Set BOINC to identify 2 GPU with different priorities..?

When a recent BOINC client finds two CUDA devices that differ in capabilities, it will ignore the less powerful one completely (message says 'not used'). Older clients will treat all CUDA devices equally and report the properties of device #0.

Different settings for different CUDA devices aren't provided for in the standard BOINC client. You are, however, free to implement them and compile your own client.

BM