[jswayze]

I've been moving up the ranks quite nicely with SETI, but with all the talk about F@H, I thought I'd throw a few GHz its way.

I downloaded the CL client, fired it up and I'm off an running. Thing is I've now got a terminal window tied up.

What's the "MacNN Way" of configuring and transparently running F@H on a dual processor machine? I use SetiDockling for Seti, but I didn't see a comparable app for F@H.

Though I haven't started a second client in parallel yet, I don't see how two clients can share the same ~/Library/Folding@Home directory.

Some help?

Thanks,

Jeff

[mikkyo]

For dual processors you can run 2 copies of the Folding Folder.
You need to configure the clients to have different machine numbers.
So dup your existing Folding directory, cd into the new copy, remove everything but the client(it is easier this way), and launch with
./OSX-3.25 -local -config
The local flag keeps your work units in the directory with the client.
Run through the config, change the advanced stuff and you will see a setting for machine number.
The first client will already be 1, so make the second 2.
I would run both with the -local flag to keep the work directories in the client folders, but I believe you can run them without local now if you want the work in your home dir. Im not positive, but at least if the machine ids are different it should work.

As to getting it to run in the background, that is easy.
I launch my clients like this as root:
/usr/bin/nohup /usr/bin/nice -20 ./OSX-3.25 -local -advmethods -forceasm &
Nohup makes it run even when you aren't logged in, nice sets the nice level to use as much of the processor as possible (20 is be friendly with other processes) and the & puts the process in the background.
Enjoy.

[billybob128]

so wat line needs to be typed in for it to run when not logged in and would this allow it to run when the machine is asleep?

[jswayze]

Thanks, I think I've got a pretty much figured out. A couple questions though:

1. Does the client "benchmarking" somehow determine your CPU speed and assign a bigger WU to faster CPUs? I'm running a Dual 1GHz G4, and each processor now has the largest WU (678 & 680) out there. Sheesh - it'll take me a week to finish one WU!

2. Along the same line, even though 678 and 680 have the greatest deadline (50 days) they don't have a proportional score (70.90 ea) - is this normal? I don't really understand their scoring system

3. I read in the FAH forums that core #64 (the one I'm running) doesn't support Altivec or any other accelerator. My question is - does any core support Altivec? If so, is there a way to get these WUs or it is it totally random?

4. It sounds like the -advmethods and -forceasm switches are geared for PCs - is there any benefit/detriment to using these on my Mac?

Thanks,

Jeff

Originally posted by mikkyo:
For dual processors you can run 2 copies of the Folding Folder.
You need to configure the clients to have different machine numbers.
So dup your existing Folding directory, cd into the new copy, remove everything but the client(it is easier this way), and launch with
./OSX-3.25 -local -config
The local flag keeps your work units in the directory with the client.
Run through the config, change the advanced stuff and you will see a setting for machine number.
The first client will already be 1, so make the second 2.
I would run both with the -local flag to keep the work directories in the client folders, but I believe you can run them without local now if you want the work in your home dir. Im not positive, but at least if the machine ids are different it should work.

As to getting it to run in the background, that is easy.
I launch my clients like this as root:
/usr/bin/nohup /usr/bin/nice -20 ./OSX-3.25 -local -advmethods -forceasm &
Nohup makes it run even when you aren't logged in, nice sets the nice level to use as much of the processor as possible (20 is be friendly with other processes) and the & puts the process in the background.
Enjoy.

[Welnic]

1) The benchmarking checks the speed of your machine, but I think it is geared for the old, slow end of the spectrum. From running it a little with some slower machines it seems to think that anything faster than 400MHz is fast.

2) Their scoring system is based on how long it takes a 500MHz Celeron to do each particular unit. The deadline is also based on this but they tend to round off the deadlines. Link to scoring and deadlines in FAQ:
http://www.stanford.edu/group/pandeg...#stats.howcred

3) The Tinker core that you are running does not support altivec. The Gromacs core does. Which one that you get is just luck, since your processor supports both.

4) At one point you had to use -advmethods to get Gromacs cores, but they are now out of beta so it is no longer needed. The -forceasm means that it will use advanced processing units like SSE or altivec even it crashed the last time that it tried using those. I don't think that there is a problem that you need to use this with a G4, but I don't really know that.

Even with the Gromacs cores I was never that impressed with how the G4 did at F@H compared with my athlons.