Installed just EVGA GTX 680 as a replacement for 560Ti. Upgraded drivers to 306.23 from 280 - 285. I cannot immediately say that 680 is significantly faster than the 560Ti. The power consumption is down by some 40W. Running two WU in parallel. GPU Load is around 85% Memory consumption 623MB.
A WU takes around 15minutes to complete.
Read Jason's page and try some of the tricks, I am doing 15 minute WU's with 3 in parallel.
http://setiathome.berkeley.edu/view_profile.php?userid=8534984
GTX680 specs:
Added 4th WU. Now Load is around 96% and mem use 1125MB.
Task completion in the range of 15-19 minutes.
GPU 75 C.
Fan 51% / 1890 RPM
Power Consumption around 70% TDP
VDDC 1.1620 V.
GPU Core 1071.3MHz
GPU Memory 1502.3MHz
I am amazed of how the GTX560Ti was able to perform on SETI compared to GTX680. The 560Ti core was 900MHz and Shaders 1800MHz. Which is actually doing the FFT? Core or shader? Wy will the GTX680 not scale better? Are there capabilities for concurrent crunching in the optimized applications? How are the calculations loaded onto the cores / (shaders? if applicable)?