View Single Post
Old 2005-03-03, 12:38   #3

13×641 Posts

Thanks very much for taking the time to go through the FFT stepping stuff - it's really appreciated, and can give us all confidence that prime95 is pretty much on the limit in terms of FFT related project throughput. I did a quick check of likely benefit to exponent test time and it was about 6% at most (for half the step size at the top end, and that's ignoring the difficulties of point 2).

Something that prompted me to post was that the (SSE2) 34.56 million crossover is rapidly approaching and would mean for me that an exponent would take 4.5 days more (over 100 hours at full speed) for a couple of exponents over the limit as opposed to under - it's something that seems a little crazy at first. Now there's an incentive to grab one of the 10 million digit exponents soon if ever I heard one!
Thanks again
  Reply With Quote