Page 2 of 3

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Tue Oct 20, 2020 6:26 pm
by MeeLee
PantherX wrote:
joncrane wrote:...what's going on? Secondly, is there some way to adjust the relative priorities of the two slots such that I can still get 5 figures of PPD from CPU and get ~2million on the GPU side?...
In addition to what is stated above, when it comes to folding, you can use Process Lasso to lock the CPU (FahCore_a7/FahCore_a8) to certain CPUs and GPU (FahCore_21/FahCore_22) to other CPUs which aren't allocated to the CPU. This would be called CPU affinity which ensures that you can get the maximum out of your system. Keep in mind that it would generate more heat so as long as your cooling system can handle it, keep on folding!
joncrane wrote:...How does one detect thermal throttling of the CPU and/or GPU?...
For GPU, you can use GPU-Z
For CPU, you can use Core Temp

I also use AIDA64 Extreme which has a lot of details displayed as a sidebar on my system.
I think MSI afterburner or EVGA XOC will be better for this, as they also have the means to adjust temperature or power requirements.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 7:12 am
by bruce
Instead of two slots: CPU:12 and CPU:2 I would probably set them to CPU:6 and CPU:8 or :9. Let them run until they both get new assignments before you begin to evaluate the changes.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 7:27 am
by bruce
PantherX wrote:
joncrane wrote:Secondly, is there some way to adjust the relative priorities of the two slots such that ...
In addition to what is stated above, when it comes to folding, you can use Process Lasso to lock the CPU
I have nothing against Process Lasso, but has anybody tried <core-priority v='idle'/> vs. <core-priority v='low'/> recently?
For a long time, it didn't set the FAHCore's priority, it set the FAHCoreWrapper's priority. I haven't checked to see if that was fixed.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 8:09 am
by PantherX
On my Windows system running V7.6.13, I have set Folding Core Priority slightly higher via the GUI (<core-priority v='low'/> in config.xml) and this is what I see in Task Manager:
FAHCoreWrapper -> Priority is Below Normal
FahCore_22 -> Priority is Low

Based on this (https://docs.microsoft.com/en-us/window ... priorities) I would say that Low has a lower priority than Below Normal.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 8:37 am
by bruce
But does it also look the same if you set it to "idle" (or leave it as default)?

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 9:15 am
by PantherX
Good point, so when I used the default which is Lowest possible, this is what I see:
FAHCoreWrapper -> Priority is Low
FahCore_22 -> Priority is Low

Not sure if that's the right behavior or not. My educated guess is that it's a bug since I expect both processes to be the same priority.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 1:11 pm
by mwroggenbuck
I also see this. Changing the priority to something a little higher via the advanced control only changes the wrapper. It does not change the actual core. Effectively, this setting means nothing. However, unless you are running your CPU resources at 100% because of other processes, this should not have any effect. Even at a lower priority, the core will get resources if it does not have to compete with other processes.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Wed Oct 21, 2020 10:13 pm
by MeeLee
mwroggenbuck wrote:I also see this. Changing the priority to something a little higher via the advanced control only changes the wrapper. It does not change the actual core. Effectively, this setting means nothing. However, unless you are running your CPU resources at 100% because of other processes, this should not have any effect. Even at a lower priority, the core will get resources if it does not have to compete with other processes.
I wouldn't say it would do nothing.
If it's set to 'low' or 'idle', it would be in the same category as the resources assigned to Windows' background tasks (like defrags/trims, auto updates, etc...).
Setting it to normal, or below normal, might give it an edge, and prevent Windows' background services to interrupt the process (at least to a lesser amount)
Image

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Thu Oct 22, 2020 12:19 am
by mwroggenbuck
I understand what you are saying, but note that I specified if the CPU resources were not at 100% there would be no competition and therefor no real change. If you have defrags/trims/etc running, you would have 100% of CPU resources used, and the low priority would have an effect because the core would then have to share.

Check the task manger for the cpu load of the system idle process. If that idle process is more than zero, bumping up the priority of a different process should not really have any effect.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Thu Oct 22, 2020 8:33 am
by bruce
And remember, higher priority doesn't mean it runs faster. A nigher priority task that consumes large quantities of resourses simply means that there are fewer resources available to whatever has a lower priority. If you're okay with shutting out some other process, why not just stop it?

FAH is designed to use all unused resources.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Thu Oct 22, 2020 11:59 am
by mwroggenbuck
FAH is designed to use all unused resources.
And then some more... :wink:

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Fri Oct 23, 2020 4:08 pm
by joncrane
bruce wrote:Instead of two slots: CPU:12 and CPU:2 I would probably set them to CPU:6 and CPU:8 or :9. Let them run until they both get new assignments before you begin to evaluate the changes.
This is an interesting idea. Can you please elaborate as to why split it up like this?

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Sat Oct 24, 2020 8:45 am
by MeeLee
joncrane wrote:
bruce wrote:Instead of two slots: CPU:12 and CPU:2 I would probably set them to CPU:6 and CPU:8 or :9. Let them run until they both get new assignments before you begin to evaluate the changes.
This is an interesting idea. Can you please elaborate as to why split it up like this?
The 12 + 2 core will get a lot more PPD than an 8 + 6 core, when fully used.
But an 8+6 core is more evenly split.

I would presume there are less 12 threaded WUs (they usually combine 8+4 or 6+6, or 4+4+4 or so) than there are 6 or 8 threaded WUs.
Another benefit is, if you have a 8 threaded chiplet (4 core), the WU will be processed faster, as it'll run from a single L3 cache bank.
But if you have 12 threaded WUs (despite them getting more PPDs, due to the higher core count), it'll have to pull from 2 L3 cache banks, causing some slowdown in the work.
So despite your WUs being processed faster, you'll probably still see lower PPD scores.

It's an issue on 'older' 3000 series Ryzens, but newer 5000 series Ryzens don't have that issue as they split L3 cache between 2 chiplets (8 cores, 16 threads max I believe).

I'm just speculating here...

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Sat Oct 24, 2020 4:12 pm
by Joe_H
joncrane wrote:
bruce wrote:Instead of two slots: CPU:12 and CPU:2 I would probably set them to CPU:6 and CPU:8 or :9. Let them run until they both get new assignments before you begin to evaluate the changes.
This is an interesting idea. Can you please elaborate as to why split it up like this?
Not bruce, but my own thoughts on this would be to try a single CPU slot of 8 threads. My reasoning is based on your i7-10875H processor having 8 cores which support 16 threads through HT. When the processor is actively using the extra hardware in the processor for the second thread per core, it will use more power resulting in more heat output. In a laptop this is going to lead to thermal throttling and some loss in performance.

The ability to run at a higher clock on 8 threads may be enough to make up the difference from using more threads through HT. It might take some expermenting to see if there is a number of threads that does result in a higher PPD.

Re: Running CPU + GPU results in lower PPD than GPU alone?

Posted: Sat Oct 24, 2020 5:29 pm
by bruce
The reason I used the word "might" is that every project has a range of assignable CPU-counts. You cannot predict that there will always be WUs that can use all 12 of your 12-cpu slot. Suppose at the moment that slot seeks work, the available projects run on <=8 threads. The assgnment will be made and the other 4 threads will be idle until that WU has been processed.