Folding@home News 27th August - 3rd September

Noticed my fah client showed "Failed" on the gpu slot tonight, think it must've been like that all day, hadn't downloaded any work, just sitting there empty with failed in bold red where it normally says running in green.

Not sure what would cause that but closed and reopened the client and it's downloaded wu now and running, dropped the clock speed by 105mhz, hoping it is just a glitch :(
 
thinking ill split my amd and NVidia gpus into different pcs to see if it allows the 970 to get the numbers it should be getting.

biffa - what gpus do you have? I know some are 750 ti's but not all can be you have a shed load lol
 
Last edited:
na its a power colour the duo turbo model I got for a " decent " price from gummy tree :D

its overclocked by around 40mhz only.

ive found another aswell for cheap :o :D
 
we'll see , I was hoping for another NVidia so I can split 2 pcs up with 2 cards in each and not have any drivers mess with each other!

pc 1 970 and ??? maybe a 760 I found cheap or what ever I see or a 780 even. ( main pc )
pc 2 290 and 280x

but my z77 has 3 pci slots so I could still get the other AMD card aswell :eek: , only problem Ill need an expensive psu.
 
Last edited:
I updated my drivers only to see my GPU submission points drop by about 50% despite benchmark and game results being better than my previous driver with performance in power mode for F@H. The client/GPU PPD was the same too with 13 WU's completed, lol odd because I've never come across this before.

Had to clean out 355.82 driver and reinstalled it using the same stock clocks, will monitor the points over the next couple of days to see if they go back up. If they go back up I will put this down to some installation issue that only affects F@H/CUDA.
 
I updated my drivers only to see my GPU submission points drop by about 50% despite benchmark and game results being better than my previous driver with performance in power mode for F@H. The client/GPU PPD was the same too with 13 WU's completed, lol odd because I've never come across this before.

Had to clean out 355.82 driver and reinstalled it using the same stock clocks, will monitor the points over the next couple of days to see if they go back up. If they go back up I will put this down to some installation issue that only affects F@H/CUDA.

I've had a couple of wu's that ended up erroring at around 30%, whilst they were running my ppd was less than half (99k as opposed to 230k). The gpu use was at 100% but the tdp was at a fixed 50% and the energy meter confirmed that, showing ~175w draw at the wall compared to normal of 320w.

Not sure if coincidence or not?
 
I've had a couple of wu's that ended up erroring at around 30%, whilst they were running my ppd was less than half (99k as opposed to 230k). The gpu use was at 100% but the tdp was at a fixed 50% and the energy meter confirmed that, showing ~175w draw at the wall compared to normal of 320w.

Not sure if coincidence or not?

Thanks for the information. Had a look over my log file and there were several instances that coincidences with time of points being dropped - I quickly deduced this is the problem:

- Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900

Means that maximum temperature must be less than 110C, odd when I limited this to 83C in Afterburner, up a little from 81C with good PPD. And temperature wait of no more than 900 seconds for cooldown.

Typical wait for reduced temperatures was 2 minutes and this should not cause a 50% reduction in PPD. However if reliability was a factor and core was keep being paused it would be reasonable to expect less points. Submission of WU in the log showed it was submitted without error.

What I will do is reduce temperature limit in Afterburner to 81C and keep everything else stock. Funny that I'm having to tame these GPUs to get more points :D
 
2 things ive noticed after monitoring folding for a week or so now

1. my amd gpu get core 21 units nearly every time were my gtx970 only gets core 18's giving much less ppd.

2.core 21's increase my temps by 5-6c over core 18's


i must have good air flow in my case because even if i remove my side panel the temps stay exactly the same or drop 1c , might buy 2 high CFM 140mm fans though and see if i can force cold air onto them to lower temps just so the dam gpus fans arnt as loud. or even try one forcing air in and the other hot air out.
 
Last edited:
After finding the client showing 'Failed' against the gpu slot the other day, I've had a couple of failed wu's and then the gpu wouldn't go above 500mhz on the core. I've reseated the card, cleaned and reinstalled drivers and the card is now running at stock core but the memory clock is showing as 0mhz in both msi afterburner and hardware monitor:

PP036gq.jpg

Is this the beginning of the end for this card?

Hmmm :( Might have to get some 280x's if yours is anything to go by andy :)
 
pause the folding slot then restart it, every now and then myn does this, if I check the folding client its still showing that its working! but when I check the gpu usage its 0% so I pause the client and then restart it, say folding is showing 20% done, its not and when the GPU goes back to 100% load the folding client will catch up and then the 20% goes back down to say 14% or something which is were it was up to.

not saying its the same for your card but try it to see if its your card or the folding client.


oh ye go for 290 or 290x if you really want a card! my 290 is doing 300k ppd on core 21's which is most of the time, you don't have members market access yeti don't think, but youd get 2 x 290s for £300 in there ;)
 
Last edited:
Rob, what were the reported memory clocks via GPU-z 0.8.5 under load?

Helpful to rule out this being a bug with Afterburner or if possible test the card in another PC. There's also Nvidia Inspector to check clocks and what performance bin it's in.
 
Thanks for the replies, downloaded nvidia inspector, memory clock greyed out:

Kkd6uIV.jpg

And GPU-Z:

WIBF8Xg.jpg

Might see if gigabyte would take it back under warranty, lost the receipt for it but they give 3 year warranty and the 780ti was launched 23 months ago.
 
Last edited:
if its still in warranty then they can go of the serial number as to who its registered to

plus side is gigabytes RMA is in London or that way , so you'll have a repair or a new card within a few days
 
From Gigabyte Support:

"Goods without proof of purchase receipt are accepted for repair attempt only but are not replaced if non repairable. For such cases, goods should be returned via the seller where possible. Please forward the full EAN code and Serial number, this may help determine who possibly sold the VGA to you from which distributor. See attached photo example that will assist in identifying the correct numbers needed."

I'll keep hold of it until it fails totally, if I send it now and they attempt a repair but fail then I'll regret it. Fingers crossed !
 
Last edited:
Yeah that memory being blank isn't normal, is it blank at idle too?

Ignore the P-state on mine because that is only the default clocks, overclocked is P0 state showing normal reported clocks while Folding:
pIFjH6r.png
Though your card would work, bad memory can cause glitches sometimes or all the time. This is when I recommend you run EVGA's OC Scanner X to check for artifacts (click the artifact preset button). Don't run anything else while this is running, may cause false positive. Give it a good 2 hour run to get an idea.
 
Back
Top Bottom