GTX 970 crashes

[NecroFlex]

Ok, so my friend came to me with this issue, his GTX 970 (1.5 years old) has been crashing on him, driver errors and such, he asked me to look into it and see if i can fix it or anything like that. I told him to send an email to Palit (it's a reference cooler Palit 970) and the store where he bought it from. Meanwhile i tested it.

 

I ran, Furmark, Catzilla, Valley, Fire Strike and Heaven. No crashes, Heaven ran for 3h, temps stopped at 80°C, Furmark ran for ~30min, same 80°C, no issues. So i thought his driver was corrupt, but i remembered reading that some benchamrks can work and on others it can crash, so i decided to (yarrrr) download Metro last light for it's benchmark tool. I max it out and ran it. Went well at first, then at the second scene, around the 70°C mark it gave me a black screen.

 

I had to force restart the PC. After i came back into windows, i opened up the event viewer and looked up for the error, ofc it was a driver crash, it said this:

Quote

Display driver nvlddmkm stopped responding and has recovered

So i reinstalled the drivers, same thing happened in the metro benchmark, i installed the oldest drivers i could find for this card, same thing, I bumped up the cooling in afterburner to soole it down better, it helped a little bit, but when i got near/hit the 70°C mark, it crashed.

 

Any suggestions? I have a feeling it's either the PCB or a memory chip.

 

 

[Starelementpoke]

9 minutes ago, NecroFlex said:

Ok, so my friend came to me with this issue, his GTX 970 (1.5 years old) has been crashing on him, driver errors and such, he asked me to look into it and see if i can fix it or anything like that. I told him to send an email to Palit (it's a reference cooler Palit 970) and the store where he bought it from. Meanwhile i tested it.

 

I ran, Furmark, Catzilla, Valley, Fire Strike and Heaven. No crashes, Heaven ran for 3h, temps stopped at 80°C, Furmark ran for ~30min, same 80°C, no issues. So i thought his driver was corrupt, but i remembered reading that some benchamrks can work and on others it can crash, so i decided to (yarrrr) download Metro last light for it's benchmark tool. I max it out and ran it. Went well at first, then at the second scene, around the 70°C mark it gave me a black screen.

 

I had to force restart the PC. After i came back into windows, i opened up the event viewer and looked up for the error, ofc it was a driver crash, it said this:

So i reinstalled the drivers, same thing happened in the metro benchmark, i installed the oldest drivers i could find for this card, same thing, I bumped up the cooling in afterburner to soole it down better, it helped a little bit, but when i got near/hit the 70°C mark, it crashed.

 

Any suggestions? I have a feeling it's either the PCB or a memory chip.

 

 

What drivers were being used?

[NecroFlex]

3 minutes ago, Starelementpoke said:

What drivers were being used?

The newest one (364.72) and the oldest i could find for this card (347.88)

[Starelementpoke]

Just now, NecroFlex said:

The newest one (364.72) and the oldest i could find for this card (347.88)

The newest one was having problems and reports of it bricking people's computers. Instead of using the oldest, try something a little more up to date? Try the one just before the newest one.

[Centurius]

Seems like it might be something wrong with the GPU on the hardware side. RMA it.

[hug0mac]

I had the same problem with my first GTX 970. I did return it and got a new one. It stopped working and all and drivers crashed and weird things. Today it's doing great.

[reptileexperts]

yep, had to RMA a card with similar signs and symptoms about 6 months ago (was my first nvidia reference card) they replaced it within a week, and the new one never gave that issue again. While the driver could be part of the issue, its likely something in the GPU itself 

[NecroFlex]

4 minutes ago, Centurius said:

Seems like it might be something wrong with the GPU on the hardware side. RMA it.

I figured it is probably that, he did send an email to both Palit and the store he got it from, since it's sunday i don't expect him to get an email back until tomorrow evening

[Rolling Potatoe]

1 minute ago, NecroFlex said:

I figured it is probably that, he did send an email to both Palit and the store he got it from, since it's sunday i don't expect him to get an email back until tomorrow evening

If you cant get an rma (dunno what your warrenty conditions are, might be that you are screwed after half a year)

try increasing the core voltage or/and decreasing the core clock a bit. That might be enough to make the card stable.

[Moonzy]

if you're overclocking, reset it to stock and see if it happens

 

alternatively, uninstall your current driver with ddu and install an older version like 361.xx or 362.xx, like the 361.91 for example

remember to get the right version for your OS

[NecroFlex]

12 minutes ago, Rolling Potatoe said:

If you cant get an rma (dunno what your warrenty conditions are, might be that you are screwed after half a year)

try increasing the core voltage or/and decreasing the core clock a bit. That might be enough to make the card stable.

Will do, it has a 2 year warranty (still 6 months to go), but the problem is that it's been modified, he cut a hole into the reference cooler to add a fan that blows air out to help with the cooling, he said that might help it if it was overheating...yeah.

[NecroFlex]

27 minutes ago, Starelementpoke said:

The newest one was having problems and reports of it bricking people's computers. Instead of using the oldest, try something a little more up to date? Try the one just before the newest one.

Ok, so i installed a 361.xx driver over the old one (347.xx) and for the first time it went trough all 3 of the Metro LL benchmarks. Running the benchmark again to see if it actually was a driver issue, tho it's still strange that it crashed with the 347 drivers but works with the 361 driver.

[joe_p231]

7 minutes ago, NecroFlex said:

Will do, it has a 2 year warranty (still 6 months to go), but the problem is that it's been modified, he cut a hole into the reference cooler to add a fan that blows air out to help with the cooling, he said that might help it if it was overheating...yeah.

with a mod like that id say its unlikely they will take the rma request since you will have most likely have void the warranty 

[NecroFlex]

10 minutes ago, joe_p231 said:

with a mod like that id say its unlikely they will take the rma request since you will have most likely have void the warranty 

''oh, so you modified the card with a fan so it runs cooler cause it was heating up cause the reference cooler sucks? You clearly broke it by making it run too cold!''

 

Basically i expect that.

[joe_p231]

2 minutes ago, NecroFlex said:

''oh, so you modified the card with a fan so it runs cooler cause it was heating up cause the reference cooler sucks? You clearly broke it by making it run too cold!''

 

Basically i expect that.

most likely response yes but i dont think they expect customers to be cutting and drilling the coolers

 

[NecroFlex]

5 minutes ago, joe_p231 said:

most likely response yes but i dont think they expect customers to be cutting and drilling the coolers

 

he cut the plastic top off to glue on a 80mm fan, no drilling

[joe_p231]

Just now, NecroFlex said:

he cut the plastic top off to glue on a 80mm fan, no drilling

still mods like that arent looked upon too nicely

 

[Kinda Bottlenecked]

6 minutes ago, NecroFlex said:

he cut the plastic top off to glue on a 80mm fan, no drilling

 

Warranty is gone bruh. 

[NecroFlex]

Well, i think the card is actually fine, with the 361.xx driver, it's been working without crashes in the metro benchmark, had it ran multiple times, no crashes. Ofc still needs further testing, but atm it seems it was a driver issue of sorts.