Folding @ Home F.A.Q.

[3.0charlie]

Thanks. I'm averaging 15 secs per frame on that 20000 frames WU.

[pscout]

You can only compare frame times between 2 wu's from the same project.

The other more common measure for us points junkies is ppd (points per day).

For benchmarking comparisons ppd/ghz also normalizes things.

For the standard single core windows client, stanford benchmarks the wu's before they are released on a p4 2.8c with HT turned off. On the smp wu's, they benchmark on a quadcore woodcrest. I think the standard core client benchmark is ~ 60 ppd/ghz. If the wu's use a lot of memory or network bandwidth for up/download, they are also given bonus points.

Some cpu's/architectures will do better than the benchmark and some will do worse. Some wu's use sseII heavily, some don't. Cache can make a huge difference on some wu's.


While the points system is good for creating some lively competition between individuals and team, it can also be a sore point. For example, when a lot of bonus wu's are available and they heavily favour p4, c2d or AMD, a few people get very annoyed, and outspoken. Others try to figure out cheats. Just like in real life, a small percentage of the population become a problem.

Fortunately, most people are committed to folding for the scientific benefit to humanity. I just like to make sure i fold as efficiently as possible and encourage everyone else to do the same.

[3.0charlie]

Thanks for the info - I'm learning bit by bit. As soon as I'm done with my customer's rig, I'dig more info on the subject.

John

[pscout]

Thought i would link Charlies other thread on a c2d 2140 test rig since it had some folding discussion .. info on userid's for points/stats and instance uniquie id's

http://www.hardwarecanucks.com/intel...-test-rig.html

[sswilson]

Really wierd snag this afternoon with the SMP beta.........

Don't know how, but it looked like the client ended up set to install as a service (clicking on the fah program opened a window for less than a second and smp was "in use" when I tried to uninstall the client).

Had a bugger of a time getting it uninstalled, and then had to delete the entire folder + dl the latest version of the beta.

Spent all afternoon playing with the mem timings on that rig so thought it was something to do with that, but returning to original settings didn't fix it........

Finally up and running again though.... should have a couple of c2d's running this week at least.

[pscout]

You will get that behaviour if you try to relaunch the cleint while it is still running .. this should be logged in the fahlog.txt - that an instance was already running.

I think you are runnning win smp? When you stopped the client before updating the expired beta, did it stop clean? If it didn't you will still have mpiexec and 1 or more fahcore's running in taskmanger. After stopping the client, if they are still running, kill mpiexec and it will stop any cores still running.

I hope you didn't lose your work in progress. You can copy the wu (work dir + queue.dat) to a backup directory when you need to do 'extreme' maintenance. You can even copy them while the client is running as long as the client isn't in the process of writing a checkpoint. I used to do this regularly in the early beta to try to avoid losing work - I would backup it up before stopping the client since the stopping was not alway clean and would sometimes cause the client to restart from the beginning of the wu. if ididn't shut down clean, then just kill the mpiexec taks, delete the current work and queue, and copy in the backup. restarts would usually start from the last checkpoint.

Happy to hear about the new c2d's
Hopefully they are 4 MB cache?
It will be a very nice production boost

And you are already leading the team in production !

And congratz on the #1 team spot and passing the 100k milestone.

[3.0charlie]

Quick question: I have 2x F@H running as services - Task manager shows both. But how do I open the console? If I do, it gives me the following message:

Timeout (something)

Also, I can't restart F@H since I can't close the Core and Console programs...

[pscout]

Originally Posted by 3.0charlie

Quick question: I have 2x F@H running as services - Task manager shows both. But how do I open the console? If I do, it gives me the following message:

Timeout (something)

Also, I can't restart F@H since I can't close the Core and Console programs...

The easiest way to monitor folding services is to use fahmon FahMon Homepage

With it's view log option you will be able to see the same thing that is written to the console. If you set up network share to the folding directories, you can monitor all your instances on all rigs via your lan from 1 machine.

You can also just open up the fahlog.txt file in the folding directory.
Everything written to the console is also written there. Fahmon is more convenient than refeshing a editor view of the log.
It will also calc ppd, eta and flag an instance if it appears to be stopped or frame times are out of whack.

Here is a link to the support thread at FC Folding-community.org :: View topic - FahMon: A monitoring tool (Linux & Win32)
Uncle_fungus, one of the mod's at FC, has taken over support from the original author. He added the support for smp and gpu clients and also has a version that can monitor from linux for the brave.

and a link to a list of folding tools there Folding-community.org :: View topic - Tools List

EMIII electron microscope used to be the favourite monitor with more function than fahmon, but it has not been maintained since gpu and smp clients came out. It should still work ok for the current standard clients but who knows when stanford releases their new V6 folding client. I don't run emIII anymore since my farm is almost completely SMP now. But EMIII had much better stats views and database function and also supported pretty protein folding views (which i never used). It also had built in support for viewing the internals of the queue, etc etc

[sswilson]

Just a quick note that fahmon now works right out of the box with the SMP beta.

Seems to me I looked at it a while back and it required a few hoops to jump through with SMP so I set it off to the side to look at later.

I installed it last month and was pleased to see that it worked without any problems.

[3.0charlie]

Thanks Pete, that's exactly what I was looking for.

John