50% GPU Utilization

[Nope.avi]

I am currently folding for project 11747 on a Geforce GTX 750Ti, and the GPU utilization is only going up to 50%, and I have the folding power at Full.

[alxbelu]

If you're checking in windows task manager, switch view from 3D to "CUDA", "Compute" or similar (I had multiple views that I had to switch between).

Another option is to use e.g. the sensor tab in GPU-Z; https://www.techpowerup.com/gpuz/

[tsunami2311]

I wish I could have F@H use only 50% of GPU, I dont mind my GPU 1070ti running when it idle to do work (medium) but I dont want it running full tilt either that just means it gona 70c for hours on end and fan will be in overdrive and that fact F@H is only thing that cause my GPU to have coil whine above 75% power confusing to me ( probably cause it using diffrent parts gpu then when it gaming) I could full load/power in games and not single whine from the gpu, I been using MSI AB to limit the power to gpu to 50% to keep temps in check but does nothing to keep fan from being in overdrive. I would prefer it if client had way to limit power/usage of gpu.

Come the summer month folding really isnt gona be doable less I want to die of heat

[Joe_H]

For using less power on your GPU, there are utilities like MSI Afterburner where you can set a power limit. That slows down the GPU clock to keep within that limit, that results in less heat production and lower fan speeds. It will show high GPU utilization, but things run a bit slower.

Your can fine tune the CPU utilization within FAHControl as well to use the exact number of cores you want to further keep a handle on power dissipation.

[tsunami2311]

I know I am using MSI AB to limit the GPU lower to 50% power limit which is as low as it goes, it keeps the temperature in check it no longer pushing 70c , but the fan is still to loud, I was hope F@H could limit the usage.

I know there is "gpu-usage" command it the V7 list but it does not work and restarting the client with that set just results in the command being removed on restart of client in theory if it works it would do what I wanted which is limit the usage to 50% which would also keep fan speed down

as for Cpu Folding my CPU(6700k) fan (noctua D-14) are at locked speeds of 1000 rpm so noise is always the same which silent, cpu temps seem to stabilize at 67c regardless with 3-6 thread if i use 1~2 thread is drops to 65c so there not much I can tweak there. The gpu fan is what is loud 40% fan speed is noticeable anything above 45% is annoying. I know the GPU is pretty much silent @50% usage once it pushing 65-70% usage it get loud.

[bruce]

If your system cannot tolerate the GPU folding at 50% plus the CPUs folding at a reduced rate, perhaps your cooling system is inadequate or the heatsink needs to be re-greased. Otherwise, you can choose to fold only with the GPU (plus 1 CPU thread) or only with the CPU. There's no doubt that folding uses lots of resources and busy resources generate heat.

[tsunami2311]

it can handle the temps it just I want to keep the fans speed of the gpu in check its GPU can hand 75c fine they actual build to run those temps it will never get that high, but fan gets loud once gpu is past 50% usage and temps push 60c The temp range for gpu right for it usage it just it get loud, which why I was hoping "gpu-usage" command actual worked. As for CPU temps only other thing to push temps like that on my CPU was prime95 it almost never over 60c other wise.

like said it not matter of cooling isnt adequate cause these are all normal temp operations for my gpu model and D-14 HSF on my CPU. I just want to keep my GPU temps and usage in check to keep the fan from getting to loud. if the client cant control "gpu-usage" like the extra command says it does or should, then that is that I just deal with it till it does.

I basicly only let gpu run anyway seeing temps on cpu are beyond what i like even if it normal, GPU temps are where they should be for it usage I just preferred I could reduce the usage to about 50% which keep it silent fans.

[alxbelu]

If you have thermal headroom left on the GPU, can't you just change the GPU fan curve? Sounds like it's potentially overly aggressive for your preference?

[lostcowboy]

about core 22 for the GPU, it looks like it is disabling temperature protection? Not sure if that is a good thing.

