I'm running W10 x64 computers (4 core CPU's, no hyperthreading) with BOINC V7.20.2 with Einstein and Milkyway projects with both CPU & GPU w/u's. With varying frequency Einstein throws some w/u's at me with short deadlines, causing BOINC to go into High Priority mode. They are usually CPU w/u's. When this happens, BOINC does not stop other CPU w/u's, it just starts the high priority ones on top of the other w/u's. For example, I had 2 Milkyway GPU w/u's running @ 0.5C CPU per w/u and 1 Milkyway CPU N Body 4 core w/u running at the same time. This was ok, but then BOINC also started running 2 "High Priority" Einstein CPU w/u's which causes everything to slow down.
Any ideas on what I can do to stop this?

---

Set a shorter cache size.

Set a shorter cache size.

I don't know how much lower I should go. Both projects are set to store 0.7 days with an additional 0.3 days for a total of 1 day of work for each of the two projects. What cache size do you recommend? I did have it set higher, but I lowered it to the current settings about a month ago just for this reason. It did get better with the new smaller cache size. Is there any way of getting the High Priority mode to work better in the program?

---

Reduce both settings even further, to say 0.5 and 0.1 days. As you've found that reducing these has improved the situation somewhat this will probably improve it further. Also, if these computers are being used for other work then consider reduce the "use at most x% of the CPUs to leave a core for other duties - in the case of your 4-core systems this would be 75%.

Reduce both settings even further, to say 0.5 and 0.1 days. As you've found that reducing these has improved the situation somewhat this will probably improve it further. Also, if these computers are being used for other work then consider reduce the "use at most x% of the CPUs to leave a core for other duties - in the case of your 4-core systems this would be 75%.

I changed the settings to 0.5 and 0.2 days. Lets see how that works out. I've been doing this for a while now, 22 years! I remember the early days @ SETI when the system up time wasn't as reliable as it is now, so you needed a larger cache size to get you past the down times. :-)

The computes are used mostly for BOINC crunching. On the occasion I do other heavy work, I have the configuration file set to pause BOINC if the CPU load for non BOINC usage exceeds 40%

Does anyone happen to know if there is a way to pause just 1 CPU core if the non BOINC load exceeds a certain percentage?

---

Does anyone happen to know if there is a way to pause just 1 CPU core if the non BOINC load exceeds a certain percentage?

Sadly that isn't an available option even by manually editing _config files. It has been suggested a number of times but I am not aware of any of the devs ever taking it on to look at.

Well, I guess I will keep plugging along and tweak my cache size if I still have issues with this.
I want to thank everyone for their help in this!

---

Does anyone happen to know if there is a way to pause just 1 CPU core if the non BOINC load exceeds a certain percentage?

I think this will be possible with the next client version 7.22, if I understand the release notes correctly.

I think this will be possible with the next client version 7.22, if I understand the release notes correctly.

Looks like you are right! I am running 7.21.0 compiled from the master at git-hub. Not having played with computing preferences for a while I hadn't noticed that option appearing.

I think this will be possible with the next client version 7.22, if I understand the release notes correctly.

Looks like you are right! I am running 7.21.0 compiled from the master at git-hub. Not having played with computing preferences for a while I hadn't noticed that option appearing.

I suppose having found I have the feature, I should probably test it.

Wait out.

Edit: Changing the number of CPUs to use when computer is not in use, seems to make no difference.
Reducing the number for when computer is in use seems to have the same effect as the old, "use at most n% of CPUs. I will look for discussion on git-hub.

Another issue with it is, you can not define what activity means computer is in use unless you tick, "Suspend when computer is in use."