Thanks :beer:Denithor:thumbsup: for this informative thread on the current GPU2 processing cards that are currently working on the F@H app.
I have been patiently awaiting Stamford's update of ATI's GPU2 core to take the plunge on those 4xxx cards that are causing havock to nVIDIA's bottom line since their month+ ago intro of those 280 & 260 cards.
My wife and i took a week off this past week - unfortunatly the cc's were using their siren voices and we (that is I) made some purchases.
.....some shorts & tops from Macy's during their 99cents only shipping charge
.....Sony XBR4 40" LCD for about $750.00 off MSRP w/no interest until Jan 2010 from Sears
.....assorted cables & a HDTV TERK antenna (don't have cable) - antenna pulled in over 30 stations over the standard 2-13 channels via the rabbit ears - Stations look really good vs the co-op's master antenna cable in our apartment. The antenna looks weird but this I consider the best purchase of the week AFTER the Sony (of course!). Very pleasantly suprized.
and, last but not least - an Evga 8800GT O/C vid card! Well, just could not wait for Stanford's ATI core upgrade and these cards are dropping in price like flies lately!
Dropped it into a Q6600 (OC to 2.8ghz) 8-gig 800 mem Vista64 OS Gigabyte 35 board housed in a Antec 900 midtower Wen. night (replaced a Gigabyte 3870 I got in March this year - don't worry, will place into a AM2 box that housing an ATI 1950 GPU1 Folding card which is no longer used since F@H has stopped sending out GPU1 workunits lately).
Well, the 8800GT didnt work with the GPU2 core until I d/l the CUDA files from nVIDIA on Thur. I also d/l'ed FAHMON to report the stats that other posters have listed on this Thread showing what their card outputs are. Unfortunatly, FAHMON keeps crashing with error #5 so I am going to have to guesstimate my ppd for this particular card.
Thursday was showing an estimate of about 3650ppd which rather perplexed me since your stats report a range of 4400 - 4800 for this card variant. Did some searches online to some o/c sites dealing with nVIDIA cards and d/l the RIVA toolkit that gives me stats on the performance of the card and o/c parameters (different from ATI's Control center but it does the job)
The graphs from RIVA was showing that the card was at the low end of it's Core clock, Shader clock and Memory clock so there was room for improvements.
I first increased the fan speed to 75%, which cooled the gpu from it's temperature of 72c (processing F@H) to the 50c zone. Didn't notice any increase in sound from this jump (hmm...maybe because of the ringing in my left ear that I have been subjected to for over a month and a half lately I can't hear it).
At this time each unit was done in 1min 30sec avg.
I raised the GPU core values from 650/1620/950 (C,S,M) to 741/1846/1025 - the first 2 values linked as I move one slider the other moves in concert.
This dropped the per unit completion time to 1min 18sec.
Looking at the Log of the GPU2 core it takes about 2hours, 12min (03 sec. from last completed work unit entry) from the start of folding to reporting back to Stanford to the starting point of the next segment.
Based on the above info and that the points for the work units completed are 480 my daily ppd is now about 5,208 :shocked: (just hope my math skills are better that my hearing!) :laugh:
The surprizing thing about this is that this PC box is also running 2 SMP's - One is completing a unit every 15min and the other 20min. This is 5 min more of a load than when I was NOT running the GPU2 from the ATI3870 card!!!
I tried raising the o/c on the nVIDIA card to 751 from 741 but instability occured and the PC crashed with the screen blanking out. Rebooted & set the o/c to the 741 values & has been running stable since.
My question on the O/C: Has anyone who o/c has the core & shader NOT linked - whereas only the core clock is o/c & the shader clock left alone. If so, did this result in a Higher core clock o/c and faster processing of the work units? Please let me know...
TurtleBlue