User TJAOI

[kladner]

Up to 4, well, 5. Another showed up while I was writing this.

M79359781 has a factor: 86018707765317281777 [TF:66:67*:mfaktc 0.20 barrett76_mul32_gs]

M79329049 has a factor: 436594536103628352239 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]
M79227053 has a factor: 465022048023243077879 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]
M79004249 has a factor: 234687434795636918167 [TF:67:68*:mfaktc 0.20 barrett76_mul32_gs]
M79033063 has a factor: 463838510863801116457 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]

248 to go.

[petrw1]

Originally Posted by petrw1

And another 4 makes 15 so far.

Manual testing	77370851	F	2015-01-23 22:27	0.0	404115972658736858983	0.3503
Manual testing	77307031	F	2015-01-23 22:27	0.0	438083349942146687417	0.4406
Manual testing	77011177	F	2015-01-23 10:27	0.0	305268169799265418919	0.0378
Manual testing	72675403	F	2015-01-22 22:27	0.0	525252945259548579551	0.6840

6 more makes 21

Manual testing 77719069 F 2015-01-24 14:27 0.0 1043300478855676685249 1.2640 
Manual testing 77711563 F 2015-01-24 14:27 0.0 819036389770912845839 0.7270 
Manual testing 77705449 F 2015-01-24 14:27 0.0 290868301272790451633 0.3766 
Manual testing 77677547 F 2015-01-24 10:26 0.0 832564635208119196727 0.7637 
Manual testing 77619533 F 2015-01-24 10:26 0.0 1027401748819195733953 1.2315 
Manual testing 77551121 F 2015-01-24 06:27 0.0 346663356744036424591 0.1789

[apocalypse]

78M-79M range finished with 14 factors found.

M78029863 has a factor: 364794058631260354823 [TF:68:69*:mfakto 0.14 cl_barrett15_69_gs_2]
M78223967 has a factor: 588248925786788681369 [TF:68:69*:mfakto 0.14 cl_barrett15_69_gs_2]
M78384349 has a factor: 31605067156927414583 [TF:64:65*:mfakto 0.14 cl_barrett15_69_gs_2]
M78513509 has a factor: 163643721462565212863 [TF:67:68*:mfakto 0.14 cl_barrett15_69_gs_2]
M78520903 has a factor: 430500336104244139121 [TF:68:69*:mfakto 0.14 cl_barrett15_69_gs_2]
M78568093 has a factor: 204664989572145544961 [TF:67:68*:mfakto 0.14 cl_barrett15_69_gs_2]
M78582437 has a factor: 483475972781848663457 [TF:68:69*:mfakto 0.14 cl_barrett15_69_gs_2]
M78605207 has a factor: 566674712226093318121 [TF:68:69*:mfakto 0.14 cl_barrett15_69_gs_2]
M78692491 has a factor: 179185150747444736353 [TF:67:68*:mfakto 0.14 cl_barrett15_69_gs_2]
M78834727 has a factor: 1031162941046042297137 [TF:69:70*:mfakto 0.14 cl_barrett15_71_gs_2]
M78838867 has a factor: 239647633094808289367 [TF:67:68*:mfakto 0.14 cl_barrett15_69_gs_2]
M78843763 has a factor: 273559567441605180793 [TF:67:68*:mfakto 0.14 cl_barrett15_69_gs_2]
M78877333 has a factor: 104244699915142927367 [TF:66:67*:mfakto 0.14 cl_barrett15_69_gs_2]
M78932629 has a factor: 537007051232466434927 [TF:68:69*:mfakto 0.14 cl_barrett15_69_gs_2]

The factor for M78513509 was already found by P-1.
No exponents were already cleared by larger factors.

[flagrantflowers]

Has anyone checked exponents of users other than sannerud.com that TJAOI found missed factors for? Is there a way that we can get to the root of the problem of missed factors? Do any of these numbers have anything in common? (factoring class and other things I don't really know about?)

Is there a systematic reason why these factors have been missed? It is distressing to think that a large number of factors could have been missed during initial sieving at very low bit levels.

[bloodIce]

Originally Posted by flagrantflowers

Has anyone checked exponents of users other than sannerud.com that TJAOI found missed factors for?

