[QUOTE=LaurV;296802]I finished and reported, the P95 residue is indeed different, something with 060CAD...[/QUOTE]

:smile:

I remember thinking "heh, CAD, right up his alley" :razz:

[QUOTE=Dubslow;296803]
I remember thinking "heh, CAD, right up his alley" :razz:[/QUOTE]
:smile: That is why I remembered it, hehehe...., so original test was BAD! (not hex, but English).

Now for all other users who are against out little truce: This is a perfect example of what we were talking before. If Dubslow would report it and I (again, not personally, but generally speaking) would get it assigned without all this communication between us, then most probably I would have been done the job with CudaLucas, wasting a day of computer power, for no credit given, and for no help for the project. If the things would turn out the other way around, like the original being correct and Dubslow's CL being wrong, again, reporting it (a bad residue) would not help the project, but would "hit" Dubslow (at least his pride :P), incresing his percent of bad reports.

My twopences: whatever anyone say, reporting a DC mismatch without TC-ing it (by yourself or using a third), helps no one, nor the project. Of course, there are people who prefer to "set it and forget it", or better, "fire it and forget it".

I like it for the speed of the TC -- there's no need to even put it on your list, and we find out two days later. I'd rather not wait three months or more for ANON to tell me if I'm right or not :P

[QUOTE=Brain;296743]I've had a dozen successfull matches but since 2.00 i've had 6 mismatches and only 2 matches. :-( All mismatches happened when CL started with fftlength=1572864 and increased after less than 100 iters to 1835008... Can anybody confirm this observation?

Additionally, when I insert a new expo to worktodo.txt (CL already running) it doesn't resume: cudalucas.ini is +1 to high in that case: Line 4 contains the next expo, cudalucas.ini says 4 but would require 3...

Am I missing something?[/QUOTE]
2.00 has not been good for me, as well. I had a lot of mismatches with it. I've been too busy at work to troubleshoot the problem. 1.69 was very stable though.

[QUOTE=flashjh;296886]2.00 has not been good for me, as well. I had a lot of mismatches with it. I've been too busy at work to troubleshoot the problem. 1.69 was very stable though.[/QUOTE]
I'm 100% with it.

[QUOTE=Dubslow;296887]I'm 100% with it.[/QUOTE]
I think it's my cards. They're very good for TF, but CL requires *perfect* memory. When I get a card to test on I'll know for sure. The timing has a lot to do with it though, because 1.69 worked very well. I also had a power failure and CL hasn't worked well since, I think it may have ruined my cards.

I still trust 2.00
 

We should keep investigating. My first GTX 560 Ti was perfect with 1.69. But when I switched to CL 2.00 I added a second GPU. Temp of perfect GPU#1 when up from 79°C up to 93°C and I got mismatches. This may or may not be the reason.

I'd like to read more of your stats as my good 2.00 DCs were all done with no FFT increase at startup. All bad experienced a FFT increase. Coincidence?

[QUOTE=Dubslow;296887]I'm 100% with it.[/QUOTE]
This means you have full trust?

By the way, no overclocking here.

All my tests have been done with a specified length, in a small range of expos so that the specified length is guaranteed not to cause problems. Within the range I have tested, I have more trust in CUDALucas then in the random 5 year old tests on PrimeNet. (Edit: Factory OC from 675 to 751 MHz for me, but I have not OCd it any further, and running temperatures never break 65C, sometimes less than 60C.)

[QUOTE=Dubslow;296933]All my tests have been done with a specified length, in a small range of expos so that the specified length is guaranteed not to cause problems. Within the range I have tested, I have more trust in CUDALucas then in the random 5 year old tests on PrimeNet. (Edit: Factory OC from 675 to 751 MHz for me, but I have not OCd it any further, and running temperatures never break 65C, sometimes less than 60C.)[/QUOTE]

My temps are probably the problem. I'll investigate more later.

@Brain: what stats do you want to see?

[QUOTE=flashjh;296945]My temps are probably the problem. I'll investigate more later.

@Brain: what stats do you want to see?[/QUOTE]

The rule-of-thumb that I recall is that nVidia's are supposed to be safe north of 90C, maybe up to 100C; OTOH, word-of-mouth/rule-of-thumb is never very accurate, and of course, that's for playing video games, not error-sensitive computations. I'd say aim for 80, maybe 85, but anything beyond that is probably too much.

[QUOTE=flashjh;296945]
@Brain: what stats do you want to see?[/QUOTE]

I think of a table like this:
Exponent Range (# Initial FFT length) # Final FFT length # Number of good DCs # Number of bad DCs (# Average GPU temp # Overclock in % # Non default CL options # more if useful)

Here my latest results, not yet in the table format, 5 vs 6:
[CODE]
[LEFT]Verified test results
ExponentUser nameComputer nameResidue Date found
27474101 Sebastian Hornbostel Manual testing 47ADA87EFB861DD7 2012-03-14 14:33
27663523 Sebastian Hornbostel Manual testing EB1F96542A321991 2012-04-05 19:34
27905873 Sebastian Hornbostel Manual testing 53BDB222E1A1380C 2012-04-14 19:15
29309279 Sebastian Hornbostel Manual testing 62028C534D442A33 2012-03-18 20:16
43758037 Sebastian Hornbostel Manual testing E88FA154335EA2C3 2012-04-01 19:36 [/LEFT]

Unverified test results
ExponentUser nameComputer nameResidueError code (if any)Date found
27227191 Sebastian Hornbostel Manual testing FC794386077112__ 2012-04-11 17:25
27637219 Sebastian Hornbostel Manual testing EE7092F4DAF5E5__ 2012-04-08 15:44
28850039 Sebastian Hornbostel Manual testing 1AEE3C955657DC__ 2012-04-18 16:12
28887407 Sebastian Hornbostel Manual testing B4A9E01183BD8A__ 2012-04-12 17:38
29359303 Sebastian Hornbostel Manual testing 183CD50CBDC04E__ 2012-03-22 21:15
29555363 Sebastian Hornbostel Manual testing B197F6DD572574__ 2012-04-18 16:12[/CODE]I had a driver failure which I suppose to have killed 2 DCs.