P-1 factoring anyone?

[cheesehead]

Originally Posted by cheesehead

OTOH, once GPUs are doing L-L, the GPU:TF/GPU:LL tradeoff point changes again, perhaps back to near the CPU:TF/CPU:LL tradeoff.

Then there's the GPU:TF/CPU:P-1/GPU:LL tradeoff and the GPU:TF/GPU:P-1/CPU:LL tradeoff and the GPU:TF/GPU:P-1/GPU:LL tradeoff, plus the GPU:TF/CPU:P-1/GPU:TF/GPU:LL tradeoff and the GPU:TF/CPU:P-1/CPU:TF/GPU:LL tradeoff and the ...

... also, remembering that the CPU-feeding-GPU handling one assignment (say, not-quite-final TF) might not be on the same system as the CPU-alone handling a subsequent assignment (say, P-1, or final TF) or the CPU-feeding-GPU handling the next assignment in the cycle (say, final TF, or L-L) ... !

[TheJudger]

Originally Posted by James Heinrich

Which is 16GHz-days of work, which means roughly 64GHz-days/day throughput.

64GHz-days/day seems a little bit low to me for a GTX 460. My stock GTX 470 can do 180-200 GHz-days/day using 3 i7-cores at 3.5GHz (measured at 100M digit numbers to 2^79).

Oliver

[firejuggler]

Don't worry, it is with 1 instance of mfaktc.
What i find straneg is that one core of my E8300 seem to be enough to 'feed' my GPU.
If I use 2 instance of Mfaktc, I should have about 85 to 90 Ghz/day

[James Heinrich]

http://v5www.mersenne.org/report_top_500_P-1/
Why are there several people on this report who have completed non-integer numbers of P-1 tests? How do you attempt 138.587 tests, for example?

[ATH]

138.587 is total Ghz-days, Attempts is 42. It's shifted to the right due to the long account name.

[James Heinrich]

Originally Posted by ATH

Attempts is 42

I feel dumb now. (I was sure I'd seen it before and it was actually wrong, but it was probably me that was actually wrong )

[nucleon]

Originally Posted by James Heinrich

Which is 16GHz-days of work, which means roughly 64GHz-days/day throughput.

Originally Posted by TheJudger

64GHz-days/day seems a little bit low to me for a GTX 460.

It really depends on the cpu that's a single core run.

My GTX460 with a i7-930 can do 60-100GHz-days/day depending on the assignment for one core, 2x cores yield an additional 20-30% on top of that.

-- Craig

[firejuggler]

I have a 'slow' core 2 duo E8300 @ 2.833 GHz

[drh]

I have a few cores working on P-1's, and I wanted to assign 1 of them to work on the smaller P1's, like the 80 remaining in the 40M range, if I'm reading that right, but when I ask for a manual assignment, it says that it can't find any in that range and assigns me a DC. Just curious if I'm doing something wrong, as it's my first time trying manual assignments, or if I'm reading the activity summary wrong.

Thanks,
Doug

[cheesehead]

Originally Posted by drh

I have a few cores working on P-1's, and I wanted to assign 1 of them to work on the smaller P1's, like the 80 remaining in the 40M range, if I'm reading that right, but when I ask for a manual assignment, it says that it can't find any in that range and assigns me a DC. Just curious if I'm doing something wrong, as it's my first time trying manual assignments, or if I'm reading the activity summary wrong.

I don't think you're doing or reading anything wrong.

As double-checks, (1) I tried the same manual assignment request you did and got the same response, and (2) I got two factoring limits "simple text" reports: both on the exponent range 40M-41M and how-far-factored = 0-999, but one with and one without "Exclude currently assigned exponents" checked.

The one that excluded currently assigned exponents showed 82 exponents on which no P-1 had been performed yet. The other showed 341, implying that 259 of the P-1-less exponents were assigned for some type of work. I don't know how many of the latter are specifically P-1 assignments, though. (I also didn't note whether all those reports were within the same clock hour. PrimeNet updates its reports on the hour.)

George sometimes manually reserves certain ranges of exponents for special purposes, such as someone's custom research or testing that doesn't fit within PrimeNet's ordinary assignment parameters. Such reservations aren't necessarily always correctly reflected in PrimeNet reports. Perhaps those 80-82 exponents are part of such special reservations.

And then, there are just plain PrimeNet bugs.

[James Heinrich]

Originally Posted by cheesehead

George sometimes manually reserves certain ranges of exponents for special purposes

I'd recently come across a number of poorly-done P-1 in the M400xxxxx range, but if they're marked as "special" then PrimeNet won't give it to you, you'll get an error like this:

Registering assignment: P-1 M4000xxxx
PrimeNet error 40: No assignment
ra: Assignment is not available to register

If you're stubborn, and do it anyways, PrimeNet will accept the results once done.