Yes, at least I did. If we look in the post of Anonuser:

Originally Posted by Anonuser

53043031 53430761 53433521 58145281 58273277 58294177 58577543 58720187 60219487 61036883 65922127 66189911 66543011

We see, that KYOJI_KAMEI, deserves fair amount of double-check. So I do that for half a week by now (more than a thousand of records). I have a spider which fetched the work in the 58M-59M range, and I am double-checking his results, so far they are without any missed factor. Of course anyone interested is welcome to join, but usually no one has interest until the first missed factor is found. Everyone is so obsessed about profit, because the GHzDay credit is directly turn-able in gold and factors found are Nobel-prize achievement, so very few people want to spend power on double-checks. You are asking because you want to join the double-check or to see how the slaves are doing?

[TheMawn]

Originally Posted by bloodIce

We see, that KYOJI_KAMEI, deserves fair amount of double-check.

Careful: It isn't sufficient that someone worked on part of a candidate with a missed factor. They have to be the ones who submitted "no factor" for the range in which the missing factor is found.

For example, for 53433521 the factor found is ~2^60.13 but KYOJI_KAMEI did 2⁶⁴ to 2⁶⁵.

Looking at a couple of others I see that this user has in fact missed a couple, but we haven't really determined the value of re-doing the TF work yet.

The value is AT BEST the same as doing TF that has never been done before. Every non-missed factor decreases the value. For example, if I am expected to have missed 50% of all factors, then re-doing all my work will yield 50% of the normal amount of factors and therefore save 50% the amount of LL tests.

For Sannerud.com's laptop, his success rate was roughly 1 in 1,000 whereas it was expected to be along the lines of 1 in 70. This means he missed roughly 13 out of 14 factors which meant his work was re-done at ~93% value.

However, for someone who missed only, say, 5% of factors, then it very rapidly becomes a sub-optimal use of resources to re-do the TF.

If the miss rate is 50%, TF anything that is one bit-level below the optimal (for example, if the candidate ought to be TF'ed to 73, only redo anything up to 72).

The miss rate is 25%. make that two bit levels.

If the miss rate is 12.5%, make that three bit levels.


It becomes more complicated when we consider the option of doing P-1 instead of TF if there is a suspected missed factor.

[bloodIce]

58145281 : Factor of 64.790 bits, missed by KYOJI_KAMEI
58273277 : Factor of 64.442 bits, missed by KYOJI_KAMEI
58294177 : Factor of 66.327 bits, missed by KYOJI_KAMEI
58577543 : Factor of 65.286 bits, missed by KYOJI_KAMEI
58720187 : Factor of 65.989 bits, missed by KYOJI_KAMEI

I do not see your point with 53433521, why putting attention to something which is irrelevant. I agree that there his name is not at the missed factor (did I say that he is the only reason for missed factors?), but please do not say that 58M range looks fine to you.

The value is the value of profit you see... unfortunately. The value of cleaning the misses (at least in DC and LL range) is immense. Not to mention that accuracy in "no factor below" statements should matter something in this project ... or is just a bullshit.

Re-checking one or two users does not use too much resources (not to mention that part of the same resources would be spend to TF75 and DC/LL some of those expos for no reason).

And yes... if George points to which machine of KYOJI_KAMEI's missed the factors and compiles a list similar to that of Sannerud.com's, it will help to reduce the work, hopefully to a hundred GHzDays.

Originally Posted by TheMawn

we haven't really determined the value of re-doing the TF work yet.

How would we do that if we do not double-check at least partially? Or "we" would wait someone to do it for us? What was the "value" of work for Sannerud.com, before people jumped and discovered hundreds of factors?
I actually expect answers to some of these questions.

[kladner]

Five more. 200 to go.

Total, so far:

M79359781 has a factor: 86018707765317281777 [TF:66:67*:mfaktc 0.20 barrett76_mul32_gs]
M79329049 has a factor: 436594536103628352239 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]
M79227053 has a factor: 465022048023243077879 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]
M79004249 has a factor: 234687434795636918167 [TF:67:68*:mfaktc 0.20 barrett76_mul32_gs]
M79033063 has a factor: 463838510863801116457 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]

M79388209 has a factor: 635545254030291494921 [TF:69:70*:mfaktc 0.20 barrett76_mul32_gs]
M79387769 has a factor: 232488207444013028009 [TF:67:68*:mfaktc 0.20 barrett76_mul32_gs]
M79383181 has a factor: 278890815675960387673 [TF:67:68*:mfaktc 0.20 barrett76_mul32_gs]
M79373999 has a factor: 42176261588351110289 [TF:65:67*:mfaktc 0.20 barrett76_mul32_gs]
M79362671 has a factor: 396626201607035962649 [TF:68:69*:mfaktc 0.20 barrett76_mul32_gs]

[Mark Rose]

I've got 200 to go, all work between 68 and 71. They'll be done by morning and I'll report then.

