F1157
 

[CODE]prp80 factor: 44415938262041215702318054838369314647679685051653873934029419991814640764656633
prp83 factor: 94487321006964632663708158812394580118141100140621757422433745399029596576822996209
[/CODE]

[QUOTE=Xyzzy;312299]F1157 and F1629 are faster with three cores on the i5. Too bad msieve doesn't test the number of cores automatically to find the fastest combo.

:mike:[/QUOTE]

But your i5 is a SB so later on I will try the 3 cores vs 4 cores on my core i5 750 (Lynnfield).

[QUOTE]But your i5 is a SB so later on I will try the 3 cores vs 4 cores on my core i5 750 (Lynnfield).[/QUOTE]If we knew of a way to sell our SB i5 CPUs for $100 each, we would probably upgrade them to IB i5 CPUs. We think our motherboards support IB.

The problem is that our SB i5 CPUs are locked so we doubt anyone would want them, let alone four of them. We don't overclock so even though the unlocked CPUs were only $10 or so more we did not buy them.

:mike:

F1629
 

[CODE]prp57 factor: 135145396313724358197471673375540780861715137726779408833
prp115 factor: 1312307691852108888263004879113290403686483944586306158672806288593295877138617295097601885092249383260638647261437[/CODE]

Is it safe to move work from one computer to another if the SVN number is different for msieve? We have SVN 755 on our i7 and SVN 761 on our i5. Also, the i7 runs Ubuntu and the i5 runs Debian.

We were thinking of speeding up 8647_61_minus1 a bit.

:max:

Yes, there have been no recent changes in the linear algebra.

np_111
 

[CODE]prp60 factor: 737604601177785566870104021969416651879315010063048887123619
prp107 factor: 19936796350084025530513708564000401800321137575077879987980006370789897353948558413914001028027096520371417[/CODE]

[QUOTE=pinhodecarlos;312499]But your i5 is a SB so later on I will try the 3 cores vs 4 cores on my core i5 750 (Lynnfield).[/QUOTE]

2.5M^2 matrix on core i5 750 (Lynnfield):

commencing Lanczos iteration (4 threads)
Sun Sep 23 22:01:47 2012 memory use: 635.0 MB
Sun Sep 23 22:02:01 2012 linear algebra at 0.1%, ETA 5h57m

commencing Lanczos iteration (3 threads)
Sun Sep 23 22:36:15 2012 memory use: 615.8 MB
Sun Sep 23 22:36:29 2012 linear algebra at 0.1%, ETA 6h27m

Why for Mike's i5 SB machine it is faster to oo the LA step on 3 instead of 4 cores? On my i5 Lynnfield it was the opposite but I prefer this way for at least 29 LP bits tasks.

Difference in the number of memory controllers? Difference in L2/L3 cache size?

That's a long queue to crunch with lasieved application...nice crunch power happening on NFS@Home.:bow:
Also I would like to take the opportunity to thank Greg for distributing GCW numbers to my machines only. I really appreciate the kindness.