I've had to limit my 3770k to two cores for folding. My temps were in the 80s. We'll see how much this will affect my PPD output, though I'm thinking it may only be about 10k. I may have to reconfigure some fans or check my thermal paste as well.
Additionally I would like to thank all of the new folders and welcome them to the team. Your efforts are greatly appreciated!
@Tushon said:
airflow, dust, thermal paste are all good things to check.
I just recently filled out my Corsair 750d with 140mm fans. As of now I have the two front fans as intake and the rear and top as exhaust. I'm thinking that the three top fans are pulling the cold air out of the case closer to the front before it even gets to HSF. I think flipping the rear fan and making an intake should help.
That's just my home PC. The GPUs are doing the heavy lifting - a pair of R9 Fury 4GBs each doing somewhere between 300-600k/day if they're stable. CPU's a 4790K doing about 10k/day.
@Snarkasm said:
That's just my home PC. The GPUs are doing the heavy lifting - a pair of R9 Fury 4GBs each doing somewhere between 300-600k/day if they're stable. CPU's a 4790K doing about 10k/day.
You could have a downclock profile if they stay stable to use while folding, and revert to normal when gaming.
@Snarkasm said:
That's just my home PC. The GPUs are doing the heavy lifting - a pair of R9 Fury 4GBs each doing somewhere between 300-600k/day if they're stable. CPU's a 4790K doing about 10k/day.
You could have a downclock profile if they stay stable to use while folding, and revert to normal when gaming.
Yeah, I'm trying to determine what the trigger is, but maybe a 10% downclock is just better. But if I can get it stable at 100%, I like the big numbers
Nobody asked me, but I'm going to tell you anyway
I currently have 2 mostly dedicated systems doing CPU with 2x GTX970 GPUs each; The 970 gets just over 250k PPD pretty consistently
(In all the excitement there might be some more hardware on the way....)
yeah, it's ends up being pretty crazy with modern GPUs (also still crazy that we have multiple teraflops of single precision math capacity in such a small add-in card)
Presumably though they're boosting the points so much because most of the value for them is in being able to trace through a simulation quickly- each of these WU's is just a handful of nanoseconds of simulated time, and the insights come from having a much longer thread of simulation.
Their software is actually really neat and novel, it automatically figures out some details about subsets of the folding solution from all the individual simulations and builds new work to test out those assumptions and connect the individual pieces that have been discovered in parallel, that's how they can simulate much longer folding proteins without actually simulating it end-to-end.
(data point: one of my GPUs would take 88 years to simulate a millisecond of time in the WU it's currently running. Molecular dynamics is complicated.)
Thanks! Gave it a run overnight last night, pulled some decent numbers. I can tell when I have the folder going in the background, so I only run it in the time I normally had the machine off. So long as I get into the habit of running the sucker. # 861 DocFrazier 25374
Are the FAHClient/Control PPD estimates known to be inaccurate? I have a GTX 670 folding 24/7 and a GTX 770 folding at night. folding.extremeoverclocking is saying I'm getting around 13k a day, which is a lot lower than what the client estimates are giving me depending on the work unit.
@Gattsu said:
Are the FAHClient/Control PPD estimates known to be inaccurate? I have a GTX 670 folding 24/7 and a GTX 770 folding at night. folding.extremeoverclocking is saying I'm getting around 13k a day, which is a lot lower than what the client estimates are giving me depending on the work unit.
Did you add a passkey to your client? There is also some documentation about leaving a CPU slot open to make F@H more efficient with OpenCL. I'm not well versed in what your GTX cards use for folding if left in a default configuration though, and I would assume that they would take advantage of their CUDA technologies, but I don't know if you have configured anything special for your setup.
@Gattsu said:
Are the FAHClient/Control PPD estimates known to be inaccurate? I have a GTX 670 folding 24/7 and a GTX 770 folding at night. folding.extremeoverclocking is saying I'm getting around 13k a day, which is a lot lower than what the client estimates are giving me depending on the work unit.
Did you add a passkey to your client? There is also some documentation about leaving a CPU slot open to make F@H more efficient with OpenCL. I'm not well versed in what your GTX cards use for folding if left in a default configuration though, and I would assume that they would take advantage of their CUDA technologies, but I don't know if you have configured anything special for your setup.
