Resource hog.

[adrianxw]

Even set to "Light", FaH grabs at least twice often three times as much CPU time as BOINC or Distributed.net on my systems, (Windows 8.1 x64). Why is that?

[JimboPalmer]

Light equals threads / 2
Medium equals threads - 1
Full equals all threads

[adrianxw]

No, that does not appear to be the case. It is a quad core, eight thread machine, (Intel 6700 CPU). The task manager process shows FaH using ~30% of the CPU, it goes up and down a bit but centres about there. Threads/2 would be 4 threads, half the system, it is always less than that. Obviously, that is more than I want though.

[Joe_H]

If you post the first 100 or so lines of your log file that will show us the system and configuration as the F@h client software sees it. The slider settings are as listed by JimboPalmer. Usage can be further modified by settings in FAHControl, some will override what the slider does.

Options include removing the CPU folding slot or explicitly setting it to a different number of threads than the coarse settings of Light, Medium and Full. A GPU slot will use the CPU resources of one thread to prepare data to be sent to the GPU and retrieve the processed data for further calculations.

One thing to add, the Light setting normally places a GPU slot to process when the system is considered "Idle", i.e. no mouse or keyboard activity, in the case of Windows the the display is put to sleep.

[adrianxw]

Sure, where is the file, or its name so I can find it.

>edit<
The Windows process manager rarely shows the client going down to 30% actually, seems to be floating about 37-38%.

[Joe_H]

How to find and post from the log (Windows) - viewtopic.php?p=327412&f=24#p327412. There are similar posts detailing how to do this for Linux and macOS.

Which process is using 30%? FAHControl, FAHClient, or one of the FAHCore_nn processes? Typically only a FAHCore_nn process should be using any significant percentage of CPU, everything else will mostly be idle and use 1 or 2% at most.

An old version of the client had a memory leak, after running a while the increased VM size would slow it down and result in higher CPU usage. But that bug was fixed long ago.

[adrianxw]

The task name us FahCore_a8.exe.

[Joe_H]

That would be the CPU folding core processing a WU, that is CPU intensive calculations. 4 threads on an 8 thread processor would max out at about 50%, as it runs at almost the lowest priority other active processes may limit it to less. You can manually set the threads used by the CPU core lower in FAHControl.

[adrianxw]

Sorry for the delay, my girl was here and needed tea and to divulge all manner of useless info concerning who said what to who, who did this, what happened with that etc. etc. She has gone off to aerobics or training or similar for a couple of hours now. I have that "Client Advanced Control" task open now, interesting. I need to set it to a single core then, some configuration option I guess? The log that it shows does not look current, the dates and times appear to be from early last year?

[PaulTV]

You can set the number of CPUs to 1 (from advanced control, click on Configure, then on Slots, double click on the slot cpu, then set CPUs to 1). But if you're not interested in CPU folding at all, and only want to fold on GPU, then from the same screen where you set the CPUs, add an extra slot option with name 'pause-on-start' and value 'true'.

[Joe_H]

In FAHControl that would be click on Configure, that opens a new window. Then click on Slots, click on the CPU slot to select it and then Edit. The top item on the edit page for a CPU slot will be to set the number of threads used, a value of -1 leaves it to the software and the slider position. You can set that to 1 up to the maximum number of threads supported by your processor. After that the slider position only influences when a GPU slot runs. Save the changes by clicking Okay and then Save.

The log show in the log window of FAHControl should be your current one. It can show the log from the very beginning, but that may require clicking on the Refresh button to reload the log. If it actually is showing old information it is possible your install is messed up in some way. The current date is posted periodically throughout the log, and times are in 24-hour UTC format such as this copied from my log:

Code: Select all

13:28:08:WU00:FS00:0xa8:Completed 3400000 out of 5000000 steps (68%)
******************************* Date: 2022-11-08 *******************************
13:33:50:WU00:FS00:0xa8:Completed 3450000 out of 5000000 steps (69%)

[PaulTV]

If you tune the CPUs, you'll want to know that lowering the number is applied immediately, but raising the number is only used with the next job.

[adrianxw]

Okay, I set CPU's to 1, it was set to -1. Task manager still shows it using 40+%, I don't know how often it reads the controls though, so I'll leave that for now, might be threaded.

[adrianxw]

Yes. This morning FaH is running at a similar resource percentage as BOINC and Distributed, and all are running equally. I expect it to continue like that. It was getting close to the point where I dropped the project AGAIN.

[toTOW]

Running multiple distributed computing projects at once is never a good idea ... they're not really designed to work together correctly and each one assume that it will be running alone.

You'll have to make a choice or manually handle ressource share.