Is there a tool to find out how many stream processors are in use by a specific application?
AFAIK the GPU always spreads the work across as many SPs as possible, even for a single triangle, so the result of a "how many SPs are being used ?" question would generally be "all" or "none".
You can get a bit more information on how this all works in the "R6xx-R7xx-3D.pdf" document at :
Generally the shader pipe is either "full" or "empty". When it's "full", the only reason for SPs being unused would be that there weren't enough pixels in a given triangle, or vertices in a vertex buffer, to completely fill each of the SIMDs.
Retrieving data ...