There's a cboard group at folding@home that has been inactive lately. I'm going to join them and try to pick up some of the slack. Would be fun to see this group gain some momentum again.
http://fah-web.stanford.edu/cgi-bin/...&teamnum=43476
There's a cboard group at folding@home that has been inactive lately. I'm going to join them and try to pick up some of the slack. Would be fun to see this group gain some momentum again.
http://fah-web.stanford.edu/cgi-bin/...&teamnum=43476
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
My i7 is a 940, "Bloomfield", but I don't call it that -- that's my "baby"!
I've been off my folding for awhile now, getting a little processing done for SETI on a C2D system. Yoohoo! ET?? Where are you??
My folding team is NOT Overclockers.net! -- ACK!, The name is confusing, but I'm on the much older Overclockers.com. We're much smaller, but per person, we fold at a very high average rate.
I'd be glad to help get some folders up and running over here, but when mid-April rolls around, I'll be folding for my team, in The Great Chimp Challenge race. This year, it's a handicap race, so for a change, we will all be running on the same "level" (more or less, but this is a race, where bringing in people to help, isn't cheating - it's encouraged!).
If you have a plan for this, or think of anything I can do to help, don't hesitate to ask. I've done a TON of folding for Stanford, over the years. Glad to pitch in.
Overclockers.com team data: http://folding.extremeoverclocking.c...ry.php?s=&t=32
As you can see, Stanford has seriously changed the point system, massively inflating the points (although the hardware advances have caused a lot of it also).
Last edited by Adak; 03-26-2011 at 06:37 PM.
Hmm... I can see you guys are running SMP and GPU clients. Has to be, no? Possibly also running multiple computers. That daily average of most of you guys is insane! I considered the idea of running the GPU client. But the thought of forcing my GPU fan to top speed for the entire day... nah!
I'm running the cpu client instead. Much slower, but what the heck. I don't even have to remember it is running. But if you don't mind stepping in for even a couple of days or donate us a week, cool! This group certainly needs a push.
EDIT: I'm not sure why my first unit only scored me 69 points. It only(?) took 5 hours to complete. I will have to go back and relearn all this stuff.
Last edited by Mario F.; 03-26-2011 at 07:02 PM.
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
So it turns out points are scored based on project, with just a small margin attributed to speed. I like that.
In any case, I've turned to GPU folding and am getting around 1400 points per work unit. Each WU takes around 2:40 to complete, so I'm estimating a PPD of 12,600. Will keep folding with the GPU for a week. Then I'll turn back to the more modest CPU folding; My GPU fan runs at 40% when folding (temps are fine at around 58c) and although not too loud, it still is a big nagging. I adore silence.
I'm amazed at the processing capacity of GPUs. One has to see it to believe it. DirectCompute and OpenCL are awesome technologies that no doubt one day will be properly tapped by applications that require this kind of computation.
Last edited by Mario F.; 03-27-2011 at 06:30 AM.
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
The client you have on your cpu is the "classic" client, or your have your settings wrong, so are limited to "classic" wu's (work units). Minimum points and benefits, at this time.
Follow the hyperlinks to the high performance cpu clients. The BIG one's are REALLY big, but if you have an i7 quad core with 8 cores (hyperthreaded turned on), you can do them, but there are several constraints. Follow the links to the Folding@Home forum for all the details.
The multi-threaded (big but not BIG) clients are a good choice for a client. Any C2D cpu or above, will have no trouble and the client is also a bit more stable. Generally, this is the client I recommend.
Yes, the team folders have several very dedicated folders who either run FAH at their business and/or at their home, on several computers, and of course, it IS the Overclockers.com team, and we do O/C quite a bit. A bit less now that the electric rates have gone up so much.
My quad's are torn apart right now. One is waiting for a graphics card, the other is being re-built prior to the Great Chimp Challenge race. I can put CBoard on a C2D, however.
I've killed a couple graphics cards folding on them - so I no longer recommend it. I watched my temps carefully, but they simply aren't designed for 24/7 folding yet, imo.
Can you be more specific about those GPU issues you mentioned you had?
What cards were those?
Where they overclocked? Heavily? Some? No?
What were their temps when folding?
Where you running GPU, GPU2 or GPU3?
Where you running 24/7?
How long before they got killed?
Couple as in "2", or couple as in "some"?
I confess, apart from clearly misuse of a GPU or manufacturing defect, I've never heard of GPU failures due to constant use.
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
They were GTX280 N'vidia based cards, and they came already overclocked. One I ran at the default overclock from the factory, the second one I overclocked manually, using the Nvidia tool program. Both were mid-range overclocks, but on the manual one, I upped the fan speed controller, so I could give it more air.Originally Posted by Mario F.
It's been almost a year, so I hate to say what the temps were. Never "obscene", and within the mid range as defined by the overclocking program tool. They ran GPU and GPU2.
Yes, they ran 24/7 unless it was a hot day. I'm in a semi-arid climate, so Summer days get too warm to want to run a wu of any kind.
The first one died within 6 months. The second died just after 1 year.
Go to the Folding@Home forum, and you'll find plenty of posts about GPU's crapping out. Coincidentally, they were folding. Judge for yourself.
