[QUOTE=airsquirrels;440129]It is official - the NVIDIA 1080 throws out garbage for 7xM and up exponents, seems solid on lower DC work though. I will wait for CUDA 8.0 final and retest.[/QUOTE]

Is this just for LL or TF too?

[QUOTE=Madpoo;440381]Well, hmm... That 37M exponent must have been in cat0 territory and it didn't get started within 10 days or whatever so it expired quicker.

I may need to pay attention to that when creating cat0 type assignments for him. I know he'll complete them in time but I forgot about that "needs to start in xx days" clause.

I tend to trawl through the exponents needing triple-checks every other day or so and I must have seen that one and snagged it...[/QUOTE]

No, it expired after 30 days. I was watching that exponent because I already guessed that my double check could be bad.

[QUOTE=rudi_m;440485]No, it expired after 30 days. I was watching that exponent because I already guessed that my double check could be bad.[/QUOTE]

We've been generating a queue of about 30 days for my GPUs, which don't check in until they are completed so I could see how we could miss the window by a few days in that category. We might need to take slightly smaller chunks.

[QUOTE=0PolarBearsHere;440483]Is this just for LL or TF too?[/QUOTE]

TF still doesn't work because of a CUDA 8 bug in the RC. So for now this card is being used for <4K FFT triple checks which seem to work out better.

[QUOTE=airsquirrels;440491]TF still doesn't work because of a CUDA 8 bug in the RC. So for now this card is being used for <4K FFT triple checks which seem to work out better.[/QUOTE]

Was it that card that did [url=http://www.mersenne.org/M38324009]38324009[/url] though?

[QUOTE=Mark Rose;440495]Was it that card that did [url=http://www.mersenne.org/M38324009]38324009[/url] though?[/QUOTE]

I had not noticed that one yet. It does seem that the card is less than reliable with the current CUDA framework. It does have a decent number of successes...

It is working on triple checks only, so is it better to keep it going and 2/3rds of the time it clears an exponent but 1/3rd of the time it wastes its time vs. sitting idle? Hopefully the CUDA 8.0 final resolves the issue. Shame as it is a very fast and efficient card.

[QUOTE=airsquirrels;440490]We've been generating a queue of about 30 days for my GPUs, which don't check in until they are completed so I could see how we could miss the window by a few days in that category. We might need to take slightly smaller chunks.[/QUOTE]

Well, I think I just need to watch out for the cat 0/cat 1 stuff and maybe see if you can get those done first or something. Ah well.

There was another instance where AirSquirrels and I had a crossover exponent... same situation where he had it first, then it expired and I got the next assignment. Fortunately I noticed his result checked in before I'd started so no duplicated effort there.

[QUOTE=airsquirrels;440497]I had not noticed that one yet. It does seem that the card is less than reliable with the current CUDA framework. It does have a decent number of successes...

It is working on triple checks only, so is it better to keep it going and 2/3rds of the time it clears an exponent but 1/3rd of the time it wastes its time vs. sitting idle? Hopefully the CUDA 8.0 final resolves the issue. Shame as it is a very fast and efficient card.[/QUOTE]

Maybe try underclocking the card and see if that fixes it? It could be a hardware issue.

Quick triple check needed
 

Hi all,

Would anyone care to run a quick triple-check on this?
[URL="http://www.mersenne.org/M43408411"]43408411[/URL]

There's a weird thing about the previous check and I just wanted to make sure mine matches before digging too deep. I'm pretty sure mine's right, but always best to make sure.

I also just checked in a mismatching triple-check, so now it needs a quad check:
[URL="http://www.mersenne.org/M40792159"]40792159[/URL]

No real urgency on that one except it's fun to do quad checks. :smile:

[QUOTE=Madpoo;440501]Hi all,

Would anyone care to run a quick triple-check on this?
[URL="http://www.mersenne.org/M43408411"]43408411[/URL]

There's a weird thing about the previous check and I just wanted to make sure mine matches before digging too deep. I'm pretty sure mine's right, but always best to make sure.

I also just checked in a mismatching triple-check, so now it needs a quad check:
[URL="http://www.mersenne.org/M40792159"]40792159[/URL]

No real urgency on that one except it's fun to do quad checks. :smile:[/QUOTE]

Queued them both. 43408411 is running right now and should be done in about 42 hours. 40792159 will be a couple of weeks.

[QUOTE=Mark Rose;440502]Queued them both. 43408411 is running right now and should be done in about 42 hours. 40792159 will be a couple of weeks.[/QUOTE]

Thanks. I have a hunch that when you turn in your result for 43408411 you'll see the weird thing I'm talking about (assuming you match my result). :smile: