mersenneforum.org > Data Getting <2k unfactored exponents for 108.3M
 User Name Remember Me? Password
 Register FAQ Search Today's Posts Mark Forums Read

 2021-10-21, 07:44 #1 Zhangrc   "University student" May 2021 Beijing, China 53 Posts Getting <2k unfactored exponents for 108.3M Currently https://www.mersenne.ca/status/tf/0/0/4/10800 shows that there are 2096 unfactored exponents. However, with the expected 4% probability of finding a factor in P-1, we may still have 2020 unfactored, and the range will fall into that "two-k" project. At that time, no PRP tests could be saved by finding a factor. Most of the exponents in that range are TFed to 2^76, some to 2^77 (all by me). But my power is just too small. (ETA is 2 years) So does anyone want to help? You ca help by either doing TF or P-1. Some recommended bounds: TF to 2^77, then run P-1 with B1=1300000, B2=60000000 TF to 2^78, then P-1 with B1=800000,B2=32000000 I recommend the latter if you have RTX 20xx/30xx. (P.S. I don't consider doing P-1 before the last bit level of TF a good choice, because Primenet may release the PRP assignment right after receiving NF-PM1 results.) Last fiddled with by Zhangrc on 2021-10-21 at 08:04
2021-10-21, 17:40   #2
petrw1
1976 Toyota Corona years forever!

"Wayne"
Nov 2006

483610 Posts

Quote:
 Originally Posted by Zhangrc Currently https://www.mersenne.ca/status/tf/0/0/4/10800 shows that there are 2096 unfactored exponents. However, with the expected 4% probability of finding a factor in P-1, we may still have 2020 unfactored, and the range will fall into that "two-k" project. At that time, no PRP tests could be saved by finding a factor. Most of the exponents in that range are TFed to 2^76, some to 2^77 (all by me). But my power is just too small. (ETA is 2 years) So does anyone want to help? You ca help by either doing TF or P-1. Some recommended bounds: TF to 2^77, then run P-1 with B1=1300000, B2=60000000 TF to 2^78, then P-1 with B1=800000,B2=32000000 I recommend the latter if you have RTX 20xx/30xx. (P.S. I don't consider doing P-1 before the last bit level of TF a good choice, because Primenet may release the PRP assignment right after receiving NF-PM1 results.)
Thanks for helping out.
My thoughts:
Most of the P-1 and PRP is being done by Ben Delo.
His P-1 bounds currently have a 4.7% expected success rate.
Like this one at 106M:
Code:
Ben Delo	106677187	NF-PM1	2021-10-21 10:39	11.6	21.8254	B1=878000, B2=52123000
If the statistics hold he should find about 98 factors; that would do it.
At the same time, if the TF horsepower is available it would be best if TF was at 77 before PRP gets there.
What I am seeing in the higher 107M ranges is that most exponents are getting TF to 77.
If that holds then 108.3M should be fine.
So if it was up to me I would wait and see what is left when the P-1/PRP finishes that range (could be a couple months) and then if any are left, then either P-1 any that were missed or TF more to 77.

2021-10-21, 23:49   #3
Zhangrc

"University student"
May 2021
Beijing, China

53 Posts

Quote:
 Originally Posted by petrw1 His P-1 bounds currently have a 4.7% expected success rate.
Yes, but there are only 1700 unverified exponents in that range. Even if Ben Delo does two third of the work, the average probability is about 4.4%, so still about 15 to go. That could be done by TF to 77 (some to 78).
Quote:
 Originally Posted by petrw1 So if it was up to me I would wait and see what is left when the P-1/PRP finishes that range
Sorry, that's too late. You may not save a PRP test by finding a factor. We'd better start now.
(Once a exponent is tested and verified, I will no longer factor it.)

Last fiddled with by Zhangrc on 2021-10-21 at 23:56

2021-10-21, 23:59   #4
chalsall
If I May

"Chris Halsall"
Sep 2002

10,039 Posts

Quote:
 Originally Posted by Zhangrc Sorry, that's too late. You may not save a PRP test by finding a factor. We'd better start now.
Go for it! Bring all the compute you have available right about now.

2021-10-22, 00:04   #5
Zhangrc

"University student"
May 2021
Beijing, China

53 Posts

Quote:
 Originally Posted by chalsall Go for it! Bring all the compute you have available right about now
I'll do that after I finish PRP M109769981.
My conputing power is too weak to finish 108.3M on time. Anyone want to help?

