HD 6950: LOCATION of temperature sensors?

[cmarino]

Hello,

how many temperature sensors does the 6950 physically have, and where are they located? I tried researching on the Internet but couldn't find it.

The reason I'm asking is because I just found out that GPU-Z comes up with THREE different temperatures, and what worries me is that one of the three is SIGNIFICANTLY higher than the other 2!



Does this look normal to you?

Since my 6950 is modded (I removed the noisy stock fan and replaced with a "fanless" Accelero S1 rev2.0, with added fans), I used to keep an eye on the Speedfan's "GPU" value (which was usually ok), but didn't know there were other sensors to check!

Can anybody tell me all the GPU-Z readings are accurate, and if they belong to ACTUAL hardware sensors on the board? (for example...: where is the "voltage regulator temperature" sensor???

Thank you!

 

[FreedomEclipse]

I cant recommend enough that you attach a 120mm fan at the bottom of that heatsink. 6950s are not the kind of cards that you can have a fanless solution with, they are just too powerful and create too much heat. If thats your cards idle temps - youre already overheating.

so unless you want a dead 6950 and a lengthly RMA period - bolt a 120mm fan on that cooler NOW

I just hope you havent been stupid enough to game on it in its current state otherwise you could have done more damage then good.

 

[erocker]

All three GPU temperatures are on the GPU die itself.

 

[cadaveca]

Temp 1 is shader cores.
Temp2 is Mem I/O
Temp 3 is Display I/O.

All are DTS sensors in the silicon itself, as erocker mentioned.

Temp1 is what will be displayed by CCC, is listed as GPU temp in GPU-Z, etc.

I'd be checking the cooler mounting/thermal-paste spread. The temps being slightly higher is normal, but not the 20c difference you got there, nope. Bad heatsink contact most likely.

 

[cmarino]

All three GPU temperatures are on the GPU die itself.

Good... Then, maybe, the reason of that "difference" means I need to reseat the copper heatsink properly on the GPU; that tiny Accelero S1 copper plate is not in complete contact with the GPU.


And... yes, OF COURSE i have a fan attached to the Accelero S1!! I have TWO Enermax 120mm ultra quiet thermistor fans attached to it. But even running both at full speed I keep getting #2 incredibly hot, while #1 and #3 don't even go over 60°.
There must be something... Probably, as cadaveca mentioned, it's a wrong GPU/heatisink match.

These S1 heatsinks have a very light structure holding them to the GPU... They're cheap and not very well made. The brackets even bend if you push them too hard, and I noticed that they must be custom bent to ensure a proper contact between GPU and copper heatsink!!

What a waste of Arctic Silver thermal paste tomorrow I'll clean and reapply the heatsink hoping that temps won't get even worse than before... I have the feeling that it will take me a long time reseating the sink and retesting the GPU temps before obtaining a proper match

 

[cadaveca]

I'm interested to see how this works out for ya. Keep us updated, PLOX.

 

[eidairaman1]

All three GPU temperatures are on the GPU die itself.

last time i seen a separate temp sensor was on an AMD Skt A Motherboard.

 

[cmarino]

last time i seen a separate temp sensor was on an AMD Skt A Motherboard.

Uhm, i believe even "modern" motherboards contain a separate temp sensor, embedded into a DFI's ITE Smart Guardian chip:

I thought the video card had something like this.

 

[eidairaman1]

Uhm, i believe even "modern" motherboards contain a separate temp sensor, embedded into a DFI's ITE Smart Guardian chip:
http://www.legitreviews.com/images/reviews/266/DFI_RDX200_ITE.jpg
I thought the video card had something like this.

im saying physically there not in an IC.

 

[W1zzard]

Uhm, i believe even "modern" motherboards contain a separate temp sensor, embedded into a DFI's ITE Smart Guardian chip:
http://www.legitreviews.com/images/reviews/266/DFI_RDX200_ITE.jpg
I thought the video card had something like this.

that's just the monitoring IC, the actual sensor is located in the cpu and connected to that IC. many of these ICs also measure their own temperature which is often called "motherboard temperature"

Temp 1 is shader cores.
Temp2 is Mem I/O
Temp 3 is Display I/O.

All are DTS sensors in the silicon itself, as erocker mentioned.

Temp1 is what will be displayed by CCC, is listed as GPU temp in GPU-Z, etc.

I'd be checking the cooler mounting/thermal-paste spread. The temps being slightly higher is normal, but not the 20c difference you got there, nope. Bad heatsink contact most likely.

while this is true for older cards (where gpuy shows those names), on newer GPUs i dont have any reliable information as to the sensor location (other than that they are located inside the gpu silicon).

thats why gpuy uses more generic names. other developers like aida64 chose to stick with the existing names, even though the locations might have changed

 

[eidairaman1]

that's just the monitoring IC, the actual sensor is located in the cpu and connected to that IC. many of these ICs also measure their own temperature which is often called "motherboard temperature"

Ya as on the SKT A CPUs the sensor is right in the middle of the socket itself.

The NB contained the system temperature (did direct temp gun measurement of both CPU and NB)

 

[hojoom]

Temp 1 is shader cores.
Temp2 is Mem I/O
Temp 3 is Display I/O.

All are DTS sensors in the silicon itself, as erocker mentioned.

Temp1 is what will be displayed by CCC, is listed as GPU temp in GPU-Z, etc.

I'd be checking the cooler mounting/thermal-paste spread. The temps being slightly higher is normal, but not the 20c difference you got there, nope. Bad heatsink contact most likely.

Hi i have R9 290 . GPU (1) is hot temp 80 c When I open network or games : 150 c temp ?

 

[hojoom]

Hi i have R9 290 . GPU (1) is hot temp 80 c When I open network or games : 150 c temp ?

Hi i have R9 290 . GPU (1) is hot temp 80 c When I open network or games : 150 c temp ?