Wouldn't mind upgrading my ram from DDR3-1600 to DDR3-2000: http://www.newegg.com/Product/Product.aspx?Item=N82E16820145329 I wouldn't do it just from the folding aspect... the business aspect would be to permit a greater number of virtual machines to run, each with a decent amount of dedicated memory allocated to them... I like that I could add an additional 12 GB later on whereas I'm currently maxed out with six 2-GB modules. I've read here that increasing memory speed can have a big increase on SMP2 folding performance. Does anyone have some numbers for a quad-core Core i7? <addition> According to reviews I read, my current ram is capped internally at 1600 which explains why my previous attempts to exceed 1600 failed.
all i can say is when 970 was at stock i change the memory from 1600 to 2000 and it knock off 1-2 min off a bigadv project as look for some low timming ram as well it all helps i didn't take down number tho just what i saw
I did some experimentation (i7-860) with DDR3-1333 MHz to DDR3-1600 and found that the significant difference (~1.5 minute) was with bigadv WUs where they use ~1GB RAM. For normal WUs, it was more along the lines of a few seconds. Reducing the RAM timings did shave a few seconds but was time consuming so if you have plenty of free time to kill, you can tweak the timings :nerd:
I assume we are talking about shaving time off the TPF... If I assume a direct relationship betwen TPF and PPD then my calculations suggest I would see an increase of about 1500 PPD. I don't need the additional memory at this moment, but this is good to know. Thanks! (if the estimated PPD increase was large I might have jumped) I'm focusing on acquiring two 580 HCs. That should increase our PPD by 31,500 (subtracted 4500 for SMP7 to SMP6 fallback)! I did as iancook suggested and ordered an auxillary 5.25" bay PSU. It will take some pressure off my 1200-watt PSU and permit me to overclock all 3 GPUs. The next unknown will be in reference to how much cooling the water loop for my GPUs can truly provide.
Trying to keep me from mowing you over? Your about too! On Topic: My testing has confirmed PantherX's findings as well. About 1.5-2 minute drop in TPF but only with BigAdvanced. No real noticable change with Regular A3 units.
Thanks for the numbers everyone! Ghost15: Nah - I think your PPD rate exceeds 80,000. When are you going to be back up? The team needs you!
Just started everything back up yesterday, New card is installed (and boy is it louder!) BigAdv from the 980x drops tomarrow AM (EST). The GPU3 client is being really fickled, I might have to start a thread about it. Log shows nothing happening but the Tooltip shows it processing fine.
Watch out - your GPU3 might be erroring on working units and dumping (returning them) every few seconds. View the client's Work folder. If you see the folder being cleared out and and new files being brought in every couple of seconds then I would suggest the following: - stop the client - clear our the Work folder (delete everything in there) - delete the FAHlog.txt, FAHlog-Prev.txt, and unitinfo.txt file - reboot the system - start the GPU3 and monitor the Work folder again I've only had to do this twice with the 480. I believe each time was after my GPU3 client errored out due to an unstable overclock of the GPU. I remember having to do it a few more times back when I was trying to get all 4 GPUs in my two 295s to fold when I first joined the team and began folding in early 2009. <Addition> Another thought... perhaps use the flag to force Fermi? Perhaps its stuck thinking you still have a pre-Fermi card??
Looks like a reboot did it, it was wierd, it was like it was dumping the log all at once, as it had the same timestamp for each % completed. After the reboot it seems to be writting to the log as expected. The "Viewer" showed it working correctly, so I wasn't concerned that the WU's were failing
