![]() |
|
|
#1 |
|
Feb 2017
158 Posts |
I have a question about the software. I was wondering if it makes more use of FP64 or FP32, i.e, double precision or single precision operations. I am thinking of acquiring a graphics card and I would like to take into account the needs of prime95.
Thank you. |
|
|
|
|
|
#2 |
|
"/X\(‘-‘)/X\"
Jan 2013
https://pedan.tech/
1100011100002 Posts |
Prime95 doesn't use a graphics card, but performs best with Intel CPUs supporting AVX2 and the all the memory channels populated with the fastest dual-rank memory supported by the CPU.
|
|
|
|
|
|
#3 |
|
Feb 2017
11012 Posts |
Ok. And what about Trial Factoring on a GPU? Does it make a difference in that case?
|
|
|
|
|
|
#4 |
|
"Curtis"
Feb 2005
Riverside, CA
133368 Posts |
The version of the software that does use a GPU, CUDALucas, uses DP operations for the lucas-lehmer test.
|
|
|
|
|
|
#5 |
|
Just call me Henry
"David"
Sep 2007
Liverpool (GMT/BST)
3·23·89 Posts |
|
|
|
|
|
|
#6 |
|
"Oliver"
Mar 2005
Germany
5·223 Posts |
TF on GPU (mfakt[co]) is using 32bit (unsigned) integers most of the time.
Last fiddled with by TheJudger on 2017-02-11 at 01:06 |
|
|
|