Odd Exponents Not Factored Deeply Enough
 

Due to heat issues, it's warm here and the pc also has a gtx-660 & gtx-970 in...I have my Haswell i7 doing ECM checks to keep the cpu temps down below 90C.

It recently found [URL="http://www.mersenne.org/report_exponent/?exp_lo=10775563&exp_hi=&full=1"]1077563[/URL] for which the factor doesn't appear to have too many digits, indeed a quick check at mersenne.ca reveals it's only [URL="http://www.mersenne.ca/exponent/10775563"]66.7 bits[/URL]

It took the cpu best part of an hour to find that, gtx-970 found it in 2m 16s.

So the question is should we be going back over all the old data and bringing it up to a minimum bit depth?

From the point of view searching for a prime clearly not, but from getting as many exponents as possible factored...

But that is partly the point of ECM. To catch factors that may have been missed. This question has come up many times before and the consensus seems to be it is not worth the effort to go over all the old exponents again. If you think otherwise, feel free to show your math.

[QUOTE=garo;407480]But that is partly the point of ECM. To catch factors that may have been missed. This question has come up many times before and the consensus seems to be it is not worth the effort to go over all the old exponents again. If you think otherwise, feel free to show your math.[/QUOTE]

It wasn't missed, we simply didn't go far enough.

Isn't that what TJAOI is doing?

[QUOTE=Gordon;407505]It wasn't missed, we simply didn't go far enough.[/QUOTE]

RIght. So, you're arguing for more ECM work, yes? Go right ahead! That's the joy of this project- everyone can contribute to whatever facet of the search speaks to them. These small exponents have been double-checked as composite, so any factoring effort is for the joy of clearing exponents- a noble pursuit, but not a priority in terms of finding mersennes.

Do brag about factors found, as each ECM hit tempts some of us a little more to try ECMing the smaller powers. Your results might just get some followers!

[QUOTE=VBCurtis;407513]RIght. So, you're ....[/QUOTE]
:tu: :goodposting: +1

[QUOTE=VBCurtis;407513]That's the joy of this project- everyone can contribute to whatever facet of the search speaks to them.[/QUOTE]

Agreed on that. I've been doing my fair share of side projects and had a blast. :smile:

One future idea for [STRIKE]wasting[/STRIKE] using CPU cycles involves me pulling up cases where an exponent had mismatched LL tests done, but was later factored. I want to do a 3rd test on those and see which of the other tests (if any) was correct so that I can identify the bad result.

Of course that benefits me by further identifying the flaky computers out there, but anyways...

Point being, if you have fun doing it and it's interesting to you, then go for it.

[QUOTE=VBCurtis;407513]RIght. So, you're arguing for more ECM work, yes? Go right ahead! That's the joy of this project- everyone can contribute to whatever facet of the search speaks to them. These small exponents have been double-checked as composite, so any factoring effort is for the joy of clearing exponents- a noble pursuit, but not a priority in terms of finding mersennes.

Do brag about factors found, as each ECM hit tempts some of us a little more to try ECMing the smaller powers. Your results might just get some followers![/QUOTE]

Actually I'm arguing for more TF work, my GTX-970 found it in 2 mins, the Haswell CPU nearly an hour...

[QUOTE=Gordon;407614]Actually I'm arguing for more TF work, my GTX-970 found it in 2 mins, the Haswell CPU nearly an hour...[/QUOTE]

What's the point of doing TF on exponents already double checked? Just for factors? Although finding factors is fun, it's not needed for the "project" as a whole I think.

EDIT: In case you're wondering why they are/were so low when LL was done on them.. this was before TF came to GPU's.

[QUOTE=kracker;407620]What's the point of doing TF on exponents already double checked? Just for factors? Although finding factors is fun, it's not needed for the "project" as a whole I think.

EDIT: In case you're wondering why they are/were so low when LL was done on them.. this was before TF came to GPU's.[/QUOTE]

Purely to find factors and to move them out of the "unknown" column, think of it as a tidying up exercise :smile:

I know all about why the TF levels are "low", well to some of you anyway, I was part of the original LMH group when factoring to 53 bits was considered extreme...

[QUOTE=Gordon;407614]Actually I'm arguing for more TF work, my GTX-970 found it in 2 mins, the Haswell CPU nearly an hour...[/QUOTE]
You need a larger sample size. I've done quite a bit of TF in the <10M range with mixed results. In the higher ranges it can be beneficial to TF a few bits more, but in the lower regions ECM is usually faster and with TF it takes ages to do a single exponent (1M exponent to 71 bits takes almost a day!).