OSDev.org

Well, it's like this...

Last christmas I bought this CPU, AMD Athlon 64 X2 6400+ (the standard "125W" edition). Vcore and CPU clock - all default, no overclocking.
For 3/4 a year, it was working flawlessly until July/August 2008, when I got system crashes every five minutes in gameplay - BSODs regarding something like "ati2dag.dll" - graphics driver. I installed newer, older versions of it, uninstalled all programs that could cause conflicts with it (RivaTuner etc) and I even re-installed Windows. Nothing fixed, of course. I was desperate that the graphics card driver was the culprit, but I hadn't in mind to open the hardware monitor, CPU: 62°C in idle, more than 70°C in full load , fan spinning at 3100RPM. I checked the heatsink and fan, everything OK. I reapplied the thermal paste and reseated the heatsink. It was correctly seated afterall, almost same temperatures - 59 or so in idle, 68-75°C in full load. Crashing again. So I took a vacuum cleaner and cleaned everything in the case with it, and the temperatures under load fell to 65°C. The "maximum temperature" as stated by AMD for this CPU is 55-62°C.

That happened between July and September. Because school has started at the beginning of Sep, I stopped playing games and I just used the PC for internet or smaller games. I thought the problem was fixed, but nope. Same crashing even today, but not BSODs. Just regular "windows has experienced an error and needs to close" stuff all over again. So I downloaded two bench programs, OCCT Perestroika and System Stability Tester for win32 (from sourceforge). I've ran OCCT, and after the first 20-30 seconds, it crashed. . I've experimented with various vcore my CPU can handle, through 1.35V to 1.40V I have set them in BIOS. Crashing, crashing, crashing. Temperature reported was 65°C again, but I've seen it crashing at 70°C or more - CPU has been damaged for sure.

I've read topics that the "auto Vcore" for this CPU in the BIOS utility doesn't work as it should, so people just set it to 1.375V. They reported that it was jumping up to 1.41V (max vcore from AMD for this CPU, is 1.4V).

My guess is, that the CPU is damaged from overheating. The last thing I was left to do - set the CPU multiplier from 16x to 15x, thus slowing from 3.2GHz to 3.01GHz and I've set the vcore fixed to 1.625V and I've ran OCCT. No failure for 15 minutes - compare that the original 3.2GHz setting crashed in 20 seconds. The maximum temperature under load is now 62°C and seems to be stable.

Now I'm testing it for one hour and report if it crashed or not. Let's hope it will be stable.

Sorry for a long topic, I am just curious for your response, if anybody, even non-overclockers, did experience this... stock HSF combo, installed properly and CPU overheats...

//EDIT: System was fully stable during the test, nothing crashed whatsoever plus I even browsed the internet. Maximum temperature during the test, reported by OCCT, was 65°C. I'm adding OCCT's output - temperature graph on both cores, in the forum attachment.
Let's hope the CPU will survive at least next three years. My old Pentium 4 (socket 478, Northwood) machine runs at full load 63-65°C without any errors regarding temperature, plus no underclocking/undervolting whatsoever...

CPU: 62°C in idle, more than 70°C in full load

My computer, which I built, was running fine, until I got my new graphics card, which I put in. Then, it crashed (by just turning off) in about15 seconds.

If I let it sit for a while, it lasted longer. I finally made it into the PC health status in the BIOS, and it went
CPU Temp: 110
CPU Temp: 111
CPU Temp: 112
And then crashed.

I took the cpu off, blew all the dust out everywhere, and replaced the CPU, and it works fine now.

This problem I had encountered before, and it turns out that the monitor has to be damaged, because I was (with my computer running) getting temps of over 90C. Intel said that that was impossible, as the CPU would've melted or been permanently damaged.

So, yeah. But you're doesn't sound like a monitoring fail, because it worked to decrease values...

-JL

All AMD CPUs released after the (Socket A!) Athlon Palomino (I may have spelt that wrong) core have included a thermal automatic shutdown process, which instructs the motherboard to kill the power right now, which shuts down at temperature levels below which the core will suffer damage. Unless this is faulty, which would be quite unlikely, then the CPU is not overheating. The thermal shutdowns are very distinctive - your PC just turns itself off.

Have you checked your motherboard can support your CPU? Not all Socket AM2 motherboards can supply the 125W/135W models - only those with beefy power circuitry. If the power circuitry cannot support your CPU, then you would experience crashes, and the circuirty would likely be permanently damaged - a process which could take time.

If you have any other AM2 motherboards, I'd also try swapping the CPU with them, if they can take the power load.

65C and above is not unheard of for a chip's operating temperature - my (original 320-bit bus) 8800GTS idles at 72C - which I agree is ridiculous (And I'm providing it with ridiculous ammounts of air! I have both 160mm and 80mm fans blowing into it's intake! I suspect the leafblower may have trapped lots of dust...). nVIDIA specifies a slowdown threshhold on that of 110C though.

Note also that there might be faults on the motherboard that trigger only at specific temperatures.

Just another thought - your BIOS supports your processor, right? Because I've had graphics driver related crashes when the BIOS didn't properly support the CPU inserted.

Yeah, it's all supported.

I did check that...

-JL