[TheMawn]

Originally Posted by bloodIce

58145281
The value is the value of profit you see... unfortunately. The value of cleaning the misses (at least in DC and LL range) is immense. Not to mention that accuracy in "no factor below" statements should matter something in this project ... or is just a bullshit.

Re-checking one or two users does not use too much resources (not to mention that part of the same resources would be spend to TF75 and DC/LL some of those expos for no reason).

And yes... if George points to which machine of KYOJI_KAMEI's missed the factors and compiles a list similar to that of Sannerud.com's, it will help to reduce the work, hopefully to a hundred GHzDays.

How would we do that if we do not double-check at least partially? Or "we" would wait someone to do it for us? What was the "value" of work for Sannerud.com, before people jumped and discovered hundreds of factors?
I actually expect answers to some of these questions.

The goal of this project is to find the next Mersenne Prime. Trial factoring eliminates potential candidates. This much is clear.

The smart question is: "How much TF do we do?" but that is easy to answer. Using easy numbers as an example: If it takes 100 GHz-Days to do an LL test, finding a factor will save 200 GHz-Days. If we have a 1 in 100 chance of finding a factor, then it is reasonable for us to invest 2 GHz-Days into that TF because on average, after 100 candidates, we will have eliminated 200 GHz-Days worth of LL and it cost us 200 GHz-Days of TF.


Now, why do we re-do the TF? Well, the goal of this project is to find the next Mersenne Prime, and re-doing the Trial Factoring eliminates potential candidates.

The smart question is: "How much TF do we do?" but that is not so easy to answer. It still takes 100 GHz-Days to do the LL test and finding the factor still saves 200 GHz-Days, but what is the chance of finding a factor now? This is hard to tell. If it WAS 1 in 100 with a "fresh" run, it has to be less than 1 in 100 now.

For example, if I attempted 10,000 TF runs and found 75 factors (as opposed to the expected 100) then it is reasonable to suggest that I might have missed 25. If you're going to repeat my work, suddenly your odds of finding a factor are 1 in 400. Which means that it is now only reasonable for you to invest 0.5 GHz-Days into that TF, because on average, after 400 candidates, we will have eliminated 200 GHz-Days worth of LL at a cost of 200 GHz-Days of TF.


You're confusing what I called the "Work Value" with "Results Value". The Results are just as valuable as they have ever been (2 saved LL's is 2 saved LL's) but the "Work Value" is less because it has a lesser chance of success.

It's analogous to me saying that the results are more expensive to get.

[LaurV]

You start repeating yourself. This is troublesome for two reasons, first of all it might be from the age, and second, I thought that such privilege on this forum is only reserved for myself

We generally agree here, but don't make the same confusion as other people/projects (including gpu72 in the beginning) did, of counting efficiency in GHzDays. Different hardware have different efficiency for different type of assignments, like in the puzzle with the hen and half making an egg and half in a day and a half... but even that problem will ask you how many eggs 9 hens make in 9 days, and not in 9 GHzDays...

We also understand your calculus about reducing the chances when you re-check the range, which is right.

If one wants to help the project most, altruistically, then he has to think how many exponents he can clear in some time unit, day, hour, week, month. This depends on his hardware. If you can factor 70 or more numbers to 70 bits in the same time you would need to LL a single one, then you re better doing TF. Otherwise, better do LL. Generally you are better doing LL with a CPU, and you are better doing TF with an AMD card, regardless of the bitlevels (in reasonable ranges) and exponents. With a nVidia card, you may go one way or the other, depending on the card, the range and the bitlevel.

A Titan, for example, can clear a DC by doing LLDC, in about 10 hours. In the same time, if the card does TF, it would succeed in taking about 80 (DC-Range) exponents to 71 bits, but it would only take 40 to 72 bits. The last one is too less, because the chances to find a factor, therefore clear an exponent, is below 0.5. Even the 71 bits, this is at the limit and only make chance for expos which had no P-1 done, for which the chance to find a factor is about 1 in 70. If P-1 was done in the range, the chances to find a factor from 70 to 71 bits is about 1 in 90, and TF to 71 does not make sense. You still can do it if you consider yourself lucky, and/or if you believe that finding a factor is nicer and cooler than finding a LL residue which is not nul. If you have a titan and want to find a prime, then better do LLDC with it, than factoring DC exponents to 71. This was an example, and I didn't write the exponent range deliberately.