Results

[James Heinrich]

I believe the only overlap between doublemersennes.org and mersenne.ca is for MM31 (http://www.doublemersennes.org/mm31.php vs https://www.mersenne.ca/MM31)
If anyone does TF on MM31 sufficient to bring it to the next bitlevel (91: k=576460752571858944) please PM or email me to let me know. I will not be monitoring this thread so please message me directly.

[kriesel]

[Sun Dec 18 00:20:55 2022]
UID: kriesel/asrock-gtx1650super, no factor for MM127 in k range: 255P to 260P (186-bit factors) [mmff 0.28 mfaktc_barrett188_M127gs]

[ET_]

Originally Posted by James Heinrich

I believe the only overlap between doublemersennes.org and mersenne.ca is for MM31 (http://www.doublemersennes.org/mm31.php vs https://www.mersenne.ca/MM31)
If anyone does TF on MM31 sufficient to bring it to the next bitlevel (91: k=576460752571858944) please PM or email me to let me know. I will not be monitoring this thread so please message me directly.

No one is working on it atm... thank you for the advice.

Luigi
---

[kriesel]

I'll give it some attention later. (Will post in ODM reservations thread then.) There are several programs that could be applied on MM31 to 92 bits, which is within the capability of mmfactor (96), mfaktc (92), mfakto (95), and Mfactor & Factor5 on the CPU side which can go quite high. I estimate from 91 bits to 92 would take 4 months in mfaktc on an RTX2080, so 90.64-91 should be about 3 weeks in mmfactor, the only suitable GPU app I know of that also accepts k value input for partial bit level execution.

[Andrew Usher]

As you may have seen, I convinced James to go back to the old P-1 limits, which are nearly optimal for any setup that doesn't use the 30.8 algorithm. Possibly after 91, and certainly after 92, P-1 would be superior to further TF (I don't think more detail is necessary there).

Of course you don't need to take it at all, but I think you see it would be nice for this exponent to get it to a whole bitlevel.

[kriesel]

[Sat Dec 24 03:39:53 2022]
UID: kriesel/asr3-rtx2080, no factor for MM127 in k range: 265P to 270P (186-bit factors) [mmff 0.28 mfaktc_barrett188_M127gs]

[kriesel]

[Tue Dec 27 07:36:53 2022]
UID: kriesel/dodo-rtx2080super, no factor for MM31 in k range: 450P to 460P (91-bit factors) [mmff 0.28 mfaktc_barrett96_M31gs]

[kriesel]

[Sun Jan 01 20:35:32 2023]
UID: kriesel/emu-gtx1650, no factor for MM127 in k range: 260P to 265P (186-bit factors) [mmff 0.28 mfaktc_barrett188_M127gs]

[James Heinrich]

[Thu Jan 5 05:19:26 2023]
UID: JamesHeinrich/RTX4090, no factor for MM31 in k range: 480P to 490P (91-bit factors) [mmff 0.28 mfaktc_barrett96_M31gs]

[kriesel]

[Fri Jan 06 22:25:54 2023]
UID: kriesel/asr3-rtx2080, no factor for MM31 in k range: 460P to 480P (91-bit factors) [mmff 0.28 mfaktc_barrett96_M31gs]

[kriesel]

Must not have been fully awake when posting that:

Originally Posted by kriesel

There are several programs that could be applied on MM31 to 92 bits, which is within the capability of mmff (96), mfaktc (95), mfakto (92), and Mfactor & Factor5 on the CPU side which can go quite high. I estimate from 91 bits to 92 would take 4 months in mfaktc on an RTX2080, so 90.64-91 should be about 3 weeks in mmff, the only suitable GPU app I know of that also accepts k value input for partial bit level execution.