Folding Forum

I'm a tech novice, and a newbie to FAH. I need help selecting the proper client(s) to run on my new office desktop PC.
Old machine ran XP with 1 session of the classic CPU client on 1 core, large WUs turned on; easy, although some work units take 2-3 weeks to complete.
Not sure if I did the installation wrong, but the classic client does not start automatically on boot, it must be manually started, and I frequently forget to do that.
New machine is an i7 2.8 with 64bit Win 7, SSD, 8gb, and ATI HD4350. With classic client total CPU utilization rarely breaks 15% and GPU isn't being used. Seems like a waste of resources.
Machine is turned on when I come in and off when I leave, it is not turned on 24x7 or left on overnight. Periodically I use a very response sensitive app, and have to shut down everything else, like FAH, while that app is in use, so FAH will need to be paused or shut down and restared several times a day.
I want to try the SMP and GPU clients, but cannot bog down getting my real work done, and they don't seem compatable with being paused several times a day. I'm concerned by the developmental beta status of the SMP client. Work units that need to be completed and returned in 2 days would likely be a problem, the machine frequently would not be on enough, and is frequently off for several days in a row.
I get the impression that this GPU is low end enough that it probably wouldn't do much useful FAH work anyway, and don't know if this would need to be paused or not to run my response sensitive app.
I understand vaguely the concept of running 8 sessions of the classic client, but having to repeatly start up, pause, restart and etc 8 sessions is too burdensome to getting my regular work done.
Am I pretty much locked into just running a single session of the classic client?

I would run 4 (or 8 (I'd run 4)) classic cpu clients with your run schedule. You could either use the system tray client or during configuration, select running the console version as a service. As long as you place each client in a separate folder with different machine IDs, either method will cause F@H to run at startup and exit gracefully when you shut down.

I'd question how much FAH would influence your "sensitive application" In my experience, running the classic FAH client has an extremely small impact on foreground activities. I'll bet you really wouldn't be able to tell it is running. (The SMP or GPU clients do have an impact, but since you've already been told that they're not right for your computer, that's not going to be a problem.

I suggest you try multiple clients and then if your application is truely so sensitive that you can't run them, cut back until it's acceptible. As you add more of them, the impact would increase and if you run the maximum number, you might be able to notice it, but you'll be the only one who can judge, no matter what we say here.

You have to create a shortcut in Startup for each client if you want it to start automatically.

Since i7's use hyperthreading, using one thread even at idle priority will slow down the other thread available for that core. To advise the OP best, I'd need to know the processing demands of that response-sensitive application. For example, if it uses only one core, then the best FAH performance obtainable with minimal impact to that other application is 6 threads spread over 3 cores. Without affinity control, however, one could run only 3 FAH threads, letting Win7 automatically set that critical application for the remaining idle core.

Wrish wrote:Since i7's use hyperthreading, using one thread even at idle priority will slow down the other thread available for that core. To advise the OP best, I'd need to know the processing demands of that response-sensitive application. For example, if it uses only one core, then the best FAH performance obtainable with minimal impact to that other application is 6 threads spread over 3 cores. Without affinity control, however, one could run only 3 FAH threads, letting Win7 automatically set that critical application for the remaining idle core.

I agree with your assessment, but I don't think even 3 would meet your requirements unless HT is turned off. If the application is so sensitive that it can't share a CPU, then 3 clinets might even be too much. Suppose 3 virtual cores are busy with FAH and the sensitive application is started. Windows will see 5 free virtual cores and will assign the application to one (or more) of them. There's still a certain statistical probability that it will pick one of the virtual cores that shares a real core with one of the FAH applications. To make sure it would work the way you'd want it to, you'd have to set affinity.

Another approach is to run 4 clients, but to throttle their CPU use to 95% so the foreground app can immediately grab SOME CPU time while F@H is relinquishing the rest. The lowest I've ever had to go to ensure foreground app response was 80% on a P3-766. My dual P3-550 worked fine at 95%.

However, if the OP really does have to shut down F@H, he could easily write a batch file to start the 4 instances when he wants, and simply close the 4 windows before he opens the other app.