Need expert for difficult computer problem - random crash/lockups

I've been building PCs for my personal use for almost 20 years, and my latest HTPC I built about 8 months ago. It's fairly basic, a Gigabyte motherboard with 8GB RAM (2 sticks), a Sandy Bridge i3, a BD drive, and a RAID card with 3 attached HDs in a case with an Antec Earthwatts 380W PSU. It ran brilliantly for about 6 months and nothing has been changed with the hardware.

Starting about 2 months ago I began experiencing random lockups. Felt like it was something I'd installed since the hardware was pretty new and had been stable with no changes, but I've since reinstalled Windows from scratch multiple times, run memory tests, sfc /scannow, etc. and I get no obvious errors, but the machine is still locking up. Here are the symptoms:

- Lockup can happen minutes or hours after boot, but often takes many hours up to a couple of days
- Happens whether I am using it (say watching a movie) or when it is idle
- Usually screen is completely black when I try to access it (and it doesn't respond to network ping or other connection attempts), but sometimes the screen is showing but garbled up
- Requires a reboot to fix
- Often, but not always, the date in Windows after such a reboot reflects current day date, but year showing is now 2000 (and enabling or disabling Internet time has not helped address this behavior)

Anyway, the inconsistent period of the lockups, the black vs. garbled screen, and the inconsistent date issue have me completely puzzled. i'd think if it was the CMOS battery then whatever is screwing with the date would send it to 1/1/2000 and i'd get some CMOS alert in the BIOS at start, and if it was the CPU or PSU the date would be a non-issue. So still leaning towards the battery or the mobo, but wanted to see what the experts here might recommend before I start RMAing everything. THANKS much for your help!

Are you monitoring the temps on your CPU cores to make sure the CPU is not overheating?

Yes, and it sits about 48-49 when not under load, and usually gets up to around 60 C when under loads like watching a movie and playing music simultaneously.

What about your graphics card temps?

If you haven't already done so, you'll want to configure your computer to generate a memory dump when it freezes/crashes so that it can be analyzed:

Dump Files - Configure Windows to Create on BSOD

Also, you might want to try running a system health report:

System Health Report - Vista Forums

It could very well be a corrupt driver, which can be verified checked by running "sfc /scannow" from an elevated command prompt. When you reinstall the OS, you must load the proper drivers from either the manufacturer's website or use Windows Update to get them. Sometimes, one or the other will cause an issue.

I am using the integrated video only, and the sensor temps all appear fine. I have run sfc /scannow from an elevated prompt repeatedly and it has never turned up an issue. It is non actually going to a BSOD or crashing in that sense, more of a lockup with or without a blank screen, and so I've got it configured to capture a dump but it doesn't generate one. i forgot to mention I've checked all the event viewer logs and no strangeness there, either. I ran the perform as instructed in the link you sent me but it shows no flags other than the fact I have UAC disabled and I don't auto-check for windows update (both intentional, and my windows updates are up to date).

This is all making me believe more and more it isn't software, but hardware. Agree? Ever seen a system clock do this kind of behavior?

When you reloaded Windows, where did you get all of the drivers for your system? What software do you have loaded on your system? It's not advised to disable UAC as there are rarely reasons to do so that offset the risk added in doing that.

It really could be either hardware or software related. How have you run your memory tests and what have you used to test it? Does your PC get proper ventilation?

That software doesn't test memory from the description on the site and what their screenshots show.

Memtest86+ is pretty much the defacto standard for testing memory. The sticks don't initially need to be stress tested separately. You'd only need/want to do that after errors were found or if you were trying to test them by just using the system.