This kind of problem cannot be definitively resolved by E-Mail based on a 100 word description, but we can take a stab a likely causes. While an engineering lab could invest tens of thousands of dollars in diagnostics and engineering services and pinpoint the exact problem, as an end user, you really are limited to a few diagnostics and then component testing by substitution. But here is how I would proceed:
00. I'm omitting software considerations, because what you have already done (reinstalling Windows clean) suggest that this is probably not a software problem. That's not QUITE good enough to rule out software issues 100.0%, but it's all I can do as far as responding to you request for assistance. However, before going further, definitely run "Windows Update" to install all OS patches, and get and install the latest drivers for all of your cards, especially your video and sound cards (or equivalent circuitry on the motherboard).
0. Before doing anything else, remove non-essential expansion cards, if you are overclocking anything STOP IT, and reset all BIOS parameters to their "default" values.
1. Test memory. TEST IT, DON'T JUST REPLACE IT (the "new" memory could be bad also). You need a good memory test program, either Memtest or Memtest-86 (two similar but now different programs that evolved from a common base). In your situation, run it overnight, and remove the media (disk ... floppy or CD) so that if the system does reboot, it does NOT continue running the memory test program and you know that it rebooted. This is one of the top causes of hardware problems and the easiest to test (note, however, a reboot during the running of the memory test where there had been no errors reported by the test does not specifically suggest a memory problem).
2. Check the power supply. Measure the voltages with a multimeter. Good readings don't tell you much, but a bad reading (more than 5% off nominal) suggests a power supply problem although most systems will work with a power supply that is 10% off.
3. The real problem is that you can't measure power supply transient response with any tools available to you. That is, the CPU might need a "surge" of over 20 amps but only for 1 nanosecond (billionth of a second). If the power supply can't supply that, the CPU crashes, which can take the form of a restart, but which may show up as just a locked up system also. There is no way for you to test transient response, so even if the voltages seem ok, consider buying or borrowing a known good replacement power supply, as this is one of the top 2 causes of the kind of problem you are having. DO NOT BUY A "CHEAP" POWER SUPPLY. Buy a brand name, high quality supply. You are going to have to spend at least 10 to 20 cents per watt. But a GOOD 350 watt supply will easily outperform a cheap 500 watt supply no matter what the rated requirements of your system. And to determine what you need, use a few different online power supply calculators (do a google search), average their results, then add 25%.
4. Double check the mounting of the CPU and heatsink. CPU overheating can cause restarts, which can be caused by improper heatsink mounting as well as dirt (dust) buildup around the heatsink & fan. Blow out the heatsink and fan with compressed air, but the real key is that the heatsink is properly mounted to the CPU with a properly applied "thermal solution" (may be either a heatsink compound or a "phase change pad"). Note, too much thermal compound is a problem also, the right amount is about 1 or 2 grains of rice, and that is about all. If you take the heatsink off, clean both the top of the CPU and the bottom of the heatsink completely with solvent to get them both smooth (DO NOT use an abrasive), the reapply new thermal compound. Thermal pads are one-time and should not be reused.
5. Now it's tough, because we are past all of the easy, cheap problems. CPUs rarely fail. Can't say never, but it's rare. The next most likely culprint is the motherboard, and obviously that is expensive and difficult to test, and you can really only do it by buying a new motherboard, which is tantamount to getting a new computer. Further, this will in some cases will require you to "reactivate" your software, which reactivation might be denied because in the software vendor's view, it IS a new system ..... an issue that goes way beyond the scope of this response. Unfortunately, if you have eliminated software, memory, the power supply, the CPU and it doesn't look like the video card or disk drive ..... well, there just isn't much left.
Hope this helps. This kind of problem can be very difficult to resolve and can take a LOT of time and effort, in part because even if you do fix it, or after every attempt to fix it, you may not really know if it's fixed for a significant period of use.
By the way, now I am going to get controversial here and come right out and say something that a lot of people think, not many will say, and some will disagree with vehemently: You have an AMD system, and people who value stability and reliability (and, at the moment, even performance) are better served by an Intel system. The issue isn't so much the CPUs, but chipsets and motherboards. Intel simply makes better chipsets (and, at the moment, I personally believe, CPUs also). Further, AMD systems generally are "cheaper" and more of the motherboards are made by "2nd tier" motherboard makers. Intel systems (and I mean systems with an Intel CPU ***AND*** an Intel chipset) ARE more expensive, but in my opinion (and that's all this is) they are simply more stable and more reliable. That said, I make this statement as a degreed engineer with over 40 years of computer experience (1967), a college instructor in Information Technology, A+ and Network+ certified and a holder of US patents on motherboard circuitry. Others are going to disagree, that's fine, this is all a bit controversial, I recognize that, but that is my view based on my experiences using, building, servicing and teaching computers for decades.