Lately I noticed that the SACM is using about 5% of the CPU (I5-6600) which is high, I read about it online but no practical explanation or solution was offered.

The machine runs BOINC 24/7 and since I last joined Primegrid the GPU (GTX 960) is used as well.

The machine have 8GB of RAM and the GPU 4GB, I never saw the RAM used more than 3.5GB and the GPU RAM is used less than 500MB,so there is no memory pressure and there is no need to compress anything

I tested several suggestions I found online which didn’t help and finally I thought about something else which in my case did.

When I started to run the low GFNs (15-17) the GPU was not fully loaded (~80%) so I allowed two WUs to run concurrently thinking that the combined elapse time will be shorter, it did but the SACM rose. I changed the app_config back to allow only one WU to run at a time and reread it during the run, the SACM went down immediately, I tried it several times and the results were the same. So now I run only one GPU WU.

My conclusion is that when the GPU is fully loaded there is high data traffic between the main RAM and the GPU RAM that somehow triggers the SACM to work hard even when there is no need for it.

Your take on the subject will be appreciated.