View Single Post
Old 2007-05-13, 09:15   #1
S485122's Avatar
Sep 2006
Brussels, Belgium

26·29 Posts
Default Optimising work for Intel Core 2 Duo or Quad Core

I want to optimise the GIMPS work on my two quadcores QX6700.

Knowing that TF fits completely in L1 cache, would it be right to set the L2 cache to 0 for the core doing TF and assigning the whole 4MB of shared L2 cache to the other core on the same dye and doing LLR tests on it ?

Another thing : I saw in a post somewhere the amount of L2 cache optimal for some FFT size. Since the value given was about 10M exponents (half the size of current double checks) the values given are not usable. My question : one can tune the amount L2 cache attributed to each instance. Given 4MB of L2 cache what would the best split between two instances, one running a double check (20M exponents), the other first time checks 40MB exponents ?

S485122 is offline   Reply With Quote