Trippple Checks

[kladner]

Originally Posted by Madpoo

..... seems like there aren't as many systems doing DC work?

But only on part of an FX CPU. I can't crank them out like Intel people do. In any case, that is one area where I don't try to pick and choose assignments (unlike LLTF, sometimes.) Whatever P95 gets is what gets run.

[petrw1]

Originally Posted by Madpoo

seems like there aren't as many systems doing DC work?

Just a random (ok the most recent) recent-cleared report:
DISCLAIMER: only based on range....can't guarantee some of the LL in the 54-72M range aren't DC
2 days 7.5 hours
3,000 results (i.e. the rest are TF,P1 or ECM --- or DC/LL out of Range)
370 DC - 17,084 GhzDays
563 LL - 96,167 GhzDays

[chalsall]

Originally Posted by Madpoo

...seems like there aren't as many systems doing DC work?

It's not a function how many systems, but the aggregate "fire power". But, yes, you are correct in the meta -- DC'ing is very much falling behind the LL'ing.

Personally I would argue that anyone who is concerned about ongoing system(s) stability should be doing DC'ing. I only do DC'ing to monitor the systems I'm responsible for (to ensure they're "sane"), and because I don't want the head-ache of who gets the credit if a new MP is found (I accept the extremely (!) small chance that a DC will find a missed prime).

[Madpoo]

Originally Posted by Madpoo

Ah, well, I went ahead and added it to one of my workers then. I just looked on the server and there are 1219 other exponents smaller than this one so I guess it would be longer before it's assigned.

Maybe once I'm done futzing around with triple-checks and other nonsense I could dedicate some time to doing actual double-checks... seems like there aren't as many systems doing DC work?

And done:
M34548961

[Madpoo]

Originally Posted by chalsall

... (I accept the extremely (!) small chance that a DC will find a missed prime).

I'm not so sure it's all that small a chance.

Well, smaller than the chance of finding one with a first time check (which is already small), but even then, with some VERY high error rate systems out there, the odds of a "hidden" are probably anywhere from 1-5% of the odds of finding one in a first-time check with a functional system.

The key there being that 1-5% is somewhere in the area of the average error rate. The error rate seemed to vary over time, so the actual rate would depend on what range of numbers are currently being tested and when the wave of first time checks swept through. Note: It's not reasonable to get a good error rate past 35M where the data is scarce. Double-checks above the current "wave" have a very high rate of self-verifications so I'm not sure that's statistically meaningful.

[Madpoo]

I figured this was a good mini-milestone for triple-checking... down to just 1000.

I snagged all of the 45M + work. I think I have enough on my plate to last at least until the end of the month for the most part.

Besides this list here, I'm also working through exponents where the same person checked it multiple (3+ times). Those are kind of interesting because clearly something was going on that they tested it over and over again (like those weird examples I mentioned).

Enjoy! And thanks again to the folks working on these.

[Madpoo]

Here's another update.

I spent a little time stocking up some workers with enough number crunching goodness to last until mid-June or so. I now have exponents from 29M - 32M, and then from 40M up. I think that'll last me for the next month anyway.

I see ATH is just about done, so just holler if you were interested in more. The ranges from 27M-29M and 32M-40M are available.

When I'm ready to tackle more, I'll probably keep working down from 40M, so feel free to take smaller work; that works for me.

[Mark Rose]

I'll do these:

DoubleCheck=27016741,71,1
DoubleCheck=27026429,71,1
DoubleCheck=27045749,71,1
DoubleCheck=27055669,71,1
DoubleCheck=27107077,71,1
DoubleCheck=27161191,71,1
DoubleCheck=27166229,71,1
DoubleCheck=27189991,71,1

[petrw1]

I'll take 28,000,000 - 28,499,999

35 Exponents....could be more than a month.....maybe 2

[cuBerBruce]

I'll reserve these before everything's been taken.

DoubleCheck=27202993,71,1
DoubleCheck=27224819,71,1
DoubleCheck=27224843,71,1
DoubleCheck=27226307,71,1

I don't expect to start them before June.

[Mark Rose]

After seeing mprime's estimate, I'll grab these, too:

DoubleCheck=27229123,71,1
DoubleCheck=27251701,71,1
DoubleCheck=27251827,71,1
DoubleCheck=27284339,71,1

Everything should be done towards the end of June.