Quote Originally Posted by STMahlberg View Post
Why 4 instances of Boinc?

When I was playing around with Linux, I had paused Boinc on Windows while I installed Boinc on Linux and ran some WEP wu's. When the instance of Boinc on Windows resumed from it's timeout, it locked up my rig or rather slowed it down that it became unresponsive.
The multiple instances of Linux, one it's so I can diversify what I'm running more easily if I'm running mulitple apps/projects. Also it's to dabble with different distributions. I'd like to get a mac os going some day. But mostly it's a simple and cheasy way to get goofygrid credit/hours, or for when WU limits are hit like Scole said. My setup on the 12c/24t t5500 is 3 linux vms all that use 4 threads each. So max threads they can use is half my total. It's my opinion that a lot of WCG projects pay slightly better on linux so what I would do is load up the linux vms first and then added windows threads of wcg until I felt I had the right overall load considering hyperthreading.

It seems that virtual threads are a bit more forgiving than the host's when it comes to CPU time. Like you said locking up. If you had a 4c/8t machine and setup a VM to use 5 threads and then you turned off your hyperthreading and launched the VM i bet it would work totally fine until you overloaded it and it would just bog but probably not crash the host.

In fact 3 of my windows hosts all have linux vms running on them. Right now since I think cosmology CAMB runs better on windows I have them all idled out just running nci. but when things change I'll often run 3threads on the Linux VM while the host runs a gpu.