GTX 970 running hot

[lexidobe]

Hello everyone, I have just picked up a 2nd free 970, but it has a problem. Long story short but a friend of a friend needed their alternator changed, so I helped them out. I was going to ask him for like $20 for my time, but he offered me a working GTX 970 that he described as "running hot" even though he replaced the thermal compound. I took it and figured it was probably normal reference model 80-85 degrees. When I put it in my system to put it to work folding, I realized that it does actually have a problem. It idles at a pretty normal temp of around 30 degrees, but as soon as I load it it instantly hits the upper 80s and starts throttling bad, and that's with the fan at full speed. With the exception of my water cooled 970 which obviously has no fan, I run the fans on my folding cards at max speed to keep temps down.

 

I then took the card out, took the cooler off, replaced the thermal compound, reassembled, and tried it again. It still runs hot, replacing the thermal compound made no difference. I suspected a contact problem between the core and the cooler, so I removed the backplate as well as all the screws holding the cooler on except the 4 around the core and re tested it and found that the temps are normal this way. Does this mean anything to anyone? Everything looked good when I had the cooler off, thermal pads were in place. I'm thinking he probably mixed the screws up when he replaced the thermal compound and put long screws where short ones belong or vice versa, although they all seem to screw in correctly. The short screws are behind the backplate and the long screws are the ones that go through the backplate, which seems right to me. I just don't know what else it could be though. Does anyone have any ideas? Thanks in advance! 

[Alienware 15 R2]

I just think that's the Normal Temprature with those 980s.

 

I'd keep all the Fans At max.

[lexidobe]

Just now, Alienware 15 R2 said:

I just think that's the Normal Temprature with those 980s.

 

I'd keep all the Fans At max.

It is normal at stock fan speeds, but at maximum fan speed it should be much lower and not throttling. Like I said when I removed all the screws holding the cooler on except the 4 around the GPU core it was running in the low 50's and not throttling, a normal temp for 100% fan speed. 

[Alienware 15 R2]

try undervolting the GPU or get a Better Cooling fan solution for the Graphics card Fan shroud.

[lexidobe]

11 minutes ago, Alienware 15 R2 said:

try undervolting the GPU or get a Better Cooling fan solution for the Graphics card Fan shroud.

As far as I know there is no way to undervolt a 970, I tried years ago and couldn't find a way to do it and it doesn't look like anyone else has either. If you know of a way feel free to let me know though. The fan isn't the problem, it's moving plenty of air. The problem seems to be with the contact between the GPU die and the cooler since temps are good with the cooler held on by only the 4 screws around the core, but it runs hot with all the screws in.

[Alienware 15 R2]

Well then if that seems to be the Issue get a New and Better Fan Shroud?

[adamr539]

57 minutes ago, Alienware 15 R2 said:

Well then if that seems to be the Issue get a New and Better Fan Shroud?

DIY a CPU cooler onto it  Wrong person sorry.

Edited March 27, 2020 by adamr539
Wrong quote.

[adamr539]

2 hours ago, lexidobe said:

Hello everyone, I have just picked up a 2nd free 970, but it has a problem. Long story short but a friend of a friend needed their alternator changed, so I helped them out. I was going to ask him for like $20 for my time, but he offered me a working GTX 970 that he described as "running hot" even though he replaced the thermal compound. I took it and figured it was probably normal reference model 80-85 degrees. When I put it in my system to put it to work folding, I realized that it does actually have a problem. It idles at a pretty normal temp of around 30 degrees, but as soon as I load it it instantly hits the upper 80s and starts throttling bad, and that's with the fan at full speed. With the exception of my water cooled 970 which obviously has no fan, I run the fans on my folding cards at max speed to keep temps down.

 

I then took the card out, took the cooler off, replaced the thermal compound, reassembled, and tried it again. It still runs hot, replacing the thermal compound made no difference. I suspected a contact problem between the core and the cooler, so I removed the backplate as well as all the screws holding the cooler on except the 4 around the core and re tested it and found that the temps are normal this way. Does this mean anything to anyone? Everything looked good when I had the cooler off, thermal pads were in place. I'm thinking he probably mixed the screws up when he replaced the thermal compound and put long screws where short ones belong or vice versa, although they all seem to screw in correctly. The short screws are behind the backplate and the long screws are the ones that go through the backplate, which seems right to me. I just don't know what else it could be though. Does anyone have any ideas? Thanks in advance! 

DIY a CPU cooler onto it

[lexidobe]

10 hours ago, adamr539 said:

DIY a CPU cooler onto it

That won't work, I have 3 cards in my system. My main card (an EVGA 970 FTW with an EK block) and 2 folding cards. Currently those folding cards are 2 reference 960's, but may as well swap one of those cards for a more powerful 970 so I have 2 970's and a 960. 