Last fiddled with by Zhangrc on 2021-10-22 at 00:06

2021-10-22, 00:10   #6
tuckerkao

"Tucker Kao"
Jan 2020

10001100102 Posts

Quote:
 Originally Posted by Zhangrc I'll do that after I finish PRP M109769981. My conputing power is too weak to finish 108.3M on time. Anyone want to help?
Finish M109769981 up to 2^79 is somewhat deep in the bit levels, but if that's your favorite exponent, then it's all good.

Last fiddled with by tuckerkao on 2021-10-22 at 00:11

2021-10-22, 00:19   #7
Zhangrc

"University student"
May 2021
Beijing, China

11111012 Posts

Quote:
 Originally Posted by tuckerkao Finish M109769981 up to 2^79 is somewhat deep in the bit levels
Off topic. Would you like to help? Any help is greatly appreciated.

 2021-10-22, 00:21 #8 petrw1 1976 Toyota Corona years forever!     "Wayne" Nov 2006 Saskatchewan, Canada 22·3·13·31 Posts With luck my GPU will free up in a few weeks. With bad luck it could be 8. Cross your fingers. Keep in mind though all PRPs (or almost all) are preceded by P1 so there shouldn't by many "wasted" PRPs.
2021-10-22, 00:22   #9
chalsall
If I May

"Chris Halsall"
Sep 2002

10,039 Posts

Quote:
 Originally Posted by Zhangrc Off topic. Would you like to help? Any help is greatly appreciated.
Fsck me...

2021-10-22, 00:34   #10
tuckerkao

"Tucker Kao"
Jan 2020

2·281 Posts

Quote:
 Originally Posted by Zhangrc Some recommended bounds: TF to 2^77, then run P-1 with B1=1300000, B2=60000000 TF to 2^78, then P-1 with B1=800000,B2=32000000 I recommend the latter if you have RTX 20xx/30xx. (P.S. I don't consider doing P-1 before the last bit level of TF a good choice, because Primenet may release the PRP assignment right after receiving NF-PM1 results.)
The PrimeNet will assign the exponents to the PRP testers because GPU72 only recommends up to 2^76 for M108.3M. Both 2^77 and 2^78 are considered optional to the server.

The recommended bounds should be more like:
TF to 2^77, then P-1 with B1=800000,B2=32000000
TF to 2^78, then P-1 with B1=1300000, B2=60000000

The higher TF bit level has to match the larger P-1 bounds.

I have Geforce GTX 780, so that's below your latter. I won't be able to meet your time.

I don't understand the needs for the extra TF levels after the GPU72 recommendations. If 2^76 to 2^77 is truly needed, it's that PrimeNet server which should be updated, so it'll assign an additional level to all the TF users to work on.

Quote:
 Originally Posted by petrw1 With luck my GPU will free up in a few weeks. With bad luck it could be 8. Cross your fingers. Keep in mind though all PRPs (or almost all) are preceded by P1 so there shouldn't by many "wasted" PRPs.
The chance of finding a factor between 2^76 to 2^77 that cannot be covered by the standard P-1 bounds is more like 1/77 * 0.75.

I definitely agree that the recommended bit levels should increase after Nvidia releases the Lovelace Geforce 4000 series in Q4 2022.

Last fiddled with by tuckerkao on 2021-10-22 at 01:06

2021-10-22, 02:42   #11
tuckerkao

"Tucker Kao"
Jan 2020

2·281 Posts

I actually have worked on an exponent in M108.3M a while ago which was M108377323

Because I submitted the results manually, the server assigned the exponent to curtisc even without the P-1.

Quote:
 Originally Posted by Zhangrc Would you like to help? Any help is greatly appreciated.
I think I can help M108285523 and M108366523, these 2 look more like my numbers.

Last fiddled with by tuckerkao on 2021-10-22 at 03:25

 Similar Threads Thread Thread Starter Forum Replies Last Post petrw1 Data 986 2021-11-27 04:53 drkirkby Miscellaneous Math 39 2021-08-24 21:08 sweety439 sweety439 0 2021-06-24 02:29 jasong Marin's Mersenne-aries 7 2006-12-22 21:59 WraithX GMP-ECM 1 2006-03-19 22:16

All times are UTC. The time now is 09:35.

Sun Nov 28 09:35:57 UTC 2021 up 128 days, 4:04, 0 users, load averages: 1.54, 1.23, 1.06