Quote Originally Posted by famish99 View Post
It is shocking to see how much of a performance gain is caused by switching to AVX, but I suppose it uses the execution units more efficiently than SSE.
What was shocking was that nobody with a SB was getting a performance gain. I tried to find out why this wasn't happening a couple of months back but no one seemed to be interested in finding out why at the time.

The reason for the increase is the new registers that are 256bit and can allow twice the throughput per clock cycle of the older 128bit SSE registers. Expect about a ~80% gain using SB with Linpack and AVX.

I finally got to do my own testing of this. You can see the results here, if interested.

Although SP1 is still beta it shouldn't be too long before it's finalized.