nVidia production

SatTech

SatTech

2[H]4U
Joined
Dec 5, 2005
Messages
3,381
Anyone crunched any of these : 5230 (R4, C30, G8)'s? I've got a 280 GTX chewing on one now and FahMon shows only 2000 PPD with these bad boys. Perhaps one of the 'ATi' optimized ones? The other one is crunching a 5016 good for 7000ppd.

Any thoughts?



 
SatTech said:
Anyone crunched any of these : 5230 (R4, C30, G8)'s? I've got a 280 GTX chewing on one now and FahMon shows only 2000 PPD with these bad boys. Perhaps one of the 'ATi' optimized ones? The other one is crunching a 5016 good for 7000ppd.

Any thoughts?



Click to expand...

You always want to look here first:

http://fah-web.stanford.edu/psummary.html

As a rule that will give you an idea or at least a rough idea how that unit might fold.

I have so say I'm no fan of FAHMonitor other then it allows you to see that the machine (S) are actually working.

Your best bet is open the log file and look at the time lapse between frames.

Keep in mind if you have two different video cards in the same machine its possible that you are forcing one to run much slower. Also, if you over clocked the card make sure it's still over clocked. Some chipsets are not fond of running say an SMP client and a GPU client and supporting the over clock program at the same time.

Hope that at least points you in the right direction

Luck man;)



 
Yeah, thanks for the reply. Both are 280GTX's. That one was folding about 1/3 as fast as the other card running a different protein. I'll have to keep my eye out for more like it and see if it was just a funky anomoly. Both cards picked up decent proteins and are running at 7500 ppd each, so that's much better. It was just kind of suprising to see a 280GTX fold that slowly.

Stranger things have happened however..



 
SatTech said:
Yeah, thanks for the reply. Both are 280GTX's. That one was folding about 1/3 as fast as the other card running a different protein. I'll have to keep my eye out for more like it and see if it was just a funky anomoly. Both cards picked up decent proteins and are running at 7500 ppd each, so that's much better. It was just kind of suprising to see a 280GTX fold that slowly.

Stranger things have happened however..



Click to expand...

I should have also mentioned that the assigned points per WU are many times no indication at all of a given work units folding times.

This is and has been a huge argument many of us have been having with Stanford for more then a couple years now. Frankly, they don't want to hear about it.;)

 
Holy shite :eek:

Please someone tell me it's a dream that a nVIDIA GTX 280 GPU would only get 2000 ppd. and that I'm coming unglued over nothing. I understand about the different WU's and the different point structures (ie 480 vs 430), but even with a EVGA 8800gs Newegg "el cheapo" special I've had a ppd of at least 3500 points (with the "lowball" WU's). Has Stanford introduced a more time consuming, more complicated and less point rewarding WU :eek:

I need some help from the experienced folders of this team. Someone was nice enough to post a deal about the GTX 260's in a price range I can almost afford (I understand they're EVGA "B" stock, with a warranty of only 90 days), but now I'm very skeptable. Maybe I should just devote my small resources to WCG and admit to myself I've been outclassed by F@H.

I've never had any illusions of become much higher in the point structure than I already am, I was just trying to stay at about 10k pernts :rolleyes:)

Foldin' and WCGn' for a CURE

 
2000 ppd off a GTX 280 is a problem, I can get only 5200 ppd with the heaviest proteins but not that low. Make sure the fahmon is configured for Effective rate, not last 3 frames since that can vary wildly.

This is also possible if the core happen to be shared with another client by accident.

 
Well, I haven't seen any more of those WU, and both cards are ticking along at 7000+ ppd currently. It might have been some wierd windows problem also because only one GPU client (the 2000 PPD one) was running at that time. The other client was stuck trying to get work. Once it finally did (some 5 shutdowns/restarts of the client later) it ran about 5000 PPD on some unit. I ran out to dinner and came home to a crashed machine (windows crashed due to nVidia driver message). Everything has been running smoothly since...Dunno.



 
SatTech - I noticed this morning that my 8800 had crunched one of those overnight last night. I didnt notice any points drop from the "old" Nvidia WU we have been getting. This looks like it may have been a problem with your PC at that point in time. Just thought I'd let you know. :)
 
BillR said:
I should have also mentioned that the assigned points per WU are many times no indication at all of a given work units folding times.

This is and has been a huge argument many of us have been having with Stanford for more then a couple years now. Frankly, they don't want to hear about it.;)

Click to expand...

That's just one of the things I like about the WCG setup. Different projects have a quorum system in place with from 2-19 machines crunching the same WU for validation. If machine A takes 3 hours to process that unit and machine B takes 4.5 hours, B will get a higher points award for the extra time involved. But by the same token, machine A will generate more points/results overall if it is faster than B as it will process more results.

It just seems like a fairer system that rewards (if points can be considered a reward) for more time/effort/power to crunch a given amount of work.


 
You must log in or register to reply here.
Back
Top