(I checked twice and did not mess up GPU<->CPU, dont worry)
(***Skip to conclusion if you dont dare reading interesting information***)
When I boot windows (or my Linux Gentoo):
Pretty often get
I remember to long ago have been playing for several hours on every setting enabled (BFBC2), but then:
Replaced CPU Thermal paste with some new in the *wrong way* and my CPU heated(81Celsius) a bit... But nothing happened then. Maybe that was totally unrelated but just saying.
However, that was not gaming, rather compiling firefox10 (which is pretty huge), and the GPU was not affected in any way(or not that I know of, that would be really strange).
But now its gotten so far that it happens in pure OS, without any game running nor anything working, just idle. (both windows and linux. Even happens in Linux TTY).
The thing is, I can switch to another TTY and it goes away exceptionally (not always) and everything is normal. So no CPU bad, just GPU.
-> CPU got once above 81 Celsius.
-> Crash happens even with cool GPU (very very minor 2D desktop decoration...)
-> I can get seldomly out of it with ctrl-alt-del, so no CPU fail.
-> Started to happen after CPU overheated.
-> The heck, are GPU and CPU in any way connected to each other? Maybe missing case fans?
Still I get those errors/crashes as shown in the similar picture without doing ANYTHING, not even firefox started.
I tried resetting BIOS, its reset now. I reset all settings in AMD Catalyst 12.x driver software, and still.
Your graphics card is either faulty or is not sitting perfectly solid in its socket, but its more likely that its faulty. Remove and reseat the graphics card and if it continues to happen replace it. I would recommend that you try it in another system and/or put another "known good" graphics card in your system to verify.
I remembered to have overclocked my RAM to 2000mhz+, which is more than my 1600mhz ram.
Of course, no hardware component, neither BIOS, could initialize and after a few auto-restarts it reset the BIOS settings. Might such stuff cause those errors ? I am running memtest for 28 Minutes now, and exactly 75% of first pass is done without any errors.
1. Test a different card in you system, and
2. *Test your card in a different system
*PSU requirements for graphics card:
#1 - The PSU must meet the overall WATTAGE, use an online calculator
#2 - The PSU must supply the required AMPS on it's +12V rail or combined rails. (May need to search a bit. Example a GTX570 may require 38Amps. Get a PSU with at LEAST 45Amps on its +12V rail/combined rails).
1. run MEMTEST as mentioned, but rerun if you change MEMORY settings in the BIOS
2. CPU diagnostics (PCMark95 multi-core or similar app. Google.)
3. FURMARK for graphics (don't bother yet as your crashing in idle, but use this LATER when things appear stable).
4. Try a different PCI-e slot, if available.
5. BIOS. Set your CPU and MEMORY to the normal settings.
1600MHz RAM does not need to be overclocked in your setup. Your CPU just isn't fast enough even overclocked at 100%.
a) make sure your system is stable at default settings (test CPU and RAM)
b) overclock the CPU
c) now adjust the multiplier for the RAM so it then correctly drops back to 1600MHz
d) rerun Memtest
e) boot to Windows and rerun the CPU stress app (CTRL-ALT-DEL -> procesess, to make sure all cores stressed to 100%).
f) finally run Furmark and 3DMark (any series)
g) Make a BACKUP of your Windows partition (I make weekly, automated backups, using Acronis True Image to a second hard drive. A FREE version of Acronis TI exists for Western Digital and Seagate. It's manual but quite easy to use.)