dont get points from SMP or GPU

[Joe_H]

I thought the cpu usage % doesn't work for SMP?

It only works correctly for uniprocessor slots. On SMP slots it just slows down one thread. That desynchronizes the threads causing the SMP WU to run much slower, but not decreasing the actual CPU usage by much. On GPU slots it can reduce the CPU usage of the core that feeds data to/from the GPU. That may have more effect on folding on ATI GPU's than nVidia as the ATI folding cores use much more CPU time. To the OP, you do not want to use the cpu-usage parameter to control CPU utilization on your machine by F@H except on uniprocessor slots. On SMP slots reducing the core count to 2 or 3 is more effective.

Going to the other changes visible in the logs that have been posted. When you added the uniprocessor slot, the SMP slot was left active. The CPU time used by the uniprocessor slot has not allowed it to complete its startup in the log section shown at any time. If you are going to run uniprocessor slots you should delete the slot, or at least pause it while testing how the uniprocessor slot works.

As for the problems with the GPU slot, someone more familiar with those may have an idea what is happening. But something does appear to be corrupting the checkpoints and preventing work from resuming.

is the v7 client avaible also on linux (ubuntu), so i could test that also?

Yes, the V7 client is available for Linux. Folding on a GPU is not supported under Linux though.

[iceman1992]

is the v7 client avaible also on linux (ubuntu), so i could test that also?

Yes it is. But GPU doesn't work on Linux by default (at least for now). Some have successfully run v6 GPU using wine..

[bruce]

Several general rules that may or may not apply to your hardware:

Computers that are "unacceptably slow" are probably using their GPU for folding. If the pauses are unacceptable, start by finishing or removing that slot/client. Either SMP or Uniprocessor client(s)/slot(s) have much less impact on foreground activities due to the way the OS handles task priorities.

Computers that miss the SMP deadlines, whether due to a slow CPU or due to not folding enough hours per day should switch to Uniprocessor slot(s)/client(s). Many of the WUs are the same but SMP-only projects have shorter deadlines than the projects that are assigned to either Uni or SMP.

A SMP slot/client will automatically configure itself for the maximum number of threads it detects in your hardware. (Four, in your case.) SMP does not perform effectively unless those threads are rarely used by things other than SMP. Running SMP + Uni is a really bad idea unless you reduce the SMP core allocation or suspend it. You can get away with GPU+ Uniprocessor if you have a classical NVidia GPU because it uses very, very little CPU time. SMP plus an AMD GPU or even SMP + Torrent will probably play well together, either.

CPUs that are hyperthreaded (Intel) or share a FPU between cores (AMD) will only gain a fraction of FAH performance by loading up all of the threads. You may CHOOSE how many uniprocessor client(s)/slot(s) you want to run and see if it makes a difference. That also does change foreground performance, but it's a small enough difference that most people don't notice either way. (Running a torrent, though, can have a much bigger impact.) [For the atom 330 (2 cores), I'd recommend 2, not 4m leaving the other two threads for other things.]

That's a lot of unnecessary information since you've already got a recommendation to run two Uniprocessor slots but it backs up what you've already been told.

[CaqKa]

Going to the other changes visible in the logs that have been posted. When you added the uniprocessor slot, the SMP slot was left active. The CPU time used by the uniprocessor slot has not allowed it to complete its startup in the log section shown at any time. If you are going to run uniprocessor slots you should delete the slot, or at least pause it while testing how the uniprocessor slot works.

somehow I managed on my own to put the smp and gpu on pause while i tested the uniprocessor.
right now i am havin one slot for gpu, one for SMP and two for uniprocessor. the smp is on pause, the gpu is running but showing no processing? the uniprocesor is doing great (~22% both).
also i have in the work queue a job for sending, which is on 0% from morning?
updated log: https://dl.dropbox.com/u/1183736/log-20120911.txt

does the gpu not work at the same time while there is a uniprocessor or SMP running?

[P5-133XL]

The GPU slot should work fine with the other slots. If it isn't there is something else going on (like you paused it).

