mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet

Reply
 
Thread Tools
Old 2007-08-16, 02:37   #1
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

17·263 Posts
Default Why would I get .001 factor points?

Twice in the last 5 days (Aug 10 and 15) I have received a .001 bump in my factoring points ... without returning any factors. I have one PC factoring at the leading edge (42-43M). Every time it completes a test it gets .112 points for NOT finding a factor. If it finds a factor it would show up as a cleared exponent and even at 63 or 64 bits would receive more than .001 points.

Similarly if my Phase I/II factoring of the LL tests of 10 Million digit numbers found a factor it would show up as cleared and awared more than .001 points.

There have been NO cleared exponents since Aug 9 ... a Phase II factor of 71 bits that awarded about .289 points.

My best guess is that I get .001 points for finishing Phase I/II tests without finding a factor. This did happen on the 10th and 15th.
petrw1 is offline   Reply With Quote
Old 2007-08-16, 04:56   #2
axn
 
axn's Avatar
 
Jun 2003

7·683 Posts
Default

Yep. P-1 gives you .001 without a factor, and .002 with a factor. Enjoy!
axn is offline   Reply With Quote
Old 2007-08-16, 06:57   #3
S485122
 
S485122's Avatar
 
Sep 2006
Brussels, Belgium

2·17·47 Posts
Default

P-1 indeed earns one 0,001 credit points if no factor is found. I suppose one gets 0,002 credit points by trial factoring for finding a factor to an already factored exponent. (I retested an exponent for testing purposes and the forgot to turn of primenet reporting and got 0,002 points for it.) But if one finds a factor one gets the same amount of points as if one had found the factor through trial factoring.

A couple of examples :
32848043 had been factored up to 68 bits. P-1 found the 78 bits factor 266513394516452020042561 in P-1 stage 2 and earned 0,170 credit points.

32670233 had been factored up to 68 bits. P-1 found the 81 bits factor 1787890320533655745697929 in P-1 stage 1 and earned 0,172 credit points.

As you can see the credit depends on the exponent and on how far it had been trial factored. (Although it is not always consistant : I got 0,287 credit points for finding a factor of M39122179, 0,455 for finding a factor of M37015621, both numbers had also been trial factored to 68 bits...)

The credit points awarded do not depend on the calculation effort, to be fair P-1 factoring should earn one a number of credit points proportional to the effort. If one looks after credit it is better to do only LL testing. Trial factoring earns about two thirds of the points and P-1 earns you about 400 times less points than LL testing... Unless you find a factor then you can even earn a little bit more than by LL testing.

Jacob
S485122 is offline   Reply With Quote
Old 2007-08-16, 10:17   #4
Xyzzy
 
Xyzzy's Avatar
 
"Mike"
Aug 2002

73·107 Posts
Default

19,136 P-1 tests completed: 19.136 years credit.

439 factors found: 1.01 years credit.

Preventing further LL tests, and thus reducing global warming: Priceless.
Xyzzy is offline   Reply With Quote
Old 2007-08-16, 11:59   #5
S485122
 
S485122's Avatar
 
Sep 2006
Brussels, Belgium

2·17·47 Posts
Default

I fully agree that the factors found represent a very usefull contribution to GIMPS. One of my 4 computers is especially designed for P-1 : XP64, 4GB of RAM...

I only keep on ranting about the unfair credit system that underrates factoring (especially P-1) except when a factor is found.

Jacob
S485122 is offline   Reply With Quote
Old 2007-08-17, 04:22   #6
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

17×263 Posts
Default

Hmmm ... .001 for P-1 seem disproportionally low.

Example:
I can factor a 42M from 62 to 68 bits in about a day and get .112 points without finding a factor and likely more if I find one depending on the bit level it is found at.

P-1 factoring on the other hand takes a couple days suggesting to me it is doing more CPU-Years of work but why only .001 points?
petrw1 is offline   Reply With Quote
Old 2007-08-18, 20:37   #7
jinydu
 
jinydu's Avatar
 
Dec 2003
Hopefully Near M48

2×3×293 Posts
Default

I can't claim to speak for Mr. Woltman, but my guess is that the disproportionately low credit for factor-less P-1 testing is to discourage cheating. It seems like it would be easy to just report "no factor found" and thereby move up the factoring rankings without doing any actual work. The 0.001 credit makes this tactic much less efficient.

However, this doesn't explain the "large" credit given for Trial Factoring work.
jinydu is offline   Reply With Quote
Old 2007-08-19, 08:29   #8
S485122
 
S485122's Avatar
 
Sep 2006
Brussels, Belgium

159810 Posts
Default

Obviously the credit awarded for P-1 factoring should take the B1 and B2 parameters into account. The prevention of cheating explanation is not satisfactory, one could do the same with trial factoring or even LL tests facking a residue as well (until found out by double checks...)

Trial factoring does not give a "large" credit. On one of my machines the processor earns 0,217 credits a day trial factoring and 0,307 when LL testing. It is when finding a factor that the 2,32 (or something like that) multiplier can rise your credit by day to 0,5. In the current ranges the probability of having an exponent with a factor that can be found by trial factoring from 62 to 68 bits is about 9% and even taking that into account one gets less than 0,24 credits a day on average. Of course the numbers will be different with another processor but the proportions would be roughly the same : 1 for P-1, 235 for trial factoring and 310 for LL testing.

Last fiddled with by S485122 on 2007-08-19 at 08:55 Reason: More likely bonus factor
S485122 is offline   Reply With Quote
Old 2007-08-19, 19:27   #9
jinydu
 
jinydu's Avatar
 
Dec 2003
Hopefully Near M48

2·3·293 Posts
Default

Quote:
Originally Posted by S485122 View Post
Obviously the credit awarded for P-1 factoring should take the B1 and B2 parameters into account. The prevention of cheating explanation is not satisfactory, one could do the same with trial factoring or even LL tests facking a residue as well (until found out by double checks...)
Well, that's what makes all the difference. Someone who always reported fake LL residues would eventually be caught through double checking; but there is no such deterrent with factoring (although admittedly, there are ways to cheat with double checking too).

Also, I meant "large" in comparison to P-1 testing. That is why I used the quotation marks.

Last fiddled with by jinydu on 2007-08-19 at 19:35
jinydu is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Two points devarajkandadai Math 19 2012-12-15 14:15
TF vs. LL Where are the breakeven points for MMs? aketilander Operazione Doppi Mersennes 6 2012-11-04 12:56
My GHz and points are off as of today???? Unregistered Information & Answers 14 2011-09-27 05:34
Lagrange points L4 and L5 davieddy Puzzles 7 2007-09-04 12:50
More points for PRP? Mystwalker Prime Sierpinski Project 6 2006-01-03 23:32

All times are UTC. The time now is 12:40.

Tue Dec 1 12:40:35 UTC 2020 up 82 days, 9:51, 1 user, load averages: 2.68, 2.61, 2.49

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.