[Alienware 15 R2]

I think that could be the cause of some Hot air going on in that Case of your that causes the GTX 970 to get that Hot?

[lexidobe]

Just now, Alienware 15 R2 said:

I think that could be the cause of some Hot air going on in that Case of your that causes the GTX 970 to get that Hot?

That's not it, I have very good airflow through my case. Also, the cards are the reference blower style, so they are exhausting their heat directly out the back of the system rather then dumping it into the case.

[Alienware 15 R2]

is the 970 a blower style?

[lexidobe]

1 minute ago, Alienware 15 R2 said:

is the 970 a blower style?

Yes, it's a reference blower style as well. 

[Alienware 15 R2]

Look,

Blower styles are Terrible versions of the GTX 970 that are Known for making the Cards overheat.

 

The Reason yours is overheating is because they’re close together and are sucking air from each other and not getting cooled properly.

[Alienware 15 R2]

Alright this is the solution,

 

Keep your fans spinning at 100% if you want the temperatures to go down instead of up,

Normal temperatures for the GTX 970 Blower Model are 75 to 80 degrees Celsius. Max Temperature is 98 degrees Celsius.

 

if the GPU reaches that temperature then crank the fan up to the max and turn off your Monitor.

 

wait 5 minutes then get back on the computer.

[lexidobe]

6 minutes ago, Alienware 15 R2 said:

Look,

Blower styles are Terrible versions of the GTX 970 that are Known for making the Cards overheat.

 

The Reason yours is overheating is because they’re close together and are sucking air from each other and not getting cooled properly.

Nope. As I said if I remove all the screws holding the cooler on except the 4 around the core it cools fine. Also, I tested that card in one of my extra systems rather than in my main system because it was easier than taking my 960's out of my main system with the water hoses in the way, so it was the only card in the system and still running hot if the cooler had all the screws in it. With the cooler held on by only the 4 screws around the core the temp doesn't exceed the low 50's. 

 

EDIT: As I said the fan speed was set to 100% the whole time during the testing and it is still running hot.

Edited March 27, 2020 by lexidobe

[Alienware 15 R2]

Try using MSI afterBurner to Lower core voltage and Memory Voltage by -100.

then retest folding.

[lexidobe]

2 minutes ago, Alienware 15 R2 said:

Try using MSI afterBurner to Lower core voltage and Memory Voltage by -100.

then retest folding.

Again- To the best of my knowledge there is no way to undervolt a 970, the slider in Afterburner will only move to the right to increase voltage, it won't move any farther to the left than zero. Same with the other programs I tried. From the Google search I did it doesn't look like anyone else has been able to undervolt a 970 either. If you know of a way feel free to let me know though.

[cobalt77]

So it is a situation where the temps were bad, you removed the back plate and temps were good, and then you screwed it all back together and temps were bad again? Or I miss something?

[lexidobe]

2 minutes ago, cobalt77 said:

So it is a situation where the temps were bad, you removed the back plate and temps were good, and then you screwed it all back together and temps were bad again? Or I miss something?

Basically yes. The card runs hot with the cooler held on by all the screws, but if I remove all the screws except the 4 around the core the temps are good.

[cobalt77]

Any major problems with leaving it without all the screws in? (I guess this is a solution you don't want)
Putting all the screws back in must twist or pull the cooler shroud in such a way that the air passes over/around the HS, rather than through it. What happens if the screws are screwed down loosely?

[lexidobe]

4 minutes ago, cobalt77 said:

Any major problems with leaving it without all the screws in? (I guess this is a solution you don't want)
Putting all the screws back in must twist or pull the cooler shroud in such a way that the air passes over/around the HS, rather than through it. What happens if the screws are screwed down loosely?

The major problem with leaving those screws out is that the cooler's thermal pads may not make good contact with the card's VRMs and VRAM chips, causing those to overheat and fail. Only the core has a temp sensor, so I wouldn't know if something else is overheating. As quickly as it heats up (within 5 seconds it's in the high 80s and throttling) it has to be a contact issue between the cooler and the core rather than an airflow issue, which would take a while to heat up. 

[cobalt77]

I did miss that you wrote above about how the temp rockets up, sorry.
BTW what make/mod of 970 is it?

Any chance that the HS is being levered off the GPU core by being pulled down on one side by screws or pushed down by the shroud?

[Mick Naughty]

What back plate is it? Don’t recall any reference cards coming with a back late until the 10 series. 

[lexidobe]

1 minute ago, Mick Naughty said:

What back plate is it? Don’t recall any reference cards coming with a back late until the 10 series. 

It's just the one that comes on it