Trial factoring to > 2^77?

[tuckerkao]

Originally Posted by Zhangrc

Me too, that doesn't hurt your throughput. Would you like to help with some wavefront TF?

Not until I get a new PC because it'll take too long to finish them. Would be more efficient to just run the PRPs on my current machine.

Since SRBase is the group that will take the advantage on the higher TF depths, it should definitely be the assignments of that group which you've just mentioned they need more works to do.

[Zhangrc]

Originally Posted by tuckerkao

Not until I get a new PC because it'll take too long to finish them.

How long exactly?
For me, TF from 2^76 to 2^77 takes 3 hours while PRP and P-1 takes 11 to 12 days. It still make sense to TF to 2^77.

[tuckerkao]

Originally Posted by Zhangrc

How long exactly?
For me, TF from 2^76 to 2^77 takes 3 hours while PRP and P-1 takes 11 to 12 days. It still make sense to TF to 2^77.

On my current machine, It'll take me 11 hours to TF from 2^76 to 2^77 for M115M, P-1 takes 8 hours, PRP takes around 12 days.

I have some cooling problems with my current GPU, it has to rest from time to time, the reason why I want to buy a new machine.

Too bad DrKirkby quit GIMPS, he prefers the wavefront exponents and has a lot of memories capable of running large amount of P-1 factoring.

I'll finish the P-1 which shows the details on the attached image below. If no one takes that exponent within a week, I probably won't run more.

[tuckerkao]

M115173323 has a factor: 64115403008215858576973359

The same F-PM1 fate as M168173323, interestingly both have a factor in the 2^80+ bits.

I don't need to run TF 2^76 to 2^77, that's almost certain no factor within that bit.

[Uncwilly]

Originally Posted by Zhangrc

Nowadays people mainly do trial factoring in the two-k project (which I have very little interest), and GPU72 is releasing TF assignments to 2^76 and seems very reluctant to give any higher bounds. However, there are hundreds of thousands of unfactored exponents in the 107-119M range

So I suggest doing trial factoring to higher than 2^77, starting from 107M. My reasons are as follows:

1. You can do TF beyond what GPU72 hands out. Get the assignment, edit the work todo, don't turn in the results until you turn in all of your TF
2. You can 'freelance' the TF after GPU72 has released the exponent.
3. mersenne.ca shows that 76 is the highest recommended level for that range (based upon how long it takes a GPU to do the TF vs the PRP.
4. It might be most profitable to do P-1 before the last bit(s) of the TF. There is overlap in what P-1 might find and what TF will find.
5. Many people are trying to help maximize the total speed of the project. Once an exponent is factored to the break even point (including P-1), it makes the most sense for the project to do the PRP.

[Uncwilly]

Originally Posted by tuckerkao

then ask the SRBase group to run every exponents up to 2^77 or 2^78.

You obviously haven't been following the discussion in the threads about SRBase's TF work. The plan has been doing breadth first TF. There have been discussion to try to change that to work near the wave front, but that is not in the (GPU)cards.

[tuckerkao]

Originally Posted by Zhangrc

How long exactly?
For me, TF from 2^76 to 2^77 takes 3 hours while PRP and P-1 takes 11 to 12 days. It still make sense to TF to 2^77.

I just learned that Intel Core i9 12900k will launch on Nov 19, 2021 with a probable retail price of $604.

If I pair it up with Nvidia Geforce 3060 Ti, the total price of the new PC shouldn't be exceed $2000. It'll be nice to have the new DDR5 memories and PCIe5 which shouldn't be less powerful than DrKirkby's machines.

Trial factoring of 2^76 to 2^77 of M115M will only take around 1 and half an hour on Nvidia Geforce 3060 Ti.

I figured that 2 of such PCs together will generate more total outputs than 1 $4000 AMD Threadripper 5970X PC with Nvidia Geforce 3080 Ti.

A lot of things can change during the meantime, I cannot make the final decision until I see the detailed descriptions of the available products.

[tuckerkao]

M106489771 served as a great example that a factor between 2^76 to 2^77 could be found through P-1.

Consider the chance of a P-1 finding a new factor is around 4x of a TF from 2^76 to 2^77, however the P-1 only takes slightly longer than the TF, as far as I understand that Kriesel is able to finish both of those within an hour and a half.

The 4x chance is a reality because digging out new factors between 2^77 to 2^100 have been quite frequent even using the standard bounds.

[tuckerkao]

Originally Posted by Zhangrc

How long exactly?
For me, TF from 2^76 to 2^77 takes 3 hours while PRP and P-1 takes 11 to 12 days. It still make sense to TF to 2^77.

You don't have to run P-1 and PRP together. If F-PM1, then don't run the PRP.

UncWilly has been correct, it's better to run P-1 on my CPU first while my GPU aiming another exponent, M168607223 has a factor between 2^77 to 2^78 which was found by P-1.

Running P-1 will be able to collect most of the low hanging fruits. If no harvests, then target the TF 1 bit higher than the GPU72 recommended level.

[Zhangrc]

M108398111 has a factor: 127634018463172816320049

The 108.3M have 2096 unfactored exponents. If F-PM1 probability stays around 4% we might still have ~2020 exponents. I suggest we should do some TF and P-1 now before it becomes too late.
Anyone like to help?