GPU TF work and its impact on P-1

[Xyzzy]

Optimally, couldn't we wait 10 years, doing nothing but saving money on our electrical bill (which could be used to purchase a few computers in 2021) and then knock out all of this work?

[davieddy]

Originally Posted by Xyzzy

Optimally, couldn't we wait 10 years, doing nothing but saving money on our electrical bill (which could be used to purchase a few computers in 2021) and then knock out all of this work?

I often get the impression that some folk would favour a moratorium
on TF and LL testing. Now why didn't Euler or Lucas think of that?

David

PS I've thought of a nice puzzle based on that proposal. Look out for it!

[R.D. Silverman]

Originally Posted by KingKurly

I do a fair amount of P-1 in the 1.9M range .

Purpose?

[KingKurly]

Originally Posted by R.D. Silverman

Purpose?

None other than to find factors. I realize the mathematical value is low (if even existent at all) but it tied in well with this year's theme of things remotely related to my birthday, somehow. Hopefully my LL on M91103891 is less offensive to your finely tuned mathematical senses?

[davieddy]

Originally Posted by KingKurly

I occasionally do P-1 work on exponents that were LL'd once but never P-1'd. (Almost all of them are by curtisc.) You would propose that I cease such activities?

My guess is that curtisc could only successfully enlist all the computers
in his university by ensuring minimal impact on their other uses, notably
by not requesting memory for P-1.

Rather than cease "such activity", I would encourage you to do a ckdo
and get rid of these eyesores as quickly as possible!

BTW the number of TF assignments is misleading as a guide to how
much is actually being done. How many exponents do these folk
have stacked up at a time?
One assumes that each LL test assigned is being worked on currently
(albeit maybe slowly). By contrast, I detect almost no change in the
number of TF or P-1 assignments below 53M, suggesting that very few
are being worked on.
Perhaps ckdo could shed more light on this.
For a 50M exponent, TF to 68 takes 0.6 Ghz-days. TF to 69 takes 1.2
and TF to 70 takes 2.4.
wtf is going on???

David

[chalsall]

Originally Posted by davieddy

By contrast, I detect almost no change in the number of TF or P-1 assignments below 53M, suggesting that very few are being worked on.
<snip>
wtf is going on???

One way to easily see the actual "bit depth" progress is to view http://www.mersenne.info/trial_facto...ular_data/1/0/ (for example), and leave the page open in your browser (or save it to disk). Then the next day open a new tab and view the page again.

Flipping between the tabs will show you the actual work being done.

This will show the progress for all exponents below 100M. You can, of course, also drill down to watch a specific 10M (or 1M, etc) range.

[science_man_88]

Originally Posted by chalsall

One way to easily see the actual "bit depth" progress is to view http://www.mersenne.info/trial_facto...ular_data/1/0/ (for example), and leave the page open in your browser (or save it to disk). Then the next day open a new tab and view the page again.

Flipping between the tabs will show you the actual work being done.

This will show the progress for all exponents below 100M. You can, of course, also drill down to watch a specific 10M (or 1M, etc) range.

you mean like : http://www.mersenne.info/trial_facto...a/4/100000000/

instead of : http://www.mersenne.info/trial_facto...a/0/100000000/

[davieddy]

Originally Posted by science_man_88

you mean like : http://www.mersenne.info/trial_facto...a/4/100000000/

instead of : http://www.mersenne.info/trial_facto...a/0/100000000/

Anyone who has read this thread will know how much importance
I attach to TFing 9 digit exponents to 64 bits, as far as finding a prime
is concerned, and how much credit gets awarded

[Radikalinsky]

Originally Posted by Christenson

Do think upon all those LL assignments that return abandoned.....if we can do a little something for those folks, it will return many-fold.

I do think about if someone could save the partial work of those folks who give up. Probably they could upload their backup files to GIMPS. The next one who picks up his number could continue where he stopped.

Some security might be necessary so that nobody can steal someone else's work. Also, the correctness of the partial result should be somehow secured so that it is worth to continue this work.

GIMPS may need a new TByte disk to save the backup files (each 5 .. 6 MBytes on my PC) and a better internet connection.

[davieddy]

Originally Posted by chalsall

One way to easily see the actual "bit depth" progress is to view http://www.mersenne.info/trial_facto...ular_data/1/0/ (for example), and leave the page open in your browser (or save it to disk). Then the next day open a new tab and view the page again.

Flipping between the tabs will show you the actual work being done.

This will show the progress for all exponents below 100M. You can, of course, also drill down to watch a specific 10M (or 1M, etc) range.

Many thanks.
I've never seen that link before.
Can you get there from mersenne.org?

In the light of this thread, that "how far factored" page makes for depressing reading: in the 40-60M range, virtually no exponents
have been factored past the "CPU limit", and many are below it.
By contrast, exponents between 23 and 33M are all factored about
2 bits further than the CPU limit for DC.

David

[chalsall]

Originally Posted by davieddy

Many thanks. I've never seen that link before. Can you get there from mersenne.org?

Yes, you can.

Originally Posted by davieddy

In the light of this thread, that "how far factored" page makes for depressing reading: in the 40-60M range, virtually no exponents have been factored past the "CPU limit", and many are below it.

Please keep in mind David that not everyone has the same agenda as you.

GIMPS allows those who participate to do what ever they want.

Even if that's closer to GIMFS than you might wish....