Hello there, this is my first post
I possibly found the reason why so many cards die when a third party cooler is installed.
Today I had the opportunity to get my hands on a thermal camera (FLIR ThermaCAM E45).
So I used the chance and found a hotspot.
My case has a good air flow and I installed an Arctic Cooling S1 with two Turbo-Modules on my card. I didn't rip off the red plate.
Furthermore, I sticked four RAM-coolers on the VITECs as I thought it might help.
I measured the temperature after app. 10-20 minutes of running Furmark.
The listed temperatures are the ones GPU-Z displayed, in this order: 1. GPU, 2. VRM and 3. RAM.
Top-component side (?)
Bottom-PCB (?)
Top
66 / 76 / 62
The blue squares are the RAM-coolers, you can see they don't get hot.
Bottom
66 / 76 / 62
A bit later bottom-side
76 / 84 / 72
Unbelievable, 119°C though the readings are ok
No wonder some cards died.
The hottest spot of the card is exactly where the long chip can be seen on the other side respectively the space for the sticker on the PCB-side.
So I made changes, deinstalled the RAM-cooler as they didn't help and instead of the Turbo-Modules I installed a 12cm PWM-fan which rotates with 1200 rpm.
Top
76 / 85 / 72
Bottom
76 / 85 / 72
Though the temperatures GPU-Z displayed didn't improve the temperature on the PCB-side decreased to a tolerable level.
I assume the reason is, that the RAM-cooler don't hinder the airflow any more and the 12cm-fan blows more than the Turbo-Modules.
The fan of the cooler installed by AMD/ATI blows some air through the hole of the red plate and therefore cools the parts under it.
It seems to be important, that these components are cooled!
I'm happy my card didn't die while the tests, the modification works but should be handled with caution .
Update
Wow! Increadible what my first post provoked.
Thanks for the replies.
Due to the big interest I decided to reinstall the ATI-cooler for a test.
(Wasn't that easy, as I cut the plug to install the 12cm fan.)
Here are the results:
IDLE - Fanspeed 22%
Top
81 / 87 / 76
Bottom
After 10-30min of FurMark – Fan 31-34%
(after 10 min the temperature settled and didn't change within 30 min)
87 / 92 / 84
Bottom
So even with the original cooler the temperature is very at this place.
(fluctuated around 110 Grad)
Though not as hot as in the first variant with the RAM-coolers.
So don't do it this way:
The temperature at this place is near to the limit and might cause the death of the card if the ventilation is not sufficient.
The second variant with the 12cm fan shows that the S1 can cool the card better than the original cooler, as long as the airflow reaches the "Hotspot".
I think you can achieve that even without leaving the red plate installed.
Sorry, I didn't read all the answers. I'm no expert but had the chance to be in the posession of a thermal camera for a short time.
The emission-coefficient was set on 1 during the measurements.
A lower coefficient would even increase the temperatures, for instance you get a temperature of 147°C with a coefficient of 0.5, while getting 90°C with 1. But the temperature can't be that high.
The emission-coefficient of 1 only applies to black material without reflection, so I assume the real temperature is even higher than the measured one.
2. Update
In my opinion the 12cm Arctic-Cooling fan was a bit to weak (at 100% only 1200 rpm) and therefore I searched for alternatives.
I found the Papst 4412 F/2 GP, the only pwm-controlled 12cm fan that goes beyond 1500 rpm I found.
2900 rpm and an airflow of 170m³/h can blow away quite much if it's needed.
So here's a last test:
Accelero S1 with Papst 4412 F/2 GP
FurMark 20min
Fanspeed 30%
75 / 82 / 69
Top
Bottom
FurMark 10min
Fanspeed 40%
69 / 77 / 67
Top
Bottom
FurMark 30min
Fanspeed 40%
72 / 81 / 67
Top
Bottom
Only 93°C on the PCB-side after 30 min of Furmark is a quite good value.
Now I got a bit temperature clearance for overclocking.
At 30% fanspeed the fan can hardly be heard, in IDLE GPU-Z displays 44 / 52 / 39 °C at that speed.
At 40% the fan whispers and can be heard, but it's acceptable.
I attached the fan directly to the graphic card, so it is controlled automatically.
I set the fanspeed to a fixed value only for the tests.
But I had to edit the characteristic-curve with a BIOS-Editor, as the fan needs more rpm to run up. It still needs a bit fine-tuning, but I'm quite happy with the result up to now.
Now a bit additionally info for the ones, who think of a fraud.
IDLE
FurMark
Und JA, diesen Eisen-SPAWA-Kühler hab ich auch abgesägt, und mit vernünftiger Paste, wieder draufgeschraubt, hihi, viele im Forum meinen es wäre Aluminium, aber ich sage ES IST EISEN
Funktioniert trotzdem gut!!!
Denn es stimmt, für sich allein genommen und mit den Standardeinstellungen ist der Kühler / der Lüfter unzureichend.
