- last edited on by
ROGBot
Hey folks,
Over the past few weeks I had started seeing some really high idle temperatures on my ROG Mars 760x2 card, and figured that it might be time to throw on some new thermal compound.
After doing a little research, i picked up a tube of "Thermal Grizzly - Kryonaut" thermal paste/compound and did some before and after tests to see what kind of results I would get.
At the start, this was really just for me and I didn't have any intention of posting my results on here but they were pretty significant so I felt that this could be important to some people, especially those of us who would assume that ROG means you're also getting the best of the best when it comes to TIM.
For the testing, I used FurMark 720p GPU Stress Test.
Prior to beginning the testing process I restarted my PC then initiated the GPU Tweak II "Gaming Booster" which turns off unnecessary system services and defragments the system memory.
I have the GPU fans set to run at a constant 80%
You'll have to bear with me for these pictures, since some of them don't clearly show the temperatures from testing.
This first picture is taken in non-sli mode at IDLE (two PC screens running) which is my main mode, so each GPU is dedicated to its own screen.
I also included the CPU temperatures from AI Suite 3 to show the difference there and account for any potential differences in room temperature, though my CPU has a H100 on it. That said, the H100 will never cool greater than my room temperature, as per thermodynamics. No perpetual motion machines here 😛
This picture shows an IDLE temperature of 36 degrees Celcius across both GPUs.
This second picture is in the same view mode (non-sli) after running FurMark 720p GPU Stress.
Ignore the right hand side of the picture, the important stuff is the 75 Celcius maximum temperature indicated in blue.
(I couldn't get GPU Tweak to actually show that in the "Max:" thing.. not sure why, ask ASUS)
Unfortunately, I did not remember to take a picture of the idle temps in SLI (single screen) mode, but I recall them being similar to the non-sli.
Here is the FurMark stress test temperatures for SLI Mode Before the thermal compound changes, indicating an average maximum temperature of 80 degrees C across both cards (81 and 79)
AFTER
Now to replacing the compound.
WARNING: If your MARS 760 is new, this may void the warranty as there did appear to be a "warranty" removal sticker on one of the cooler screws.
If you're okay with this, you'll have to remove pretty much every screw on the board, including the two on the silver plate by the DVI ports.
The fan/LED cables are tucked under the MARS logo area, so be careful when separating the board from the cooler.
I performed a two step Arctic Cooling solution cleaning on the card, then applied small amounts of thermal paste in an x pattern.
I opted for the x pattern because unlike a CPU which has an IHS, the GPU dies on this card are exposed and complete coverage is necessary, where as a small dot in the center of a CPU IHS typically leaves the corners uncovered but is not a problem.
After application and reassembly, I performed three 10 minute burn-in tests without recording the results to set the compound (though this is a no cure compound, so it wasn't really necessary)
Paydirt.
My IDLE temperatures in non-sli mode (similar process, 80% fan, gaming booster) were an astonishing 23 degrees celcius.
While my room temperature has not significantly changes (personally it feels warmer), this does not seem like a realistic result.
The weird part is that this IDLE temperature was after the burn-in tests, so I cannot even claim that it was just because my PC was cold.
I will update in a few days to see if I can still achieve similar temperatures.
The non-sli stress test yielded a much more realistic result with a maximum temperature of 71 degrees celcius, a change of about 4 degrees from before the reapplication.
The SLI stress test yielded a maximum average temperature of 76 C (75,77) Once again, a change of about 4 degrees.
Conclusion:
I know there will probably be quite a few of you on here to **** on my methodology and what not, my intention was not to post this for you guys but merely to see what kind of results I would get.
I would be entirely satisfied if I could see a similar change in the idle temperature of 4 degrees, but would be thrilled if that 23 C result stayed consistent.
Over the past few weeks I had started seeing some really high idle temperatures on my ROG Mars 760x2 card, and figured that it might be time to throw on some new thermal compound.
After doing a little research, i picked up a tube of "Thermal Grizzly - Kryonaut" thermal paste/compound and did some before and after tests to see what kind of results I would get.
At the start, this was really just for me and I didn't have any intention of posting my results on here but they were pretty significant so I felt that this could be important to some people, especially those of us who would assume that ROG means you're also getting the best of the best when it comes to TIM.
For the testing, I used FurMark 720p GPU Stress Test.
Prior to beginning the testing process I restarted my PC then initiated the GPU Tweak II "Gaming Booster" which turns off unnecessary system services and defragments the system memory.
I have the GPU fans set to run at a constant 80%
You'll have to bear with me for these pictures, since some of them don't clearly show the temperatures from testing.
This first picture is taken in non-sli mode at IDLE (two PC screens running) which is my main mode, so each GPU is dedicated to its own screen.
I also included the CPU temperatures from AI Suite 3 to show the difference there and account for any potential differences in room temperature, though my CPU has a H100 on it. That said, the H100 will never cool greater than my room temperature, as per thermodynamics. No perpetual motion machines here 😛
This picture shows an IDLE temperature of 36 degrees Celcius across both GPUs.
This second picture is in the same view mode (non-sli) after running FurMark 720p GPU Stress.
Ignore the right hand side of the picture, the important stuff is the 75 Celcius maximum temperature indicated in blue.
(I couldn't get GPU Tweak to actually show that in the "Max:" thing.. not sure why, ask ASUS)
Unfortunately, I did not remember to take a picture of the idle temps in SLI (single screen) mode, but I recall them being similar to the non-sli.
Here is the FurMark stress test temperatures for SLI Mode Before the thermal compound changes, indicating an average maximum temperature of 80 degrees C across both cards (81 and 79)
AFTER
Now to replacing the compound.
WARNING: If your MARS 760 is new, this may void the warranty as there did appear to be a "warranty" removal sticker on one of the cooler screws.
If you're okay with this, you'll have to remove pretty much every screw on the board, including the two on the silver plate by the DVI ports.
The fan/LED cables are tucked under the MARS logo area, so be careful when separating the board from the cooler.
I performed a two step Arctic Cooling solution cleaning on the card, then applied small amounts of thermal paste in an x pattern.
I opted for the x pattern because unlike a CPU which has an IHS, the GPU dies on this card are exposed and complete coverage is necessary, where as a small dot in the center of a CPU IHS typically leaves the corners uncovered but is not a problem.
After application and reassembly, I performed three 10 minute burn-in tests without recording the results to set the compound (though this is a no cure compound, so it wasn't really necessary)
Paydirt.
My IDLE temperatures in non-sli mode (similar process, 80% fan, gaming booster) were an astonishing 23 degrees celcius.
While my room temperature has not significantly changes (personally it feels warmer), this does not seem like a realistic result.
The weird part is that this IDLE temperature was after the burn-in tests, so I cannot even claim that it was just because my PC was cold.
I will update in a few days to see if I can still achieve similar temperatures.
The non-sli stress test yielded a much more realistic result with a maximum temperature of 71 degrees celcius, a change of about 4 degrees from before the reapplication.
The SLI stress test yielded a maximum average temperature of 76 C (75,77) Once again, a change of about 4 degrees.
Conclusion:
I know there will probably be quite a few of you on here to **** on my methodology and what not, my intention was not to post this for you guys but merely to see what kind of results I would get.
I would be entirely satisfied if I could see a similar change in the idle temperature of 4 degrees, but would be thrilled if that 23 C result stayed consistent.
