I just encountered one of the strangest problems I've had in my 25 years of computer use. As to why I am posting this in the cpu section, that will become obvious shortly. I thought that maybe someone might benefit from my resolution of this problem. I've condensed the story into a few paragraphs, but note that these events took place over a three-day period. Also, popcorn is being served in the main sitting room.
The machine is my wife's--she always gets my hand-me-downs. She wanted to play a 3d game, so I installed one of her Agatha Christie games for her and bid her good luck. Ten minutes later she called me over and said (paraphrased), "Walt dear, you unmitigated hunk of resplendent and gorgeous man, the machine has just turned itself off, so fix it before you see the wrong side of this roller pin I have in my hand!" I walked over and hit the power switch and the system booted right up. She opened the game again and, ten minutes later, called me back over because the machine had mysteriously powered back down--just as if someone had yanked loose the power cord--no bsod--just a simple power off--without bothering to properly shut down, of course. At this point I began to analyze the situation. The shut downs never occurred while browsing or doing anything Windows/browser/non-game software related--I could do that for hours without a problem--they only occurred when running a 3d game for approximately 5 minutes. And I mean 5 minutes--it didn't matter what game, what settings I used, what resolution, etc.--you could time the shutdowns by the clock. Long ago, when I used the machine myself at home as my primary box, I never had this problem or anything remotely similar.
System as follows:
Asus A8N5X motherboard with the 1003 bios installed--latest version from Asus
Athlon x2 4200+ running at stock 2.2GHz
Windows 7 HP
3 gigs of DDR 2, 266MHz ram configured as 2 one-gig DIMMs in Bank A, 2 512MB DIMMs in bank B.
BFG GeForce 8800GT OC in the board's only PCIe x16 slot (factory overclocked 25MHz above stock)
nForce 4 core logic
Audigy 2 PCI sound card
Linksys wireless PCI adaptor card
1 DVD/CD Writer configured as IDE 0 Master
1 300-gig Sata hard drive, configured as Sata port 3 Master
Antec True Power 430W
At first it seemed that heat was the likely culprit, so I used Riva Tuner to rev up the 8800's fan to 100%, and took off the side cover. It didn't matter how many games I tried--I even used a powerful portable fan to blow air into the case directly onto the gpu--every 5 minutes the machine just turned itself off. At that point I ran the temperature sensors in RT and from the resulting logs after each power off I could see that the 8800GT was topping out at ~70C, which really was fine, and much lower than I thought it would be, as the darn card would nearly burn me when touching it. According to the thermal sensors, maximum 70C under load gpu temps should not be a cause of a power off condition.
At this point I figured that either the 8800 GT was in some way defective and needed replacing, or else it might be that the Antec PSU wasn't able to handle the load. I voted for the PSU and picked up a 500W Earthpower Antec and installed it. No dice--system would power off every five minutes when playing a 3d game. The PSU did not seem to be the cause. I replaced the 430W and returned the 500W PSU. I didn't immediately buy another gpu because I wasn't satisfied that the culprit was the 8800GT--I wasn't getting blue screens with driver errors, there were no screen artifacts to see of any kind--the system was just shutting itself off for some reason while playing a game.
Next I decided to check out my ram by running Win7's memory test. At this point I noticed something odd--running the advanced test, the test would stall at 21%, although the machine itself did not lock up and I could exit out of the test via the keyboard, but this was the case even when I tested one DIMM at a time, two at a time, or all four. Then when I ran the basic test, one DIMM at a time, or in any other combination, I experienced the power-off symptoms I'd been getting in 3d games. Although it at least superficially appeared that I had some suspect ram, I wasn't really ready to accept that because it didn't seem likely to me that all four DIMMs had gone bad at the same time. Also, I had run extensive testing with the ram set at slower speeds and timings--to no avail.
It was at this point in the process that I began thinking about the cpu. I thought it might be a good idea to bump up the cpu voltage just a bit to see what happened, and set the voltage to 1.5v from its default of 1.37V. I rebooted and before I could get into the bios to check on the system temps and voltages, the machine just powered off. Another reboot and again the machine powered off before I could get into the bios. I reset the CMOS and booted back up successfully. Reseated my ram, gpu and other cards for the umpteenth time, checked all of my cabling, checked everywhere for dust, etc., and decided to see if bumping it up to 1.5v would shut me down again--or whether the first shut-downs were a fluke. Changed to 1.5V in bios, rebooted and...shutdown before I could get back into the bios. This time I reset CMOS immediately and successfully rebooted.
By then I had decided that in some fashion the cpu was the culprit--of course it would have to be the very last thing I checked! I powered down, unplugged the PSU from the power strip, and turned the PSU switch to off. When the last motherboard standby light winked out, I removed the heatsink and then carefully retrieved the cpu from the insertion slot. At this point in the retelling things get a little hazy--I noticed that there seemed to be an inordinate amount of Arctic Silver thermal paste on top of the cpu. It really was gooped, like a thick sludge. Some of it had dripped down and inundated a few pins on the bottom of the cpu. Immediately I set out to remove the goop from the pins and that's when I noticed that several of the cpu pins were bent at severe angles, and that's also when I became somewhat pissed--at myself, of course. I still cannot say whether the pins were bent before I pulled the cpu from its slot, which would mean that when I had installed the cpu into this slot better than a year ago I had carelessly put it in while somehow managing to bend several pins at the same time, or whether in my haste to remove the gooping, slopping Arctic Silver paste from the pins in the present I inadvertently bent them--which would also have been my fault since I had apparently put far too much paste on top of the cpu when I had first seated it in this motherboard.
I whipped out the old magnifier glasses, got a very small screwdriver, and proceeded with care to straighten every one of the bent pins while at the same time trying mightily not to break even one of them... Oh, joy that was fun! Somehow I succeeded to the degree that the pins finally all looked more or less straight enough to seat, and then I set about cleaning the excess arctic silver from the cpu and the heat sink with 70% isopropyl alcohol and a huge reservoir of patience. Finally got everything dry enough and clean enough to risk reinsertion and powering up. The cpu appeared to seat with no problems as I had checked a couple times to ensure no pins were being bent. This time I used but a very thin coat of Silver, and locked the heat sink back into place, double checked all wiring--and hit the power on switch and held my breath. Either the machine would short itself into oblivion, or it would work like it should, or, and this was my biggest fear, it would successfully boot and run only to continue powering off every five minutes inside a 3d game.
Well, call my monkey's uncle a rogue and a Snipe, it worked! By George, the damn box was working as it should. No more powering off--at all. Ram tests complete without a problem. Now the machine works as it did before I turned it over to the wife! The system was once again capable of running 3d games normally, and I didn't have to replace a single component!
Epilogue: I still don't know whether the powering off problem was caused by mangled cpu pins that had been mangled ever since I installed this cpu into this motherboard, or whether the Arctic Silver had spilled out from the top of the cpu and run down the side of the cpu to temporarily short a couple of the cpu pins, and they became bent in my frantic efforts to clean the paste from the pins when I discovered it. I'm sure I did have a short of some kind--one that was serious enough to cause the machine to power down due to cpu conditions that occurred during gaming--but that did not occur when doing normal, everyday, mostly 2d computing. A short among the cpu pins would also explain why the machine powered off almost immediately after I tried jumping the voltage a mere .13v.
I hope this story might help someone with a similar problem. All's well that ends well. Please kindly recycle the popcorn bags on your way out!