Yesterday, just to see that I can take the heatsink off, I tried to twist and pull it right after running the OCMT for a minute or so. Of course I turned off the computer and unplugged it before pulling the heatsink. It did come off very easily. Then I orders Arctic Silver 5, but it takes time to get it, you know, and the local stores have only the low quality thermal material. I was kind of impatient and wanted to test a couple of other things. So guess what I did. Hehe. I wiped off most of the compound on the surfaces of the CPU and heatsink and attached the heatsink back into its position (aren't I crazy? DON'T TRY THIS AT HOME!) And then I put an industrial strength super fan in front of the stock fan of AMD and started it. And then I turned on the computer. It just started as if nothing has changed (although there was almost no thermal compound between the CPU and the heatsink) and then the softwares reported the same temperatures as before. OCCT test failed after the first minute because of high temperature (>62).
Then I turned off the super fan while the computer was on, and watched the temperature but nothing had changed. I tell you, AMD's stock fan is real good. Anyway, while risking my 200 dollars or maybe even more, I realized that maybe the problem was not due to the thermal compound, or the fan or the heatsink... Then I checked the CPU voltage with SpeedFan and OCCT, and it was set to something around 1. But after running OCCT, the voltage was shooting the 1.60-2 range in the first minute. The recommended range of voltage for my CPU is stated to be between 0.825 and 1.40
So obviously the CPU voltage volatility was the problem. But why was it volatile? Because of the EVIL AMD Cool'n'Quiet option in BIOS which is enabled as default, as if it does an awesome job. And they don't know how EVIL it is. EVIL. The general idea of this Cool'n'Quiet is, as we know, to set a low voltage when not much task is around, and increase the voltage if more work is needed. But then I understood that when it gives very low voltage to the CPU, Windows cannot even start up and HENCE is "the restart loops" problem mentioned in my initial post. And when CPU is loaded by OCCT, or other regular tasks, it gives a very high voltage to the CPU, HENCE the temperature gets very high, BIOS starts beeping, and Windows eventually gives a blue screen and the computer shuts down, which were again mentioned in my inital post as the pieces of the problem.
So what I did was: I turned off the EVIL Cool'n'Quiet option in BIOS, set the CPU voltage manually to 1.40. I got a very good idle/full loaded (34-56 C) temperature range with this voltage but I wasn't satisfied, so I set the CPU voltage to 1.225 (the lowest stable under full load). With this setting OCCT auto test runs for 1 hour without giving any errors. The min and max temperatures are 32-52 C. Isn't it amazing???
Now the computer runs without any problems or beeps, idling around 32, and I am a happy man.
And ah, Arctic Silver 5 is coming tomorrow and I will put some there for sure (yes, just a little). Hopefully it will decrease the temperatures even more like a degree or two...
I wonder how AMD guys don't know about this problem with the Cool'n'Quiet and let people use it as default...
07:32 AM by