I just acquired 2 w9100s. The use case for them is a double precision OpenCL application for finding large prime numbers. These were to replace 7990s, which were executing tasks at approximately 1.8% per hour per core. Each core should have 1/3 or less the double-precision performance of a w9100, yet the w9100 only processes tasks at 1.05% per hour, or 41% worse performance. I have tweaked every setting I could find. Using 15.301.2601 on Windows 10/64-bit. They are not overheating, and they are staying at 930mhz/1250mhz core/memory. Running the application pegs utilization at 100%
Anyone have any suggestions for things I can tweak/check/etc?