I have 3x4850's and 1x6950 in a quad core amd rig. Ccc 12.3 and Boinc Manager 7.0.25. While running BM 6.12.34 my CPU usage was .05, very light. After upgrading to BM 7.0.25 my CPU usage jumped to almost a full core per card/wu. This ties up all 4 cores of my quad so I can't crunch any cpu projects. Anyone else seeing this? Is there a setting I'm missing?

I have 3x4850's and 1x6950 in a quad core amd rig. Ccc 12.3 and Boinc Manager 7.0.25. While running BM 6.12.34 my CPU usage was .05, very light. After upgrading to BM 7.0.25 my CPU usage jumped to almost a full core per card/wu. This ties up all 4 cores of my quad so I can't crunch any cpu projects. Anyone else seeing this? Is there a setting I'm missing?

Probably not but it is typical of where AMD is trying to go with their processing, Nvidia has been that way for a long time now and AMD is starting to catch up. As you are aware gpu's put out over 10 times the credit a cpu does, IF you can feed it enough work at the RIGHT time, Nvidia has it figured out and is producing more credits than AMD using the same software, AMD is now starting to catch back up again. Most hard core gpu users dedicate a cpu to each gpu and MUST do it if they want to get max work out of their Nvidia card. I personally do it when I see what you do, the cpu usage go up over 0.5% on a work unit. Dropping back to version 11.9 of the AMD software may drop that but I am not sure of it, it could be the Boinc software contributing.

Have a look at this thread http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2779 there is some mor einfo on there. Personally I still can't fix this issue on my 6990s. I tried fiddling with polling mode and wait factor but i made no difference except to reduce GPU usage and make WUs take longer. I am running catalyist 11.9 on this system and while the 12.x series drivers get rid of the high CPU usage, I get constant BSOD with them so they are no good. 12.3 drivers seem fine on my 7970s though and I get normal cpu usage and no BSOD.

Similar thing here: after updating to BOINC 7.0.25, CPU usage for MilkyWay jumped from 0.05 to 0.972. Running 2 OpenCL MilkyWay tasks on 2 HD5870 ties up one (of 12) logical core on my i7-990x CPU. That core is mostly idle, the only activity observed through Task Manager is when new MW workunit is loaded for processing. Other CPU cores are running Einstein@home 24/7.

MW output has slightly increased since runtimes are now 2-3 seconds shorter - it seems that one free logical core is enough to speed up two MW OpenCL tasks (compared to my previous BOINC 6.12.34 configuration when all 12 cores were running Einstein@home simultaneously with two MW OpenCL tasks on my HD5870s).

With Collatz, things are same as before - CPU usage is still 0.01, all 12 cores are free for Einstein@home.