anyone else get -worse- performance on 64bit linux by compiling the sources yourself for linux? If i compile the source and run the linux ELF binary i get about 4500 pps, if I run version 7.1 (need to update so i get the right share value shown though) through 64-bit wine I get about 7200 pps.... although with the linux version it actually takes about 5 minutes to "auto-configure" the sieving while the windows version doesn't say anything about this to me, that is the configuring part, it does say that auto-config is on.
Just curious...
Just curious...
You should probably check 6, preferably 7+ chains per hour over a couple of hours for a descent comparison. Primes per second has been proven to be an inaccurate performance index numerous times
You can also use 4 and 5 chains per hour to determine possibilities for higher chains.
For instance, one of my machines with an AMD Phenom II X4 910 pulls about 5195.19 4ch/h and 499.58 5ch/h.
Using the following maths, you can get a decent idea of higher chains:
Code:
5195.19 / 499.58 = 10.399 (truncated for ease)
6ch.h: 499.58 / 10.399 = 48.04 (41.90 is my real measurement)
7ch: 48.04 / 10.399 = 4.62 (3.95 is real number)
8ch: 4.62 / 10.399 = 0.44 (0.62 is real number)
and so on
6ch.h: 499.58 / 10.399 = 48.04 (41.90 is my real measurement)
7ch: 48.04 / 10.399 = 4.62 (3.95 is real number)
8ch: 4.62 / 10.399 = 0.44 (0.62 is real number)
and so on