need) to test. AFAIK, those units were test units for Fermi architecture that weren't actually doing any reportable science.

While I'm in here... somebody over on the FAH support forums posted this link with Mr. (or is it doctor?) Pande speaking about GPU computing.....

Folding@home: Petaflops on the Cheap Today; Exaflops Soon? - 2010 GPU Technology Conference, San Jose Convention Center San Jose, CA

Awesome Link, it really helps us understand why we are doing what we are doing.
A lot of this I spoke with when I was working where i used to , as one of the Bio-CHermists at my work did the initial 'folding' in the 70's.

ST

[Dead Things]

Originally Posted by 10e

I decided to just give 'em a core and go to -smp5 and use WinAFC to lock them to the last core.

On an i7 860, I've got a GTX460, a GTS450 and a GTX260. With the heightened CPU cycle times of the Fermi cards, I tried going to SMP7 and locking the GPU clients to an unused core, and that improved things. But I find I get my best overall results by continuing to run SMP8, while locking each GPU client to a different core at -BelowNormal priority (in my case, CPU1, CPU3 and CPU5). I get a grand total of about 2000 extra PPD this way compared to SMP7 with the GPUs all on one core. Might be worth trying out to see if you get different results.

[sswilson]

Looks like the GTS450 honeymoon is officially over folks!!!!!!

Folding Forum • View topic - GPU3/Fermi P6800 on fah

6800 WUs are hitting the non -advmethods pool. Just pulled one myself.

Keep an eye on your EUEs for the next little while, I know I had to drop some of my OCs down to be able to run these.

[Xeven]

Oh, so that's why my PPD just tanked (@work can't confirm). I thought my twins crashed and burned. Back to reality I guess