Looks like its having problems getting a WU. Can you attach the actual file?

Sometimes some of my computers do this when my internet is being flaky. The client will then hang indefinitely at "Download."

Run the following commands.
1. Pause all slots
2. Run the following in terminal to attempt to stop the client

Code:

sudo /etc/init.d/FAHClient stop

3. The previous was probably unsuccessful (FAHControl should disconnect). If FAHControl is still connected to the client run this to force kill it

Code:

sudo killall -7 FAHClient

4. Restart the client

Code:

sudo /etc/init.d/FAHClient start

5. Resume all slots

Hopefully it should start downloading. You can also try deleting the WU thats running on the CPU slot (pause it first) and see what happens. WUs can be found in the /var/lib/fahclient/work directory. You will probably have to run an elevated terminal to delete the directories.

 

Ah, sorry... should have clarified that I went back to Win7.

Originally the slot started at 10 threads, so I adjusted it to 11. I think that was my issue, because I removed the slot and returned it back to 10. Now it's folding fine... but I want that extra thread running

 

11 threads is definitely the issue. Folding@home doesn't play nice with prime numbers of threads. If you ask for 11 threads you are effectively telling it to use all of your CPU since two threads = 1 core, and any one thread can use 100% of a core's resources. Unless you have a machine dedicated to CPU folding you should always leave one full core free for other tasks or one core per additional slot, else it will massively slow down the whole client. Reason is because the client attempts to run the threads all with the same workload, so if one gets bogged down with other tasks, all of the other threads slow down too.

 

Back on my 4820K I could have sworn that I used to run it at 7 threads... Even in Ubuntu (without configuring the GPU slot at all) it would run the CPU at 11 threads just fine (before I messed everything up).

 

The code base that FAH has been using doesn't break down the work into prime numbers very well. WUs that allowed 5/7/11 threads were always limited compared to even number threads. This year they stopped allowing 7 and 11 from what I remember. Depending on the CPU that can be the default setup with a GPU so Stanford is shooting themselves in the foot.

 

Good to know! Stanford is definitely missing out on some serious power... I'd guess about 1 thread per user with hyperthreading.

 

Hyperthreading allows two threads or processes to run on one CPU core but they are both competing for that core's resources. As in, for both threads to go the same speed they would each have to use 50% of the core's processing power. But, one thread can run at 0% while the other thread runs at 100% and uses the core's full processing power. So in effect the same amount of work would be completed running 11 threads vs 12.

The problem is that most SMP work units do not play nice with prime numbers since SMP stands for symmetric multi processing and mathematically speaking its not possible to be symmetrical when working with prime numbers.