lifeskills
2010-07-02T03:14:04Z

I read an article in Fast Company recently about a program called World Community Grid, which seems to be a program similar to folding@home.  I have a few PC's lookin for work, and this seems to be just the job for them.

Could anyone point me in the right direction to get me started?  

Does folding@home rely on CPU or GPU?

Does this program run in the background? (Might keep it off of my main rig if it is too memory intensive)

What are the hardware requirements?

Excuse me if these questions are answered elsewhere.

Thanks!

AKwyn
  •  AKwyn
  • 76.5% (Friendly)
  • Advanced Member
2010-07-03T21:10:06Z

Folding@home has a regular client for the CPU but a special high-performance client for dual-core CPU's and GPU's. So it relies on both but only depending on the version.

I don't know if this program runs in the background but you might as well snoop around and check to see if it does.

crowTrobot
2010-07-05T19:12:06Z

Technically any computer can do the work, the only problem is that there are deadlines to be met and if your computer is slower then the data you worked on will just be discarded because its late.  (recommended at least a core 2 duo or better). GPU wise Nvidia cards have an advantage as the work output you get from the nvidia cards are much higher, AFAIK its still not very optimized for the ATI cards and the work output isn't as good as Nvidia's.  The client can run in the background and utilize unused CPU work (idling, etc).  IIRC you can adjust it.  Also it can get stressful to your power supply (a dedicated folding machine can use up a lot of electricity).

acarzt
  •  acarzt
  • 100% (Exalted)
  • Advanced Member
2010-07-08T02:16:19Z

I was able to run 8 instances of the CPU client on my Quad core with hyperthreading with some customization. I specified each instance to a specific core.

Then I also ran 2 instances of the GPU client and specified it to a specific GPU.

It takes a little customization but once it's done, you make a batch files to run them all at once. I turned my rig into a folding beast lol

It's supposed to run in the background and other apps will take priority. You still might see some delay tho. Try it out and see if the performance impacts your computer too much or not.

slugbug
2010-07-08T13:40:22Z

With a Core i7 920 you really should be using the Windows SMP client with a passkey to get the bonus points. Also if it is overclocked you could run the -bigadv work units which can give upwards of 60,000 points.

werty316
2010-07-08T23:02:03Z

You can use your CPU and/or GPU for folding@home.
It can be set to run in the background as a service and requires little memory.
There aren't any hardware requirements unless you run the GPU client.

slugbug
2010-07-11T12:45:42Z

Your GTX295 should be able to run two GPU clients, and One Windows SMP client for your 920.

lifeskills
2010-07-11T14:16:42Z

finally just installed this.  Acarzt, how did you run multiple instances?  I am going to wait until my HD5850 gets back from RMA before I install the GPU client.

werty316
2010-07-11T17:53:48Z

The easiest way is to make another folder that contains all the needed file for folding.

 

This guide should help: http://forum.xcpus.com/f91/guide-running-client-multiple-nv-gpus-12627.html

acarzt
  •  acarzt
  • 100% (Exalted)
  • Advanced Member
2010-07-11T18:08:39Z

I downloaded one of the multi core clients but when I would run it, i noticed it wasnt putting my CPU at full load.

So what I did was.. I made a folder on my desktop and placd 8 shortcuts in the folder all leading to the exe file of the application. Number your shortcuts 1-8 (or however many cores/threads you have)

My CPU can run 8 threads at once so I had 8 shortcuts. Right-click and click properties of each shortcut. In the target line enter C:\Windows\System32\cmd.exe /C START " " /affinity 1 at the beginning of the shortcut so it should look like this

C:\Windows\System32\cmd.exe /C START " " /affinity 1 "c:\program files\whatever\Program.exe"

Change the number on each shortcut 1-8 to assign 1 instance to each core.

I don't remember if I used the single core or multi-core client for this tho lol So you might have to do a little experimentation. But using this method I was able to acheive 100% CPU usage.

I did the same thing with the GPU, assigning each instance to a specif GPU. But the syntax is different so just let me know if you ever need that as well.

acarzt
  •  acarzt
  • 100% (Exalted)
  • Advanced Member
2010-07-11T18:11:44Z

It's been a LONG time since I had this setup. So if you run into issues, let me know and i'll do what I can to help you figure it out :-)

werty316
2010-07-11T22:24:57Z

The SMP client is multi-threaded so you should really only need to run one client window especially if you are running SMP2 WUs; if one client alone doesn't put a 100% load on your CPU you can run two clients and set the affinity of each to -smp 4.

slugbug
2010-07-12T14:04:42Z

http://folding.stanford.edu/English/DownloadWinOther

Scroll down to Windows: V6 Beta SMP2/CPU clients and get the 6.29 beta for MPICH (32-bit or 64-bit)

http://www.evga.com/forumsarchive/tm.asp?m=503846 

http://fah-web.stanford.edu/cgi-bin/getpasskey.py 

Definitely more productive than running 8 single core clients.

acarzt
  •  acarzt
  • 100% (Exalted)
  • Advanced Member
2010-07-12T18:07:02Z

I really don't remember what kind of points I was racking up. I remember it being a lot tho. But I had my GTX295 heping out too.

lifeskills
2010-07-12T22:38:19Z

Thank you guys for all the help.  Got my first WU earlier today.  Now that I kind of understand this stuff, im going to start with the multiple core stuff/ SMP client later