Originally posted by: Ken_g6
Or, I imagine, you could launch a BOINC instance for 1 core while the non-BOINC instance is going, and when that WU is done you could up it to 2 cores.
WCG seems to be BOINC compatible, but not exclusively BOINC. I wonder if anyone's benchmarked to see if either way is faster/gets more points?
Yes, it should be possible to run UD-agent alongside BOINC.
WCG added BOINC as an option late 2005, to get Linux-support, but due to users asking for it, they also added windows-application.
As for speed, the UD-agent is as default limited to 60% cpu-usage and needs some form of hack to use more, and as already mentioned doesn't work on multi-cpu-systems.
As for "points", I've never run the UD-agent under WCG, but do remember it's "features" then ran by UD 'think', not sure on whatever they're calling themselves now... The cpu-part of benchmark was only 35% of the total score, disk-size, memory-size and internet-speed was also counted... So, if WCG is using the same, with cpu being 35%, it means 3 slow 1 GHz-computers that each crunches 1 wu/day, will get significantly more "points" than 1 fast 3 GHz-computer that crunches 3 wu/day...
For BOINC-client on the other hand, it doesn't matter which computer you've crunched a specific wu on, you'll get the same "points" regardless of 200 MHz or 4 GHz. But, since the 4 GHz can crunch much more wu/day, you'll also get more "points"/day.
So... if you've got a bunch of p2/p3-systems, the UD-agent will likely give higher points/day than the BOINC-client. But, for fast computers, the BOINC-client will likely give more points/day...
As for the actual application-speed, I don't know of any significant speed-difference between them, so as long as you've applied the "hack" to use 100%, you'll likely do as much useful scientific work regardless of UD or BOINC... But of course, this is only true for single-cpu...