Mini-Projects - Page 3

Christenson · 2011-10-08, 02:34

You do know that a given TF bit level gets easier as the size of the exponent increases, because only factors of form 2kp-1 are considered, where 2^p-1 is the number we are trial factoring.

If using P95, I believe you can do a DC yourself, because of the randomisation of the data. But CUDALucas might be quicker....

Dubslow · 2011-10-08, 03:01

Just got a week of TF-DC assignments. Early judgement indicates I can knock out quite a few DC's.

LaurV · 2011-10-08, 04:43

Quote:

Originally Posted by Christenson

because only factors of form 2kp-1 are considered, where 2^p-1 is the number we are trial factoring

2kp+1

Dubslow · 2011-10-10, 02:28

In about a day, maybe a bit more of work, going from 67 to 68, I've knocked out 2 exponents already. Quite a bit faster than DC, even on CUDALucas. That's 1 factor a day going to 69 bits, so that is still faster than the DC on CUDALucas (around 2.5 days on my card).

(Incidentally ckdo, as of 20 mins ago (already one unreported assignment done) I've reported all results up to and including M25494233.)

Dubslow · 2011-10-10, 02:30

Dang. After finishing the previous post, I checked that one unreported assignment I mentioned. Turns out it found a factor. So that's 3 in around a day.

Dubslow · 2011-10-10, 02:38

Just counted. In around 20 +/- 1 hr, I've reported 91 results, with 2 factors found, or if you add an hour, 3 factors in 95 candidates.

Christenson · 2011-10-10, 04:11

Quote:

Originally Posted by Dubslow

Just counted. In around 20 +/- 1 hr, I've reported 91 results, with 2 factors found, or if you add an hour, 3 factors in 95 candidates.

That's the statement that we need to do one or two more bit levels or so on the range you are working, before returns diminish to the point where it's better to do LL testing.

Remember that we can ask your card to do a DC with CUDAlucas at the same time we can ask a core to do it, so 1.25 days per single factor for you is the optimum, the point of indifference.

Keep up the good work!

Dubslow · 2011-10-10, 05:15

Well, actually my card's at about 90% on mfaktc, running CUDALucas at the same time would slow them both down quite a bit

MrHappy · 2011-10-10, 08:13

Quote:

Originally Posted by MrHappy

26M: 21387 at 67 bit (only 7158 still needing a DC)
27M: 11719 at 67 bit and 9902 at 68
28M: 21421 at 68 bit

54M-59M: 3631 at 68 bit and 17646 at 69 bit

26M: 21353 at 67 bit (only 7101 still needing a DC)
27M: 11303 at 67 bit and 10305 at 68
28M: 21051 at 68 bit

54M-59M: 3405 at 68 bit and 16785 at 69

Can someone easily get the TF-Depth for exponents in a range that are not DCed already?
Like: in 27M range XXX are at 67 bit and have no DC-LL and YYY at 68 bit ...

ckdo · 2011-10-10, 15:21

Quote:

Originally Posted by MrHappy

Can someone easily get the TF-Depth for exponents in a range that are not DCed already?

Nope.

For the sake of it, here's the count of exponents in my TF pool, by range:

Code:

25M:  583 @ 67
26M: 2392 @ 67
27M: 1612 @ 67
28M: 6313 @ 68

That will keep the gang busy until late November at yesterday's rate.

ixfd64 · 2011-10-10, 17:39

On the subject of TF, I'm surprised that George hasn't added GPU support to Prime95 yet. I mean, the code (mfaktc, etc.) is already available, so what is he waiting for?

2011-10-10, 02:28	#26
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	In about a day, maybe a bit more of work, going from 67 to 68, I've knocked out 2 exponents already. Quite a bit faster than DC, even on CUDALucas. That's 1 factor a day going to 69 bits, so that is still faster than the DC on CUDALucas (around 2.5 days on my card). (Incidentally ckdo, as of 20 mins ago (already one unreported assignment done) I've reported all results up to and including M25494233.) Last fiddled with by Dubslow on 2011-10-10 at 02:29

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Mini ITX with LGA 2011 (4 memory channels)	bgbeuning	Hardware	7	2016-06-18 10:32
Mini ITX in server case	bgbeuning	Hardware	8	2016-05-11 14:13
Mini STX	bgbeuning	Hardware	1	2016-04-20 13:35
What can I fit in mini-ITX and a 120W PSU	fivemack	Hardware	2	2016-01-01 17:50
Riesel base 3 - mini-drive I	gd_barnes	Conjectures 'R Us	199	2009-09-30 18:44

2011-10-08, 02:34	#23
Christenson Dec 2010 Monticello 5·359 Posts	You do know that a given TF bit level gets easier as the size of the exponent increases, because only factors of form 2kp-1 are considered, where 2^p-1 is the number we are trial factoring. If using P95, I believe you can do a DC yourself, because of the randomisation of the data. But CUDALucas might be quicker....

2011-10-08, 03:01	#24
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	Just got a week of TF-DC assignments. Early judgement indicates I can knock out quite a few DC's.

2011-10-10, 02:30	#27
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	Dang. After finishing the previous post, I checked that one unreported assignment I mentioned. Turns out it found a factor. So that's 3 in around a day.

2011-10-10, 02:38	#28
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	Just counted. In around 20 +/- 1 hr, I've reported 91 results, with 2 factors found, or if you add an hour, 3 factors in 95 candidates.

2011-10-10, 05:15	#30
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1110000110101₂ Posts	Well, actually my card's at about 90% on mfaktc, running CUDALucas at the same time would slow them both down quite a bit

2011-10-10, 17:39	#33
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 2·5·239 Posts	On the subject of TF, I'm surprised that George hasn't added GPU support to Prime95 yet. I mean, the code (mfaktc, etc.) is already available, so what is he waiting for?