F@H Linux advice.

JosiahBradley

[H]ard|Gawd
Joined
Mar 19, 2006
Messages
1,791
Hello everyone, I'm about to hit 1000 WU and I have mainly been using my PS3 for the past few years. I just installed fah6 on my workstation and was wondering what type of work it should be doing given it's specs. I can dedicate 4-6 cores (Thuban) depending on time of day (each at 3.2Ghz) and I have 16GB RAM.

Is there any particular way I should be running the client or WU that are suited to this machine. Everything seems to run smooth even running the client as I work. The only problem I;ve encountered was stopping the client (Ctrl-C) and restarting it to switch the number of cores (6 down to 4). This resulted in a loss of a WU. I may have stopped it while it was writing a checkpoint or perhaps lowering the cores caused it to fail.

Also I am seeing a lot of mention on -bigadv, is my system capable of such WUs?

Any suggestions to ensure smooth operations form here on. Thanks for any advice.

Also I have an ATI V4800 GPU and was wondering if I could put it to use as well. I've read some guides about setting up WINE to run the windows GPU client and was wondering if it was worth it.

 
Last edited:
Regular SMP is where you need to look. -bigadv currently requires 8 cores, but is increasing to either 12 or 16 supposedly.

Changing cores in the middle of the work unit can very well bork the unit.
 
I'd suggest regular smp on 4 cores if you really notice the client on 6. Changing core count mid-WU is not a great idea.

I also wouldn't bother with the video card. ATI cards just do not do well folding.
 
Thanks for the advice. I'm going to stick to 4 cores. And thanks for saving me from having to attempt tackling wine once again. The only GPU I folded on before was my 1900XT when the first GPU client was released but I shortly lost Internet connectivity after that so never got back to it on my gaming PC.
 
I have solved my 4 vs 6 core problem with a simple solution. I am running a 4core client during the day and night requesting big WU that take about 22 hours to complete. At night when I'm not using the machine I am running a second client with 2cores requesting small WU to churn though.
 
I would suggest against the '^C' approach for switch between 4 and 6 core. I have seen more than enough WU lost doing it. Just don't. With the how well the linux scheduler works, you will not see any noticeable difference between 4 thread and 6 threads running unless you start off another app that uses multiple threads for an extended period of time. Eliminate those free CPU cycles :D

ATI cards are not supported to fold under wine.... at least the last time I checked. Even if they are now... if you are folding purely for points, the GPUs are not worth it point wise. If you are more interested interested in the science, for sure. The more different WU processed the better.
 
Yeah my goal is simply to fold. While I like getting points to mark my progress I'm not really too interested in maximizing points but helping the project out.
 
Especially if you're running 64-bit linux, getting the ATI cards to fold under wine will be painful at best and impossible at worst. I spent about an hour setting up CUDA just-so to fold on my nvidia cards under gentoo recently, only to find out I was getting a 27:00 frame time, which is several times longer than what they get under windows. I would say give the cpu client as many cores as you can afford, run standard smp, and enjoy the science =)
 
Back
Top