Message boards : Questions and problems : Windows XP Pro Agonizingly Slow w/BOINC 6.6.36
Message board moderation
Author | Message |
---|---|
Send message Joined: 17 Jul 09 Posts: 2 ![]() |
I am running BOINC 6.6.36. Machine is a Dell Precision T3500, with Intel Zeon W3520 CPU @ 2.67 GHz and 3 GB of RAM. OS is Win XP Pro SP3. When I run BOINC, because the system has eight (8) cores, eight processes start. (Well, actually nine, because the video card is CUDA.) The system becomes painfully slow. Click to open any application (e.g., Word, IE, anything!), and it takes 10 - 15 seconds for the application to start. I have tried changing BOINC's configuration to use only 50% of the CPUs, but I still have long delays, but not quite as bad. If I kill BOINC the system is wickedly fast - no delays at all. What am I doing wrong? Any help would be appreciated. |
Send message Joined: 5 Oct 06 Posts: 5149 ![]() |
I run an earlier Dell Precision 490, also with 8 cores. It started BOINC under Windows XP, and continues under Vista. No problem, though saturation of the quad-channel FB-DIMM memory sub-system means that with the wrong mix of tasks, it doesn't scale well compared to two separate quads. I suspect the slowness is due to your CUDA card. What projects/tasks are you running on it? Some (AQUA, SETI/VLAR) aren't very well behaved at the moment. |
Send message Joined: 17 Jul 09 Posts: 2 ![]() |
That was an excellent guess on your part. When I disabled the use of the GPU and set BOINC to use all 8 cores, rebooted, and started BOINC again, things are nice and snappy. Funny, because I run this at home and my video card is also CUDA, and I have never seen any issues. Hmm... Thanks again for the suggestion. |
Send message Joined: 5 Oct 06 Posts: 5149 ![]() |
Not a guess, bitter experience. As well as the Dell, I run three quads with CUDA cards - two of them are currently in a similar state because of badly-behaved CUDA tasks (AQUA, in these two cases). Anyway, glad the machine is working fast enough for you to finish off the detective work. |
Copyright © 2025 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.