HD6950 over-heating

[weeble]

Started getting "Display driver stopped responding and has recovered" messages popping up about a week ago, and in event viewer. Previously, re-installing .NET4 and CCC fixed this, but not this time.

Had a squizz at this http://support.amd.com/us/kbarticle...TIKMDAGhasstoppedrespondingerrormessages.aspx so thought what the heck, backed up vital stuff and reinstalled windows 7 64-bit. That got rid of the error, but non-intensive games would play for a while, then become stuttery and then freeze (sometimes to a black screen). Alt+Tab out and could then go back in, for a minute or 2. Under-clocking the card helped make it last a little longer, but nothing more.

Fired up GPUz, and noticed idle temps of 50 degrees. So set the GPU fan to 100% via overdrive. Play any game and temps rocket upto 106! Even web games get to 95ish. Cleaned all filters, fans - still same temps. Was even getting 101-102 with PC right by the open balcony doors.

Quick butchers today at http://forum-en.msi.com/index.php?topic=171259.0 suggests remove GPU cooler and apply some MX4 - how do I go about this? Guessing I'm barking up the right tree?

Bought MSI ATI Radeon HD 6950 OC Twin FrozR III Power Edition 2048MB GDDR5 back in May 2011 (no idea where the receipt is though )
PSU = 550W Antec True Power
Case = Zalman Z9 Plus

 

[Drangueos]

well you can try removing the cooler and reapplying the thermal paste.

 

[Zuma]

there is no reason why it should have failed so reapplying the thermal paste is unlikely to work.
check if the fan is spinning properly. also i assume there isn't another card blocking the fan intake for the card.
have you tried running with the side of your case off?
also try touching the heatsink with your finger when it is hot. if it's just bearably hot then thats ok, if it's so hot you can't bear to touch it then that's not good.
just thiking that the temp sensor could have failed and be giving false reading

 

[shankly1985]

Display driver stopped responding and has recovered Message, this mostly happens when you have a Overclock and not enough Voltage.

You Overclocking the GPU? Even if its a factory Overclock you might still need a tiny touch more voltage or lower the overclock.

 

[weeble]

there is no reason why it should have failed so reapplying the thermal paste is unlikely to work.
check if the fan is spinning properly. also i assume there isn't another card blocking the fan intake for the card.
have you tried running with the side of your case off?
also try touching the heatsink with your finger when it is hot. if it's just bearably hot then thats ok, if it's so hot you can't bear to touch it then that's not good.

Pretty sure all fans we're spinning, certainly the case ones were when I checked after a spring clean. Although, I am wondering if the rig is quieter now than it used to be... Yeah, it's just the one card - will try with the side off tonight (hearing reports that side fans don't always help with airflow sometimes) and the finger test...

Display driver stopped responding and has recovered Message, this mostly happens when you have a Overclock and not enough Voltage.

You Overclocking the GPU? Even if its a factory Overclock you might still need a tiny touch more voltage or lower the overclock.

Nope - no overclocks applied. Under-clocked it a while ago from Core Clock 850MHz > 800, and Memory 1300 > 1250.

 

[wazza300]

undo the four tiny screws that hold on the heatsink/fans,and pull out the fan connector from the pcb

then wipe off the old paste from the gpu die and blow out all the crud and dust from the heatsink,apply some new thermal paste and screw the cooler back on and connect the fan plug

then see how the temps are

 

[weeble]

undo the four tiny screws that hold on the heatsink/fans,and pull out the fan connector from the pcb

then wipe off the old paste from the gpu die and blow out all the crud and dust from the heatsink,apply some new thermal paste and screw the cooler back on and connect the fan plug

then see how the temps are

Cheers, never removed one before - ordered some MX-4 too. No harm in trying anything.

 

[RetrogamerX]

Cheers, never removed one before - ordered some MX-4 too. No harm in trying anything.

MX-4 is great paste. Good choice.

 

[C64]

try underclocking the card slightly lie 870mhz and so on

 

[faxfan]

You have a balcony?

 

[weeble]

You have a balcony?

I have 2 Have to open the doors as ambient temperature inside is 24 degrees.


Update from last night's fiddlings having taken case side off...and noticed both GPU fan's were misbehaving.
Fan nearest the back spins slowly, with an odd intermittent clicking type noise - nothing obstructing it.
Fan nearest the front was doing nada. Gave it a flick, and came to life at full speed.

Tinkering with CCC overdrive fan speeds, my PC's noise is restored (as had thought it was quieter than I remember) much to the annoyance of my Mrs.

However, back fan sounds like it's trying to accelerate RPM and failing, like a motor or something won't engage. Front fan seems to behave. Tested on increments of fan speed 50%,60%, 70%, 80% and 90%.

Rebooted 3 times - each time the front fan fires up, but back fan just literally ticks along. Shouldn't they both whizz along?

Can't figure out formatting here to include GPUz results - got readings for before, during and after.

 

[Zuma]

seems like that back fan is knackered.
it probably affects the other fan by sucking all the power from the fan controller which is why the other fan doesn't start up properly.
you need to rma it

or if it's out of rma then you need to replace the fans or fan, or the whole cooler

 

[weeble]

Cheers...It came with a 3 year warranty, which I bought from here in May 2011, and promptly threw away the receipt (oddly I've still got the box).

Warranty sticky implies a chat with OcUK "If bought from OcUK, they will deal with Warranty, typically 10 upto 25 days RMA..."

Fingers crossed I don't need the receipt.

 

[Screv]

I have a MSI 6970 which for ages ran way to hot, were talking up in the 90's when playing simple games on one screen, over 100 when playing The Witcher II on more than a few occasions. Eventually I to had driver crashes and all sorts of issues, partly caused by poor drivers and some related issues with memory dying on me.

So your prepared: the GPU on the AMD PCB is not covered like a CPU. What this means is you want to use a TIM that is not conductive. You can get away with them but you just need to be damn careful about how much you apply and where it goes. I felt it better to err on the side of caution and get some non conductive paste for the job.

When I got the cooler off my 6970 I found a mess of TIM all over the GPU and surrounding PCB. Used some Arctic Silver Articlean solution to get the original TIM off and cleaned up the area. Then applied Arctic Cooling MX-4 and did a few test fits before I was finaly satisfied with the application. Temps dropped by 25 odd degrees instantly using the same fan profile as used before with the factory applied TIM.

One of the things I noticed with the card which may or may not apply to you, the Twin Frozor II Cooler that is on my MSI card is damn heavy and the weight makes the card flex slightly, the heatsink in effect is putting pressure on the card in a direction away from the GPU. Not sure if I termed that in an easy to understand way but the gist is it does nothing to help with a good contact between the cooler and GPU. I read some people have added brackets or supports for the card to take the strain off.

As for the driver crashes: find a driver that works with a power profile your happy with. I increased the voltage for my card and under clocked it for a while to get a consistent stability and then increased the speed as and when I felt I had eliminated any other sources of instability. I still use the card now though I fully intend to replace it when AMD announce the next line up this winter. the 6000 series are still good contenders for current games and will happily contend with more recent cards if you can get them to be stable. By no means are most of them as temperamental as yours and mine have shown to be, but I have read about more than a few with similar issues.

Good luck with your trouble shooting

 

[weeble]

Thanks for the pointers - was quite looking forward to attacking it, but the RMA process began promptly.

Unfortunately MSI hadn't responded to OcUK after 30 days, so a 7870 is in the post.

At least I don't have to watch period drama's now!