GPU folding problems

[jtmc713]

Started this question in the install guide thread, making a new one here, slightly different issue though.  Ever since trying to use two cards to fold, my GTX980 doesn't fold nearly as fast as it used to.  It's gone from 2-3 hours per WU, to 7, 8, or even 9 days per WU.  Everything is on a clean install, latest drivers except for chipset; AMD crimson version drivers refuse to install for some reason or another.  Any help would be appreciated!

[Imakuni]

Your unit is still in the very early stages, where estimates are (usually) over the top. Wait a few % and it should go back to normal.

[tobben]

If the above mentioned isn't the case, and it's not helping or making any

progress at all. It might be mixing up the cuda id. Which is definitely a thing

that happens when mixing amd/nvidia cards.

 

If that is the case though, it wouldn't just increase the completion time, it

pretty much wouldn't make any progress at all, unless it is bugging out 

and hitting your cpu instead.

 

If you click configure in the top left, go to the slots tab, select your 980

and click edit. From there you can manually set your cuda id, Try 0 

and/or 1 for abit, after a few minutes it should work on one of them. 

 

That is if the 980 in your slot actually is the 980, and not just the amd

gpu showing up as the 980.

 

Actually, what is your current situation, how is the gpu setup, did you 
remove the second gpu?

[jtmc713]

My setup right now is just the 980 installed.  I took the extra step of formatting and reinstalling windows to get rid of any mixed gpu problems.  The ETA did go down though, it went down to 5 hours after some time. However, the ETA/point match up are still far off from what it used to be before hand.  Maybe it all just needs time? 

 

I'll try manually setting the cuda id on the next WU, if things still seem off.

[tobben]

If you use gpu-z or something similar and you are seeing proper 

utilization. Somewhat stable 90-100% cuda id shouldn't be the
problem.

 

I can't really suggest a driver cleanup when you did a full re-install.

 

Remember that different wu's will have different sizes and time to
complete aswell as varying point return. It also takes a few percent
into a wu before you can expect any proper estimates.

 

I'd have a look at the gpu utilization over 10 minutes or so, maybe
post a screenshot here.

 

Actually, scratch that. I forgot you did that in the op. Just how far off
is your ppd? You should be golden as long as the load is fine, and
you id/passkey is correct.

 

Looking at your stats i am starting to suspect you might not be using
a passkey that qualifies  for QRB.

[jtmc713]

How would I come by a passkey that qualifies for QRB? I just recently set up my id and passkey on the same day I started folding.

 

Just an update though - I tried the 270x again, using older drivers that are catalyst, and not crimson.  Everything installed successfully, however, it's returning errors in the log and requesting new WU's every minute or two. Log shows the same potential energy error.  Only thing that differs is the subsequent numbers. 

 

07:51:51:WU01:FS01:0x17:ERROR:exception: Potential energy error of 146.865, threshold of 10
07:51:51:WU01:FS01:0x17:ERROR:Reference Potential Energy: -721751 | Given Potential Energy: -721604

 

PSU not strong enough? I couldn't find any other info on the error.

[tobben]

Heh, i doubt that is the problem. I think i have encountered that before, but

not very familiar with that error .. Maybe it is references the calculation itself

and you are outside the error threshold, not sure.

 

You seem to be way off course with the 146.865 considering the threshold 

of 10. 

 

Got any potential sour overclocks or the like? How early do you get the error,

does both cards even start folding?

 

As for the passkey, you need to return atleast 10 QRB qualified wu's with a 

return rate above 80% to get QRB. Hence why you are probably not getting

it.(or maybe you qualify by now)

[jtmc713]

No overclocks whatsoever! My 980 would fold normally while the 270x would continuously error.  The error would come immediately or within a few minutes, % completion would be between 0.00 and 0.02%.  That's the reason I have a high WU count; the bad WU's would rack up quickly and get turned in. 

 

It might just be a bad card though.  Same error produced in another setup, so I'm assuming that's the case here.

[tobben]

Could very well be the case, try installing the gpu on its own.

clean out and re-install the drivers, then run some unigine 

heaven and furmark on it and see what happens.

 

Monitor the load while doing so, look for any irregularities.