Log in

View Full Version : P-1 Factoring having troubles keeping up with PRP firstpass testing



vjs
05-15-2006, 10:20 AM
Hey all,

Just letting everyone know that we have lost a few heavy hitters in factoring.

For those of you who don't know P-1 factoring trys to factor k/n pair just ahead of those we are going to test for primality. What this does is remove canidiates from the que which are certainly not prime faster and more effectively than regular testing.

This is somewhat of a big deal since it's easy to P-1 test in theory and those which P-1 eliminates won't have to be tested in the future (EVER!!!).

What this mean for regular testing is if a particular wasn't P-1 there is less chance that your testing the sought after prime.

The only way to help this is slow down main testing (not a good solution) by doing more secondpass testing. (I.E. entering your username followed by QQQ secondpass , for myself vjsQQQsecondpass). Or have just a few more people factor using P-1.

There is alot more in the factoring thread but personally I'd like to see all tests factored to at least B1=30000 B2= 340000 prior to P-1 testing. Perfered is somewhere around B1=50K B2=600K, but your best bet is probably using the prime95 default. Garo once said a little P-1 is better than no P-1 at all I can't agree more.

E, has been helping some people get started in the co-ordination thread but we still probably need another 5 people to build up the factoring vs testing buffer we once had.

If your interested ask some question here or headover to the factoring subsection and pick yourself up a range. It's not that tough but a little difficult to get started for the first timer, we are all here to help. I'd also like to point out there is a guide and prime95 is probably the best program to use.

Joe O
05-30-2006, 04:38 PM
Just to add my two cents. P-1 factoring could use some more help now. PRP testing is progressing faster than P-1 factoring.

em99010pepe
05-30-2006, 05:15 PM
Joe O or vjs,

How much an AMD 64 3000+ with 1GB would do per day?

Thanks in advance,

Carlos

Frodo42
05-30-2006, 11:45 PM
Joe O or vjs,

How much an AMD 64 3000+ with 1GB would do per day?

Thanks in advance,

Carlos

It all depends on what kind of parameters you set, my P4 3GHz, with mrpime does a P-1 test in about 1.5 hours in the area of the the current psp with the these parameters:
available ram for factring 512MB
sieve point: 2^50
factor value: 1.7

Tha means the computer will do ~17 test pr. day when it runs mprme 24/7 and does not run anything else.
With current factor density that means my P4 progresses about 800 in n pr. day.
From what I know AMD's are a little slower at factoring but your AMD would sure be a big help ...from my above calculation we need something like 10 3 GHz P4's running full time to keep up with prp at the current rate of this, we sure could use more power that would just give room for bigger factor densities which I think would still be efficient.

vjs
05-31-2006, 03:42 PM
Thanks for the update frodo,

what's your urrent success rate on finding factors?
Can you still factor out tests faster than it take to test them?


Note to others:
Generally one factor is worth a little less than 2 prp tests since there is no need to double check a factor. There was just a point not so long ago that we could actually remove tests faster by factoring them.

Frodo42
05-31-2006, 04:23 PM
Thanks for the update frodo,

what's your urrent success rate on finding factors?
Can you still factor out tests faster than it take to test them?


Well I would not be able to say, I think the estimated chance reported by mprime isn't of by much (it says chance 1.04% with the parameters given above) and it's been quite a while since I have had a 3GHz P4 run a prp test 24/7. My "gut feeling" tells me we are pretty close to the limit where I remove numbers from the firstpass queue equally fast by factoring and prp'ing. If I am right then it is still better to factor than prp since it saves the secondpass test for a k for which there is not found a prime before secondpass reached the n in question.
You guys are doing far too good a job at sieving, there aren't the big treasures lying around to pick up by factoring that once was there anymore and that is just a good thing :thumbs: