Newer milestone thread

[davieddy]

Originally Posted by kladner

I mis-wrote regarding assignments needing one bit-level to reach the release point. However, those need doing, too. I take such assignments sometimes, though I admit to trolling for newly released exponents which are starting from 68 or 69. In any case, I take them to the requisite release level.

It also seems that I may have posted this response to Davieddy in the wrong thread.....or the item I was responding to has been moved. It's hard to keep track when the same complaints show up anywhere and everywhere.

I have always admired Ludwig Boltzmann and although my only adversary is pond life, I empathize with him increasingly.

David
Age 62

[Uncwilly]

Originally Posted by NBtarheel_33

July 22, 2012. All exponents below 39 million have been tested at least once.

Just a little visual based upon the status files.

[NBtarheel_33]

Originally Posted by Uncwilly

Just a little visual based upon the status files.

Interesting to see the effect of the new v5 assignment rules on the minimum first-LL. Also note that the minimum two-LL number seems to maintain a linear trend.

If we do not find a bigger Mersenne prime in the immediate near future, we will have for the first time the situation where the first-LL minimum is above the largest known prime!

[patrik]

But don't forget that the minimum first-LL doesn't really say anything. There are presently about 284000 exponents below 43M that are available or out for double-check. Assuming a conservative estimate of a 1% error rate with zero error code, there are presently 2800 tests out there which had the wrong residue, (and one of these may have the correct residue zero!). They far outnumber (almost by a factor of 10) the pending first-time LL tests below 43M.

[Dubslow]

Originally Posted by patrik

But don't forget that the minimum first-LL doesn't really say anything. There are presently about 284000 exponents below 43M that are available or out for double-check. Assuming a conservative estimate of a 1% error rate with zero error code, there are presently 2800 tests out there which had the wrong residue, (and one of these may have the correct residue zero!). They far outnumber (almost by a factor of 10) the pending first-time LL tests below 43M.

That's why I personally haven't done a first time test in a while (excepting <43M tests). I highly encourage others to join!! With a few heavy hitters, we might even get as many DCs per day as LLs!

[kladner]

Originally Posted by Dubslow

That's why I personally haven't done a first time test in a while (excepting <43M tests). I highly encourage others to join!! With a few heavy hitters, we might even get as many DCs per day as LLs!

I'll look into when I resolve some technical difficulties. I should then be able to throw a GTX 460 at it. EDIT: (.....at least some of the time.)

[imwithid]

Originally Posted by NBtarheel_33

Interesting to see the effect of the new v5 assignment rules on the minimum first-LL. Also note that the minimum two-LL number seems to maintain a linear trend.

If we do not find a bigger Mersenne prime in the immediate near future, we will have for the first time the situation where the first-LL minimum is above the largest known prime!

I vaguely recall seeing these rules somewhere. Link anyone? I get the sense, however, that the rules don't always work too well. I now set all my work manually to double checking on some systems and TF for others (as most of my systems are now showing their age) as I was assigned a large P-1 number which has almost taken two years on a less often used system (Core 2 6300 - getting a bit old now, will relegate to either D-LL or TF) and another P4 system was assigned a large first time LL which will have taken almost four months to complete (fortunately, this one is on 24/7, will relegate to TF as the wiring is now a concern - extreme environment).

Despite the long turnover, I still adhere to the "marginally faster is still better" rule.

[Dubslow]

Originally Posted by imwithid

I vaguely recall seeing these rules somewhere. Link anyone? I get the sense, however, that the rules don't always work too well. I now set all my work manually to double checking on some systems and TF for others (as most of my systems are now showing their age) as I was assigned a large P-1 number which has almost taken two years on a less often used system (Core 2 6300 - getting a bit old now, will relegate to either D-LL or TF) and another P4 system was assigned a large first time LL which will have taken almost four months to complete (fortunately, this one is on 24/7, will relegate to TF as the wiring is now a concern - extreme environment).

Despite the long turnover, I still adhere to the "marginally faster is still better" rule.

The link is separate from your computer preferences, I have no idea why those aren't working.

[Dubslow]

This page seems to indicate there's only one LL remaining under 40M. You-know-what says the estimated completion is 808 days, and the computer name is "Netbook".

Exponent      WorkType  Stage, %     agedays  days to go EstimatedCompletion	   Next Update   Updated     Assigned	Userid	CPU Name
39286963	LL	LL, 59.20%	1200	808	  2014-10-18	         2012-07-26	2012-07-25  2009-04-19	Andrew	Netbook

Thoughts? I think many can guess what mine are.

[bcp19]

Originally Posted by Dubslow

This page seems to indicate there's only one LL remaining under 40M. You-know-what says the estimated completion is 808 days, and the computer name is "Netbook".

Exponent      WorkType  Stage, %     agedays  days to go EstimatedCompletion       Next Update   Updated     Assigned    Userid    CPU Name
39286963    LL    LL, 59.20%    1200    808      2014-10-18             2012-07-26    2012-07-25  2009-04-19    Andrew    Netbook

Thoughts? I think many can guess what mine are.

You mean other than 'an LL that's been running for ~4 years is likely to be bad'?

[garo]

Working through the remainder of my assignments. I still have a lot of 25Ms but at least they are all being worked on and will be finished in a month or so. I doubt the number of exponents out assigned to people not me will reduce by more than half in the same period.

I have 27/83 25-26Ms that have one LL test only. Let us keep track of this ratio.