I think I would dump the smp slot and just fully use uniprocessor slots + GPU if you can get it to work properly.

[Rel25917]

thats utorrent, dropbox, chrome, launchy, googledrive sync. others are minor...

Not sure if torrents could cause the checkpoint problem but might be possible. I know they can definitely cause the connection issues in the log when the gpu slot tried to upload(although uni slots upload fine). Not sure exactly how google drive or dropbox work but if the work folder is being monitored by one of them I can see it getting in the way if they try to sync it.

[CaqKa]

no, the gpu is not paused, it says "Running" but is on 0%. seems like it is connected with the issue that some data is not send.

updated log: https://dl.dropbox.com/u/1183736/log-20120911.txt

dropbox and google drive are monitoring other directories... i mentioned the programs because process explorer says they are the heavest on I/O.

[CaqKa]

the server to which my client is sending results is "not accepting" more than one month??? http://fah-web.stanford.edu/logs/171.67.108.25.log.html
why is the client even sending the results to this server?

[bollix47]

171.67.108.25 is a collection server that has not worked for a long time. The server that we need to see is the work server the client tries to send before trying the collection server.

Troubleshooting Server Connectivity Issues (Do This First)

[CaqKa]

i do not understand why is there a fallback on a server thats not working? shouldnt that be programmed so that it finds it self a working server?

logs:
https://dl.dropbox.com/u/1183736/log-20 ... 140854.txt
https://dl.dropbox.com/u/1183736/log-20 ... 014240.txt
https://dl.dropbox.com/u/1183736/log-20 ... 055613.txt
https://dl.dropbox.com/u/1183736/log.txt

[CaqKa]

when clicking on the box with send in the queue, the right side says that the work server is this:
http://fah-web.stanford.edu/logs/171.67.108.11.log.html

which seems to be very busy...

[bollix47]

First, let's look at why the client didn't send back any results to the WS(171.67.108.11). It appears that your GPU is unable to start folding and when the client tries to send the results back there's either nothing there or it's corrupted:

"Failed to send results to work server: Failed to read stream"

You should Pause and Remove your GPU slot until this problem is resolved.

What video drivers are you using? What version of Windows are you using? (e.g. Windows XP 32-bit)

[Rel25917]

Busy, but working normally. If you put this in your browsers address bar do you get an OK? http://171.67.108.11:8080/ Can you pause the torrents for a while to see if the connection issues go away?

Edit- After a closer look at the logs it seems your gpu is failing at the start of all work units now no matter what and that makes me think hardware problem. How old is this computer? If it has been around a while are the vents full of dust? Could be gpu overheating as soon as the core starts up.

[CaqKa]

First, let's look at why the client didn't send back any results to the WS(171.67.108.11). It appears that your GPU is unable to start folding and when the client tries to send the results back there's either nothing there or it's corrupted:

"Failed to send results to work server: Failed to read stream"

if nothings there, why would the client even bother to send? wouldnt it be better if it send to the server, that he got nothing to send and that the server should assign the wu to another computer?
how should I know that "read stream" is related to the client and not the server side?

You should Pause and Remove your GPU slot until this problem is resolved.

the gpu slot is removed at least for two days.

What video drivers are you using?

1/2 weeks ago when i installed fah v7, I also then installed the newest drivers.

[CaqKa]

Busy, but working normally. If you put this in your browsers address bar do you get an OK? http://171.67.108.11:8080/ Can you pause the torrents for a while to see if the connection issues go away?

Edit- After a closer look at the logs it seems your gpu is failing at the start of all work units now no matter what and that makes me think hardware problem. How old is this computer? If it has been around a while are the vents full of dust? Could be gpu overheating as soon as the core starts up.

getting "OK"
yes i can shutdown utorrent.

the computer is.. hmm from januar 2010. i clean the laptop regularly with preassureized air, so there is no dust.
but there is a chrome

hmm there is now somethings other that came to my mind. the task manager of chrome, says it has a "GPU process", does anyone know if that makes the problems?