N'Vidia makes GPU cards for workstations, and super computers. They run considerably slower than these graphics cards, but they use the same chips. IMO it's because the business cards are designed for longer life, and the graphics cards are designed for maximum possible performance, regardless of how much it might shorten the lifespan of the card.
If you want to read a funny (aka horror repair) story, google "oven fix for graphics card".
If you aren't using a monitoring tool yet, you might want to grab one. This page has links to the best one's: * HFM.NET - Client Monitoring Application for Folding@Home * - Overclockers Forums
HFM.NET is the most thorough. I use FahMon.
Checking the C2D, I have just started a 30 hour to crunch "Astropulse" wu in SETI. I'll be joining Cboard for a while, right after that.
Last edited by Adak; 03-28-2011 at 09:48 AM.
Yeah. Got HFM yesterday. Installed and set it up but can't honestly interpret some of the data. Decided to bother with it later.
As for the GPUs... scary stuff. I can't even fully grasp what makes a gpu processor less able in dealing with intensive work than any normal cpu with similar manufacturing processes. I suppose they probably die on the memory chips, no? I mean, due to the constant (almost 24/7) heat... I always looked at memory as the Achilles' heel of any GPU board design.
Anyways, as I said I don't have plans to run the GPU client for more than a week. Will then switch to the CPU client and turn the smp switch on. Won't be folding bigadv though. My only i7 is a laptop and I won't be using that for folding.
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
Well, I got into a fine mess with BOINC (the wu manager for SETI). I have some wu's that need to be returned quickly, so I suspended the wu's I was working on, and tried to get BOINC, to start crunching on those, instead -- but it wouldn't do that.
Instead, it stupidly went down the entire page of wu's that are already d/l'd on my system, and tried to run each one, in the order they were received - regardless of the date they are due back.
So, I wound up with some 70 odd wu's, that were all started, but immediately suspended, before I got to the right one's that had to be finished, before I could start folding@home work.
Now, the problem is, each one of those 70 wu's is using a 5-50k of memory, and this system is barely usable at this time. Folding is out of the question, until I get a lot of these SETI wu's crunched and the memory freed.
It's working it's way down, but it will be three more days before I have enough memory left to start folding@home work.
You're a very generous person to let distributed computing take over your computer like that. Well, you're really searching for aliums, so maybe I'm being a bit sarcastic about that, but still.
Figured I would bring it back with you Mario
Woop?
Yeah. I've been seeing you for the past 3 or 4 days, prog-bman
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
Yep. Got a couple of servers here that aren't being used to full capacity quite yet, so why not use them for something good. And running it on my desktop and laptop.
Woop?
Thanks a lot, but the protein folding at Folding@Home, is by far, the biggest project I donate to**. Overclockers.com has a SETI team as well however, and they pitch in big time to help my Folding@Home team, in our team races.
So my last 11 months have been a big "thanks" for last years Great Chimp Challenge Race help from the SETI team.
I believe it's all but impossible to detect a radio signal from another galaxy. The amount of dust and other interferences in space would scatter such a signal, making it all but impossible to discern such a man made signal, from the naturally occurring EMF generated in space.
But, the SETI guys love their SETI project, and I want to thank them for their support, so... come on, ET!
Anyway, the Great Chimp Challenge is one of the most fun things you can do with a computer, with your clothes on -- you wouldn't believe the excitement (grown?) men get into over a race for the right to display a plaque of a monkey on their website, and add some buzz about Folding@Home. It gets crazy! (Some of the really intense race teams begin organizing and optimizing their hardware and software in JANUARY, for the race which begins May 5th.)
** Adak's contributions to Folding@home
Last edited by Adak; 03-30-2011 at 07:54 AM.
Codeplug, ProgBMan and I have been steadily increasing Cboarder rating over the past 2 weeks.
CBoarders climbed 12 positions over the past week alone and the list of groups' positions we are threatening is very large.
Within the list of CBoarders users, we are pulverizing everyone's past efforts. Codeplug raised three positions and is bound to raise a lot more (and fast) due to his adoption of GPU folding. ProgBman has it harder because he has close to 100,000 points to bridge in order to climb to the 2nd position in the group. Meanwhile, I've climbed 51 positions on the past 2 weeks taking advantage of GPU folding and no competition, until Codeplug rejoined.
I'd like to see a lot more of you folks rejoining. Currently you have three options for folding:
CPU: Less intensive single core client. Folds projects in the 60-80 points per work unit, for maybe an overall 210 points a day.
SMP: Multi-core/multi-processor client. Folds projects in the 700-900 points per work unit, for a daily total of 2,000 to 4,000 points a day. It uses all your core processors, but my tests revealed it is conservative. For instance, it doesn't raise your CPU temperatures above Prime95 levels. Good option overall and it really operates at the lowest possible priority which means you can use your computer normally.
GPU: Multi-Core GPU client. The fastest of them all, taking advantage of GPU's preference for floating-point calculations. It folds projects in the 1200-1500 points per work unit and offers something like 10,000 to 17,000 points per day. If your GPU is a modern one and your case is well cooled, I'd suggest this one as long as you don't keep it running 24/7.
PS: Playstations. Also a very fast client, just a tad bit under the GPU. But I know nothing of it. Didn't test it because I don't own a PS.
Originally Posted by brewbuck:
Reimplementing a large system in another language to get a 25% performance boost is nonsense. It would be cheaper to just get a computer which is 25% faster.
