GPU temp 3 is too hot, 6970 lightning

[metallicaband]

After running 15 minutes of 100% furmark load, here are the GPU temps as reported by GPU-Z:

GPU Temp 1: 64
GPU temp 2: 76
GPU temp 3: 96 !!

Idle temps are close though at around 33, 31 & 37 degrees.

I'm running the fan at pretty high speed @65%, even though I live in a hot place my room's AC is on so the ambient temps are normal (I'd guess around 20-24). If I run the fan at lower speeds the third GPU temp would easily hit 100+.

GPU-Z says that the exact location of those specific parts of the GPU are unknown, so if you guys have any idea about the actual locations of those GPU parts, please let me know so I'd have an idea about where to look at to decrease the GPU temps.

My overall case airflow management is pretty decent since the HAF X is easy to deal with even for a newbie, I'm using the first PCI-E lane on my sabertooth motherboard.

Any tips to decrease the third GPU temp would be great.

 

[TheGamerZ]

I'd imagine that those are the PWM's. They regulate the power and they will get quite hot. They're built to tolerate some extreme temperatures.

 

[laukinisvilkas]

From my long and hard experience - gpu temp #3 stands for something inside gpu die.
I had pretty much the same problem with high gpu temp #3 and 6850 (ok not the same architecture but it should describe similar things).

To put it simple - bad factory TIM application. Once I changed TIM my gpu temp#3 went from ~98 in furmark to ~62C (gpu temp #1 #2 and #3 should stay within few degrees because all stand for something in-die).
And since I have not seen anything similar to your or my problem, I think it is really really really rare. And original application didnt look bad at all - it was evenly spread on all gpu, no psychically distinguishable blind spots.

 

[kcmastrpc]

After 5 mins of Furmark my XFX 6970 was cooking...

GPU1: 94C
GPU2: 102C
GPU3: 102C

I'm going to apply some arctic silver and see if that helps.

 

[kcmastrpc]

I cracked open my 6970 to discover the TIM on the GPU was thin and there were noticable spots which weren't even covered at all. I cleaned up the board, applied a thin layer of AS across the gpu and then did the same for the RAM and power components.

My card went from 61C to 40C idle and under serious load it won't go above 91C.

 

[metallicaband]

Thank you all for the response.

I assume I'll void warranty if I tried to take off the cooler to reapply the thermal paste?

 

[tangoseal]

Airflow in your case. Airflow in your case. And..... drumroll............... Airflow in your case.

of course that is in addition to your regooping the heatsinks.

 

[magoo]

If you do not know what you are measuring, how do you know how hot it is supposed to be during load?

You need to find out what the specifications are on the measuring points.....before you go tearing off the heatsink and actually making it worse?
or applying a slightly conductive TIM and causing a short?

 

[Kenjiwing]

water cool

 

[Kenjiwing]

If you do not know what you are measuring, how do you know how hot it is supposed to be during load?

You need to find out what the specifications are on the measuring points.....before you go tearing off the heatsink and actually making it worse?
or applying a slightly conductive TIM and causing a short?

really? afraid to reapply TIM?

 

[magoo]

really? afraid to reapply TIM?

No not at all.
But....some are actually conductive....like AS and smeared on the wrong component would fuck up your nice video card. Which is why I use Noctura.....non-conductive.

Plus if you tear off the heatsink and foul all the thermal pads, and not having any replacements are shit out of luck until you can get some.

I just don't see the need to fix something that probably isn't broken......only reason I'd remove a stock heatsink would be to place a waterblock.....and those come with thermal pads.