[QUOTE=Prime95;310759]It might be helpful if people would publish their optimal GPUSievePrimes values for testing various Fermat numbers and double-Mersennes. To get the best throughput, it is important to get this parameter set somewhere in the vicinity of optimal.[/QUOTE]
For my assignment: MMFactor=127,179,2.5e15

These settings work best on a 580 (GPU usage @ 99%):

GPUSievePrimes=816000
GPUSieveSize=128
GPUSieveProcessSize=16

Just for reference, about how long [I]should[/I] this take (so I know if I'm on track or not?)

[QUOTE=rcv;310837]... I suggest you tune the h/o digit of GPUSievePrimes (800000 or 700000 or 900000). Then tune the second digit, to max out performance (810000 or 820000 or 830000).[/QUOTE]This is a perfect job for the software to handle for itself. How about making the code do this automatically?

[QUOTE=flashjh;310841]Just for reference, about how long [I]should[/I] this take (so I know if I'm on track or not?)[/QUOTE]
About 3.5 days, more or less. (17.2 * 2/9 * 15/16 * 0.98 = 3.51. 17.2 days for my assignment of size 1.125e15; your assignment is about 2/9 the size of mine; your card has 16/15 as many cores as mine; George improved the code by about 2%.) Perhaps a little less, based on clock speed of your card.

Am I interpreting the table correctly that
MMFactor=89,6.9e12,100e12
would be a reasonable, open reservation? It looks like it will take about 24 hours on my GTX 480.

Edit: Considering how relatively little work it is now, I will double-check the lower ranges as well. The lowest that doesn't give an error is 128, so I'll do
MMFactor=89,128,100e12

[QUOTE=frmky;310847]Am I interpreting the table correctly that
MMFactor=89,6.9e12,100e12
would be a reasonable, open reservation? It looks like it will take about 24 hours on my GTX 480.

Edit: Considering how relatively little work it is now, I will double-check the lower ranges as well. The lowest that doesn't give an error is 128, so I'll do
MMFactor=89,128,100e12[/QUOTE]

Thank you for your reservation.

Would you like to apepar as "frmky" or with your name?

Luigi

New subforum?
 

We are messing up this thread :smile:

May I propose the opening of a subforum (like "Double Mersennes") to manage information related to reservation and completion ranges, benchmarks, tables and stuff like that, leaving this thread to the development of mmff?

Luigi

[QUOTE=ET_;310850]Thank you for your reservation.

Would you like to apepar as "frmky" or with your name?

Luigi[/QUOTE]
Let's use my real name.

Arg!
Well, forget my earlier reservation, I have to rerun it from the start and i'm going away tommorow morning.
I can re -run FermatFactor=46,300e12,325e12 with GPUSieveProcessSize =8 .

[QUOTE=firejuggler;310853]Arg!
Well, forget my earlier reservation, I have to rerun it from the start and i'm going away tommorow morning.
I can re -run FermatFactor=46,300e12,325e12 with GPUSieveProcessSize =8 .[/QUOTE]

I will keep both your reservations active, so you'll know what to run as you come back.

Luigi

I'll also take
MMFactor=107,140,100e12

[QUOTE=rcv;310844]About 3.5 days, more or less. (17.2 * 2/9 * 15/16 * 0.98 = 3.51. 17.2 days for my assignment of size 1.125e15; your assignment is about 2/9 the size of mine; your card has 16/15 as many cores as mine; George improved the code by about 2%.) Perhaps a little less, based on clock speed of your card.[/QUOTE]
Very good calculation. Once I got it tuned it was 3 days 5 hours (3.2 days).