I am running BOINC 6.4.5 on a system with an integrated GeForce 8200 onboard, plus a PCI-express 8500GT with 512MB. BOINC detects two CUDA capable devices (but classifies both as 8200's) but is not using the CUDA version of the application. I have the latest combined drivers from NVidia with CUDA support, and CUDA 2.0 installed, but no joy.
Other pertinent info:
AMD 64X2 5000+ Black Edition
3Gb system ram
Windows XP Pro SP3 32 bit

---

To follow up on my own post, and perhaps shed some light on this issue, I used GPU-Z to check out my GPUs. The integrated GPU was reporting only 64MB of memory. I attempted to increase the frame buffer size in the BIOS to 256MB but this caused Windows XP to crash upon loading. I then disabled the integrated GPU in the BIOS, allowing use of the PCI-E 8500GT only. BOINC detected a single CUDA device, correctly identified the model, and began using it to process.
Now, even though I am attached to other BOINC projects, my CPU is only loaded at 50% (i.e. only one of my two cores is actively processing non-CUDA data). I get the feeling that, although the CUDA workunits are only using ~2-3% of my available CPU cycles, having CUDA active is excluding BOINC from establishing another instance of CPU-based processing.
Any thoughts on how to extract the maximum number of effective cycles from all my computers various processers?

---

cc_config.xml

---

Thanks, that worked a treat :)

---