[SOLVED] Cooling ATi Radeon AGP video card (front and back)

13hm13

Distinguished
Aug 14, 2013
34
0
18,530
I have an old (2010) ATi video card -- ASUS ATi (AMD) Radeon HD 3450 512mb AGP DVI/VGA/HDMI Graphics Card -- that is still working (PC is always on!!) .

1145077.jpg


GPU temps (per my PC utility SpeedFan, report 35-41C avg. )

Twice a year, the fan motor needs oiling, and the fans and heat-sink needs blowing out.

On my motherboard, the fan side faces down and this side is face up:

s-l1600.jpg


Notice the white square pad with the smaller white rectangle protruding thru the cutout. I think this is the bottom of the same (??) main GPU that is on the fan side.
I put my finger on the small white rectangle and it feels very warm. Using an IR thermometer, I measured ~52C on the white rectangle.
So what exactly is the white rectangle, what's that soft white square pad around it, and why isn't it heat-sinked?
 
Solution
But what is that "stuff" on the non-fan side?

Also a weird reporting anomaly ...
In Windows, using SpeedFan utility, the same GPU's readings (when PC is idling, with min, load) are 7-8C lower than when I'm running Linux (using any Linux sensors utility, the GPU temp is 7-8C warmer than in Win, when PC is idling, with min, load).

If I had to make an educated guess, probably an integrated chip providing a PCI/PCIe to AGP interface. The GPU is new enough that it was probably built with a PCIe interface on silicon, and they needed the additional chip to make it backwards compatible with AGP.

13hm13

Distinguished
Aug 14, 2013
34
0
18,530
52C isn't very warm for electronics, certainly well above skin temperature which is why it seems hot.

I wouldn't worry until you see something like 85C and up.
But what is that "stuff" on the non-fan side?

Also a weird reporting anomaly ...
In Windows, using SpeedFan utility, the same GPU's readings (when PC is idling, with min, load) are 7-8C lower than when I'm running Linux (using any Linux sensors utility, the GPU temp is 7-8C warmer than in Win, when PC is idling, with min, load).
 

kanewolf

Titan
Moderator
But what is that "stuff" on the non-fan side?

Also a weird reporting anomaly ...
In Windows, using SpeedFan utility, the same GPU's readings (when PC is idling, with min, load) are 7-8C lower than when I'm running Linux (using any Linux sensors utility, the GPU temp is 7-8C warmer than in Win, when PC is idling, with min, load).
It looks like thermal tape. But every picture I have found of that card, even "new" cards have that same configuration. I thought maybe there was a heatsink that had been thermal taped on, but I can't find any photographic evidence of that.
 

13hm13

Distinguished
Aug 14, 2013
34
0
18,530
It looks like thermal tape. But every picture I have found of that card, even "new" cards have that same configuration. I thought maybe there was a heatsink that had been thermal taped on, but I can't find any photographic evidence of that.
I bought mine new. It has been that way since I took it out of the wrapper, Yes, all other photos I've found show same white pad. Heatsink and pad are on the other side. It might be flat "micro" heatsink -- many phones and SBCs are like that.
I did use some double sided thermal tape to affix a small heatsink to that surface. But I'm only seeing a 3C drop in GPU sensor temp and (perhaps) a 7C drop in IR thermo when I laser the taped heatsink interface.
 

Eximo

Titan
Ambassador
But what is that "stuff" on the non-fan side?

Also a weird reporting anomaly ...
In Windows, using SpeedFan utility, the same GPU's readings (when PC is idling, with min, load) are 7-8C lower than when I'm running Linux (using any Linux sensors utility, the GPU temp is 7-8C warmer than in Win, when PC is idling, with min, load).

If I had to make an educated guess, probably an integrated chip providing a PCI/PCIe to AGP interface. The GPU is new enough that it was probably built with a PCIe interface on silicon, and they needed the additional chip to make it backwards compatible with AGP.
 
Solution