Code: Select all

07:09:38:WU01:FS01:0x22:        CPU: AMD Ryzen 5 3600 6-Core Processor
07:09:38:WU01:FS01:0x22:     CPU ID: AuthenticAMD Family 23 Model 113 Stepping 0
07:09:38:WU01:FS01:0x22:       CPUs: 12
07:09:38:WU01:FS01:0x22:     Memory: 31.93GiB
07:09:38:WU01:FS01:0x22:Free Memory: 23.69GiB
07:09:38:WU01:FS01:0x22:    Threads: WINDOWS_THREADS
07:09:38:WU01:FS01:0x22: OS Version: 6.2
07:09:38:WU01:FS01:0x22:Has Battery: false
07:09:38:WU01:FS01:0x22: On Battery: false
07:09:38:WU01:FS01:0x22: UTC Offset: -5
07:09:38:WU01:FS01:0x22:        PID: 33744
07:09:38:WU01:FS01:0x22:        CWD: C:\Users\Clifford\AppData\Roaming\FAHClient\work
07:09:38:WU01:FS01:0x22:         OS: Windows 10 Pro
07:09:38:WU01:FS01:0x22:    OS Arch: AMD64
07:09:38:WU01:FS01:0x22:********************************************************************************
07:09:38:WU01:FS01:0x22:Project: 11748 (Run 0, Clone 31, Gen 5)
07:09:38:WU01:FS01:0x22:Unit: 0x000000138ca304e75e6a7fdfc5bb3715
07:09:38:WU01:FS01:0x22:Reading tar file core.xml
07:09:38:WU01:FS01:0x22:Reading tar file integrator.xml
07:09:38:WU01:FS01:0x22:Reading tar file state.xml
07:09:39:WU01:FS01:0x22:Reading tar file system.xml
07:09:39:WU01:FS01:0x22:Digital signatures verified
07:09:39:WU01:FS01:0x22:Folding@home GPU Core22 Folding@home Core
07:09:39:WU01:FS01:0x22:Version 0.0.2
07:09:45:WU01:FS01:0x22:Completed 0 out of 2000000 steps (0%)
07:09:45:WU01:FS01:0x22:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
07:10:18:WU01:FS01:0x22:Completed 20000 out of 2000000 steps (1%)

[Joe_H]

That is a leftover message from an attempt to keep GPUs form overheating from within the folding client, as a last ditch protection. It never worked well, and is slated for removal. It does not disable normal temperature protection coded into the GPU card firmware and the driver, and which has gotten much better than when this was added.

[bruce]

The temperature protection that's being disabled is inside of FAH. The right place to do temperature protection is inside of the driver. Any application, including FAH, isn't able to do it right.

[WhiteHorsePT]

I have the same issue with my GTX970. It's only using 50% of its processing capacity, even when the folding power is set to full.
On the Windows Task Manager GPU view, I can see that the job is only being processed by the Compute_0 (100%) but not by the Compute_1 (0%).

Do I need to create an extra slot in FAHControl to have it used?

[Joe_H]

Do I need to create an extra slot in FAHControl to have it used?

No, just the way Task Manager works it shows different aspects of the load on the GPU under "cryptic" Compute n designations. Other versions of Task Manager have an explicit CUDA entry. The part of your card used by F@h is being fully utilized.

[WhiteHorsePT]

Do I need to create an extra slot in FAHControl to have it used?

No, just the way Task Manager works it shows different aspects of the load on the GPU under "cryptic" Compute n designations. Other versions of Task Manager have an explicit CUDA entry. The part of your card used by F@h is being fully utilized.

Roger that. Thanks for clarifying.

[r0k0]

If you're checking in windows task manager, switch view from 3D to "CUDA", "Compute" or similar (I had multiple views that I had to switch between).

Another option is to use e.g. the sensor tab in GPU-Z; https://www.techpowerup.com/gpuz/

i have gpu-z but in the sensor tab, what do i look for ?

my 3 gpus ( Quadro K4000 - GTX 1060 6GB - 5700XT ) only run at 1-5% utilization but temps are high on all 3 when folding