How to make the CPU idle but the GPU to crunch points

[gopinathms]

I want my FAHClient to be configured in such a way that the client must not use CPU but use the GPU to full load

[ComputerGenie]

...configured in such a way that the client must not use CPU but use the GPU to full load

Open the Advanced Control...
Push slider to "Full"

Then
Configure..
Slots...
cpu .... remove..
Save

[Joe_H]

Welcome to the folding support forum.

As already mentioned, you can remove the CPU folding slot. If it is already processing a WU, it is preferred that you set it to Finish first, then delete the slot.

The GPU folding slot will still use some CPU time for preparing the WU data to be processed on the GPU, and also for creating the periodic checkpoints and validation checks.

[gopinathms]

but still the cpu runs at 50% load taken by fahcore21.exe

i want the app to take the help of cpu sometimes not always

the load is very important to me, im not running a powerhouse , im running a pentium 4 dual core processor with gt630 gpu

if i push the client to light , it always stays idle

[Joe_H]

The CPU core that is required by the nVidia drivers to handle assigning OpenCL commands and WU data can not be avoided. Those who have looked into it further believed based on the evidence they saw that the code has the CPU core in a spin wait for handling transfer of commands and data to and from the GPU. From what they saw, if I understood their explanations, the CPU core is always active, but can handle processing for other processes on the computer.

[silverpulser]

I am running a G620 CPU in conjunction with a NVidia GTX 750 Ti GPU and also only use the GPU for FAH projects and have deleted the CPU slot (only gave me about 1000 points per day). I can see a small loss of performance but not enough to stop me using the computer for not only everyday tasks but also light gaming (Asphalt and Airborne for example).

[bruce]

Whether you use a CPU slot or not is certainly your choice. Personally, I choose to do so. While the points awarded for CPU processing are much less than for GPUs, that's a realistic measure of the amount of folding work being completed -- and IMHO, every bit helps.

Processing by the CPU is specifically designed NOT to interfere with other things you might be doing. The operating system manages CPU priorities and the FAHCore_a* runs at the lowes possible priority so it backs off whenever there's something else for the CPU to do. Unfortunately the GPU has no operating system and ther is no capability to lower the priority for GPU processing. Thus when a FAHCore_1*/2* has assigned work to the GPU, a request from the OS to update the screen has to wait for its turn. The only way to avoid screen lag is to pause all GPU folding until the OS says your system is idle. (Screen lag can be noticeable if you have a slow GPU. Lags introduced by the CPU are only noticeable if you have less than about 1 MB of main ram which is not a problem with current systems.)