The Unhappy Me thread

[fivemack]

Tue Feb  8 13:57:42 2011  lanczos halted after 303981 iterations (dim = 19221961)
Tue Feb  8 13:58:07 2011  lanczos error: only trivial dependencies found
Tue Feb  8 13:58:07 2011  BLanczosTime: 1710603

It turns out there is such a thing as too much oversieving

[bsquared]

Originally Posted by fivemack

Tue Feb  8 13:57:42 2011  lanczos halted after 303981 iterations (dim = 19221961)
Tue Feb  8 13:58:07 2011  lanczos error: only trivial dependencies found
Tue Feb  8 13:58:07 2011  BLanczosTime: 1710603

It turns out there is such a thing as too much oversieving

D'oh!.

[xilman]

If I'm a good dog they sometimes throw me a bone in.

[davieddy]

Originally Posted by xilman

If I'm a good dog they sometimes throw me a bone in.

Not sleeping in the kennel again are you?
Join the club.

David

[xilman]

Originally Posted by davieddy

Not sleeping in the kennel again are you?
Join the club.

David

Don't look so frightened, this is just a passing phase, one of my bad days,

Is there anybody out there?

[davieddy]

Originally Posted by xilman

Don't look so frightened, this is just a passing phase, one of my bad days,

Is there anybody out there?

Not sure whether this will make you feel any better,
but one of my pet hates is that BT woman's tone of voice
when she says "Please Hang Up".

David

[lycorn]

Originally Posted by xilman

Is there anybody out there?

Siitting in a bunker, here behind my wall
Waiting for the worms to come...

[lycorn]

A couple of days ago a thunderstorm provoked a rather significant voltage fluctuation in my area of residence. I was working at the computer and could clearly hear the blowers noise changing pitch, as if they were going to stop, and then resuming normal operating speed. The operation of the PC, from the point of view of the operator, was not at all affected. The house lights have also flickered, although there was no actual power cut.
End result: 4 DCs most probably gone to waste.
As the tests were close to the end, I let them finish, to realize that none of the residues matched the previous ones.
The curious fact about it was that no errors were reported (yes, I have error checking turned on...) and 3 of the 4 tests ended with zero error code. Just one of them gave out a sumout error a couple of hours after the voltage spike.
The power stability in my area is very good, and severe thunderstorms aren´t that frequent, so I have never bothered spending money on UPSs, voltage stabilizers or similar stuff.
Oh well, I guess that´s what Murphy Laws are all about...

[cheesehead]

Originally Posted by lycorn

End result: 4 DCs most probably gone to waste.
As the tests were close to the end, I let them finish, to realize that none of the residues matched the previous ones.

Perhaps if you wait for the triple-checks, you may find that one or more of the discrepancies are in your favor, and be very glad you let the DCs finish. :)

The curious fact about it was that no errors were reported (yes, I have error checking turned on...) and 3 of the 4 tests ended with zero error code. Just one of them gave out a sumout error a couple of hours after the voltage spike.

... consistent with the possibility that your DCs actually computed correct residues, rather than wrong ones.

The power stability in my area is very good, and severe thunderstorms aren´t that frequent, so I have never bothered spending money on UPSs, voltage stabilizers or similar stuff.
Oh well, I guess that´s what Murphy Laws are all about...

Okay -- and it surely would be a long (, long, long) shot if all the triplechecks were in your favor -- but Murphy gets around a lot, in places that might not be so obvious.

[ixfd64]

One of my computers just returned two double-checks with non-matching residues. What are the chances that something is wrong with the machine?

I guess I'll have to wait until the triple-checks are completed to know for sure.

[Mini-Geek]

Originally Posted by ixfd64

One of my computers just returned two double-checks with non-matching residues. What are the chances that something is wrong with the machine?

I guess I'll have to wait until the triple-checks are completed to know for sure.

If possible, run a torture test on the machine for a day or two, that way you can be sure (pretty sure, not 100% of course) one way or the other without waiting for the triple-checks.
The error rate is roughly 4%. The chance of both old residues being wrong whether yours are wrong or not, unless it's a special case like being some of the smallest 10M digit ones (they were run on older hardware, often overclocked, and took so long the chances of an error were much higher) or being close to an FFT cutoff, is about 1 in 625. The chances that over all your time with GIMPS you could really hit that 1 in 625 chance on one of your machines at some point could be very high, depending on your level of participation. After all, GIMPS found two primes about two weeks apart, which seems incredibly unlikely, but in fact over its then-12-year life, the chances of something like that were quite high (~40% IIRC).

Besides the obvious conclusion of faulty hardware, it's always possible a power brownout or other error, not necessarily the fault of the hardware, could throw off both tests.