My F@H client did this automatically (re the CPU slot staying open), so hopefully they did for him as well. Passkey is good to look for as you indicated.
Comments
Jibbers Crabst.... reddit.com/r/PCMasterRace is crushing it. Gaining on us by 11M ppd. o_O
Yeah we cannot hang with that level of sheer volume of community members.
They just started 11 days ago! Go congratulate their new team here
I've had to limit my 3770k to two cores for folding. My temps were in the 80s. We'll see how much this will affect my PPD output, though I'm thinking it may only be about 10k. I may have to reconfigure some fans or check my thermal paste as well.
Additionally I would like to thank all of the new folders and welcome them to the team. Your efforts are greatly appreciated!
airflow, dust, thermal paste are all good things to check.
I just recently filled out my Corsair 750d with 140mm fans. As of now I have the two front fans as intake and the rear and top as exhaust. I'm thinking that the three top fans are pulling the cold air out of the case closer to the front before it even gets to HSF. I think flipping the rear fan and making an intake should help.
Dang @Snarkasm!
Also: We're up to 30 active now
Yeah, except it's intermittently stable. It appears to have crashed overnight and I didn't get to check it this morning. 0 PPD day sad face
For real! What are you folding with?
That's just my home PC. The GPUs are doing the heavy lifting - a pair of R9 Fury 4GBs each doing somewhere between 300-600k/day if they're stable. CPU's a 4790K doing about 10k/day.
fuck youuuuuuu
You could have a downclock profile if they stay stable to use while folding, and revert to normal when gaming.
Yeah, I'm trying to determine what the trigger is, but maybe a 10% downclock is just better. But if I can get it stable at 100%, I like the big numbers
Nobody asked me, but I'm going to tell you anyway
I currently have 2 mostly dedicated systems doing CPU with 2x GTX970 GPUs each; The 970 gets just over 250k PPD pretty consistently
(In all the excitement there might be some more hardware on the way....)
Only two rigs @sgstair? Damn dude I assumed you had a farm
That's what I mean - GPU points are nuts. If you have two Crossfire or SLI systems, you're nearing a mil pretty easy.
yeah, it's ends up being pretty crazy with modern GPUs (also still crazy that we have multiple teraflops of single precision math capacity in such a small add-in card)
Presumably though they're boosting the points so much because most of the value for them is in being able to trace through a simulation quickly- each of these WU's is just a handful of nanoseconds of simulated time, and the insights come from having a much longer thread of simulation.
Their software is actually really neat and novel, it automatically figures out some details about subsets of the folding solution from all the individual simulations and builds new work to test out those assumptions and connect the individual pieces that have been discovered in parallel, that's how they can simulate much longer folding proteins without actually simulating it end-to-end.
(data point: one of my GPUs would take 88 years to simulate a millisecond of time in the WU it's currently running. Molecular dynamics is complicated.)
Welcome @docfrazier!
Thanks! Gave it a run overnight last night, pulled some decent numbers. I can tell when I have the folder going in the background, so I only run it in the time I normally had the machine off. So long as I get into the habit of running the sucker.
# 861 DocFrazier 25374
I'm back too, though I fell out of the top 10 since this thread started.
Welcome back @sliquid and welcome new folder @Necroldia!
Welcome back @Ghoosdum and @cb!
Welcome back @GnomeQueen!
Are the FAHClient/Control PPD estimates known to be inaccurate? I have a GTX 670 folding 24/7 and a GTX 770 folding at night. folding.extremeoverclocking is saying I'm getting around 13k a day, which is a lot lower than what the client estimates are giving me depending on the work unit.
Did you add a passkey to your client? There is also some documentation about leaving a CPU slot open to make F@H more efficient with OpenCL. I'm not well versed in what your GTX cards use for folding if left in a default configuration though, and I would assume that they would take advantage of their CUDA technologies, but I don't know if you have configured anything special for your setup.
My F@H client did this automatically (re the CPU slot staying open), so hopefully they did for him as well. Passkey is good to look for as you indicated.
@Sonorous
Yes, I am using a passkey and no, I have not configured anything special.
hey @gattsu what is your folding name on the team?
@primesuspect
DJ_The-Quickness