fond of a factor? Urn yourself to become remains

[axn]

Originally Posted by storm5510

My son has a Ryzen 5 3600. 6 cores / 12 threads. Why only one core?

1 core per worker, 6 workers, all running P-1.

EDIT:- What Viliam said.

[storm5510]

Originally Posted by axn

1 core per worker, 6 workers, all running P-1.

EDIT:- What Viliam said.

Ah, I get it now. I have done something similar with Prime95 in the past, but not to this extent. 2 workers using 2 cores each on an i7. IIRC, each worker will use the memory setting in local.txt. In my case, it was 4,096. 8,192 for both.

[petrw1]

28385759
Factor: 4728836973118701446897 / TF: 72-73*
72.002 Bits!!!

[slandrum]

Originally Posted by petrw1

28385759
Factor: 4728836973118701446897 / TF: 72-73*
72.002 Bits!!!

I'm guessing that was worth about 0.001 GHz-days

[firejuggler]

M27142097 has a 174.143-bit (53-digit) composite (P27+P27) factor: 26436212642162007070517822654531755460254859231089553 (P-1,B1=1350000,B2=109350000)

[James Heinrich]

Originally Posted by slandrum

I'm guessing that was worth about 0.001 GHz-days

Mfaktc (and most factoring programs) don't look for factors in order of bitsize but by class spread throughout the target range. In this case the factor is found in class 3412 (of 4620) so it took ~74% of the full-range effort to find even though it happened to be right at the bottom of the range.

[petrw1]

Originally Posted by slandrum

I'm guessing that was worth about 0.001 GHz-days

0.0666 (Diabolical)

Originally Posted by James Heinrich

Mfaktc (and most factoring programs) don't look for factors in order of bitsize but by class spread throughout the target range. In this case the factor is found in class 3412 (of 4620) so it took ~74% of the full-range effort to find even though it happened to be right at the bottom of the range.

74% of the work for 0.2% of the credit.
That's ok sometimes it's the other way around.

[James Heinrich]

Originally Posted by petrw1

74% of the work for 0.2% of the credit.
That's ok sometimes it's the other way around.

I couldn't remember (and didn't bother to look up) if mersenne.org gives proper class-based credit or just takes the lazy approach, apparently it's the latter. For an individual assignment it's unfair, but on average as you say it all balances out.

[LaurV]

IIRC, I asked this in the past, as I got some minuscule credit, close to zero, for a factor and I couldn't figure out why, as it was not on one of the first classes, mfaktc still had to do some work to get the factor. I think George replied something along the lines that PrimeNet gives TF credit class-based, except that the classes are P95-like, and not mfaktc-like. P95 splits the classes differently, whatever that means, maybe he can clarify. On average, this doesn't matter. As Wayne said, sometimes that is the other way around.

[James Heinrich]

Originally Posted by LaurV

P95 splits the classes differently, whatever that means, maybe he can clarify.

I believe Prime95 uses 420 classes whereas mfaktx uses 4620 classes (except for the less-classes version/setting when it also uses 420).

[LaurV]

Well, it could be easily checked in which class the factors were, in both cases.