Current recommended TF bit depth?

[endless mike]

Any way to find out what the current recommended trial factoring bit depth is? I previously had Misfit set to adjust assignments to 72 bit depth; lots of assignments through, and decent amounts of factors. More recently, I tried to 75; obviously much longer for each, and less frequent factors. I'm currently running a bunch to 74, but that's just a SWAG as to what's most efficient. I'm wondering what's best for my GPUs and is most helpful for the project.
I haven't signed up for GPU to 72. Would doing so and setting it to let GPUto72 decide work best?

[chalsall]

Originally Posted by endless mike

I haven't signed up for GPU to 72. Would doing so and setting it to let GPUto72 decide work best?

If you take a look at the Current Trial Factoring Depth report on GPU72 it will give you a good idea as to the optimal cut-offs we're currently working towards -- the yellow cells are where we release back to Primenet. At the moment, for LL TF'ing, that means going to 75.

TF'ing to a lower bit-level is fine and helpful to the project, and you will find more factors per time unit. But, at the end of the day, all candidates will be taken to 75 by someone(s) before being released back to Primenet for the first LL test (of two).

Signing up to GPU72 has the advantage in that it coordinates with all the other GPU72 TF'ing workers, and balances the "fire power" according to what is needed at the various "wave fronts" (including the P-1'ers) at the current moment.

[Gordon]

Originally Posted by chalsall

If you take a look at the Current Trial Factoring Depth report on GPU72 it will give you a good idea as to the optimal cut-offs we're currently working towards -- the yellow cells are where we release back to Primenet. At the moment, for LL TF'ing, that means going to 75.

TF'ing to a lower bit-level is fine and helpful to the project, and you will find more factors per time unit. But, at the end of the day, all candidates will be taken to 75 by someone(s) before being released back to Primenet for the first LL test (of two).

Signing up to GPU72 has the advantage in that it coordinates with all the other GPU72 TF'ing workers, and balances the "fire power" according to what is needed at the various "wave fronts" (including the P-1'ers) at the current moment.

Clicked on that to have a look, sort of interesting, while there I also had a look at the weekly progress reports.Think something must be broke. I've turned in best part of 2000 results in the 40-50m range in the last week yet the table shows no movement of any exponents at all and they are all apparently sitting at 58 bits?

[chalsall]

Originally Posted by Gordon

Clicked on that to have a look, sort of interesting, while there I also had a look at the weekly progress reports.Think something must be broke. I've turned in best part of 2000 results in the 40-50m range in the last week yet the table shows no movement of any exponents at all and they are all apparently sitting at 58 bits?

Sigh... Yeah, I know...

Mersenne.info is broken at the moment. Not sure when I'm going to be able to devote the time required to bring it back online.

The Current Level Report is fully functional, but only gives instantaneous status, rather than the historical delta reports.

Sorry about that.