Strategic Double Clicking

[Mark Rose]

Originally Posted by LaurV

Next time however, when you find a mistake in my tests, you should find a prime! Otherwise we both are wasting the time...

Always good advice.

[Madpoo]

Originally Posted by LaurV

So this is not a hardware failure, for the peace of my heart
You may find more like that, from the same period.

I did a spot check of your machines and only found that one that seemed like it had any possible trouble and I picked up all of it's exponents for double-checking. So far (and I might be done?) that was the only one that didn't match my test.

[Madpoo]

I was doodling around again and thought I'd see what happens if I break down by account, CPU, year, and then also by the app version. I tried the app version before, but not by year. I thought it could be interesting to see if it gives me any more leads on strings of bad results.

It might... I'm running a couple tests. Here's an example of a machine that I would have basically skipped over given it's annual good/bad, but looking at the app version as another factor, it gets curious.

CpuId-year-app	Bad	Good	Sus	Unk	Solo	Mis
xx-yy_2007-99	0	4	0	0	0	0
xx-yy_2007-136	6	1	1	3	2	2

(the app versions are "Prime95,v23,NT service" and "Prime95,v24 final,NT service" respectively)

In aggregate it had 6 bad and 5 good for 2007... still more bad than good but not quite at the point where I'd be going after it with lower hanging fruit. But something strange seems to have happened after it was upgraded to v24... it's not v24's fault I'm sure but probably another temporal coincidence that works in my favor.

I found a couple examples like that where I'm testing an exponent of theirs to see how it goes. I'm not sure how useful it'll be, but I just wanted to mention it in case it does end up being another useful vector.

[Madpoo]

Originally Posted by Madpoo

I found a couple examples like that where I'm testing an exponent of theirs to see how it goes. I'm not sure how useful it'll be, but I just wanted to mention it in case it does end up being another useful vector.

Oh, what the heck, if anyone wants to try some of these out, here's a group where the cpu in question had a 5:1 bad/good ratio:

exponent	Bad	Good	Unk	Sus	Solo	Mis	worktodo
36099269	11	2	8	2	6	4	DoubleCheck=36099269,71,1
36527723	5	1	6	1	5	2	DoubleCheck=36527723,71,1
36574903	10	2	7	0	5	2	DoubleCheck=36574903,71,1
36631909	11	2	8	2	6	4	DoubleCheck=36631909,71,1
36717617	11	2	8	2	6	4	DoubleCheck=36717617,71,1
36850763	10	2	5	0	4	1	DoubleCheck=36850763,71,1
36883499	10	2	7	0	5	2	DoubleCheck=36883499,71,1
37029217	11	2	8	2	6	4	DoubleCheck=37029217,71,1
37093729	10	2	5	0	4	1	DoubleCheck=37093729,71,1
37107457	5	1	6	1	5	2	DoubleCheck=37107457,71,1
37281071	11	2	8	2	6	4	DoubleCheck=37281071,71,1
37308521	5	1	6	1	5	2	DoubleCheck=37308521,71,1
37902607	10	2	5	0	4	1	DoubleCheck=37902607,71,1
37952581	10	2	7	0	5	2	DoubleCheck=37952581,71,1
37968001	11	2	8	2	6	4	DoubleCheck=37968001,71,1
38022109	5	1	6	1	5	2	DoubleCheck=38022109,71,1
38089477	5	1	6	1	5	2	DoubleCheck=38089477,71,1
43780291	25	5	5	2	5	2	DoubleCheck=43780291,72,1
44352641	25	5	5	2	5	2	DoubleCheck=44352641,72,1
51184253	25	5	5	2	5	2	DoubleCheck=51184253,73,1
51416447	25	5	5	2	5	2	DoubleCheck=51416447,73,1
51907231	25	5	5	2	5	2	DoubleCheck=51907231,73,1

[sdbardwick]

I grabbed the first (lowest) 5.

[Madpoo]

Oh good grief, really?

Anyone feel like doing the 5th check on this poor thing and put it out of its misery?
M39751183

[S485122]

M39751183 is already assigned to Hung Chien-Hua...

[ric]

Originally Posted by Madpoo

Oh, what the heck, if anyone wants to try some of these out, here's a group where the cpu in question had a 5:1 bad/good ratio:

exponent	Bad	Good	Unk	Sus	Solo	Mis	worktodo
36850763	10	2	5	0	4	1	DoubleCheck=36850763,71,1
36883499	10	2	7	0	5	2	DoubleCheck=36883499,71,1
37029217	11	2	8	2	6	4	DoubleCheck=37029217,71,1
37093729	10	2	5	0	4	1	DoubleCheck=37093729,71,1
37107457	5	1	6	1	5	2	DoubleCheck=37107457,71,1
37281071	11	2	8	2	6	4	DoubleCheck=37281071,71,1
37308521	5	1	6	1	5	2	DoubleCheck=37308521,71,1
37902607	10	2	5	0	4	1	DoubleCheck=37902607,71,1
37952581	10	2	7	0	5	2	DoubleCheck=37952581,71,1
37968001	11	2	8	2	6	4	DoubleCheck=37968001,71,1

Taking 10, the remaining ones below 38M.

[Madpoo]

Originally Posted by S485122

M39751183 is already assigned to Hung Chien-Hua...

That was fast. I hope they do it right or it'll need a 6th check.

[Madpoo]

Originally Posted by Madpoo

I found a couple examples like that where I'm testing an exponent of theirs to see how it goes. I'm not sure how useful it'll be, but I just wanted to mention it in case it does end up being another useful vector.

Results are in... both of the tests I did came back with mismatches (a good thing, in this case). So that seems like it might be useful... neither one of those came from systems I would have picked out before.

I'll wait and see how some more of these other ones do before getting too optimistic about it. In the meantime I've still been going through a bunch of exponents that already had mismatches and need a triple-check. With a special focus on exponents where neither machine had much of a record, so my triple-check would potentially knock one or the other into the "let's take a closer look" category.

I did find one machine that way where it had zero good and bad to start out, but had a couple of mismatches which seemed suspicious. I did a triple-check and that machine was the loser, so I started doing DC's on its other work and it ended up with something like 8 bad, 10 good. Not too bad (from my perspective that is...horrible for that machine). I may have one or two more from that system to go, but basically that's the kind of thing I'm hoping to uncover.

[Madpoo]

Originally Posted by ric

Taking 10, the remaining ones below 38M.

I took the rest (38M and up) for myself.