Probable Memory Problems

[Tegeril]

So, system is in my sig. Past couple days, I've been finding that the system will lock up completely. It started with one core of the cpu pegging at 100% followed by applications freezing and then the mouse freezing. More recently (we're talking a few days total), the cpu pegging hasn't been happening, but the freezing has, with little to no predictability (once right after boot, another time, 12 hours later). The system will just hang there for an indeterminate amount of time, and then it reboots (no blue screen, no mention of anything in event viewer except windows stating that its most recent shutdown was unexpected).

I suspected a few recent software changes (nVidia drivers from 100.65 to 101.70, but I reverted and the problem remained and McAfee Patch 1 applied to VirusScan 8.5 Enterprise but uninstalling McAfee entirely yielded no greater success).

I then went to memory immediately. I had been running the chips at 4-4-4-12 @ 2.2v(rated 5-5-5-15 @ 2.2v), and I ran the Windows memory test in its standard mode with no problems. Changed it to extended mode and it will halt on 21% complete on a pass, state there was a problem, and then not proceed beyond that point. I went and reverted the memory settings back to stock timings and the same results occurred (halt on 21% extended, no problems standard).

I decided that I trust Memtest86 more, and ran that. It was on test #5...13% in, and it froze entirely (the plus in the logo kept animating and the blinking cursor under L2 Cache stayed blinking, but the time stopped incrementing and the program did not respond to any of its key commands).

Now, just to be sure, I reverted the cpu to stock speed, 266 fsb -- 2.66ghz and that forced the memory up to ddr2-1067 on this motherboard, but same timings as before. I'm now seeing the same problems with Memtest86+ on test 5, this time at 0%, totally halted, time not incrementing.

So in conclusion, what is the best way to determine which stick/sticks are at fault, and in which memory slots should I be testing -- should I remove one stick at a time and test the other 3, or should I test each stick individually, pairs?

Edit: Also, how much testing would be considered safe to determine if a stick is ok? (Full pass from the Windows extended test? All 10 tests from Memtest86?). I also just disabled legacy USB support in the BIOS as there were some reports via Google that this causes Memtest to freeze.

Thanks in advance everyone.

 

[mechBgon]

Test each of the four modules by itself. If all of them test faulty, what do you bet it's the motherboard that's the problem. If one tests faulty, try a pair of the remaining ones next.

If no single module tests faulty, test a pair of them. If that pair tests OK, move that same pair to the other two memory slots and test them again without changing anything else. Use a pair of slots that results in dual-channel operation, meaning either both of the yellow slots or both of the black slots.

If the test fails in the alternate pair of memory slots, now try to determine whether it's just one slot that's bad. That can be done with one module, or with a pair. You can also test whether just one memory controller is bad (slots A1 and A2 are on one controller, slots B1 and B2 are on the other).

If the pair does not test OK, test the other pair and go through the deductive process above using that pair.

Eventually you should have the possibilities sorted out. If it were me, I'd try not to push the voltage envelope any further than necessary on the RAM... 2.2 volts is 0.4V beyond the "official" DDR2 spec, so I'd try to back down a little if the RAM will run error-free at something lower.

 

[Tegeril]

This ram is rated specifically at 2.2v, I'll nudge it down a bit just to be safe in a minute after Memtest gets another chance on all four sticks. I set the motherboard to handle timings on its own and it has come up with 5-2-2-18...After that, I'll start single module tests.

 

[mechBgon]

If electromigration has occurred, lowering the voltage probably won't help anything afterwards. I don't know exactly what the RAM chiplets can handle for how long, but that much voltage might be unhealthy for them.

If you have any other DDR2-equipped systems, testing the modules in a different motherboard could also shed some light on what's going on.

 

[Tegeril]

Good to know, thanks for the advice 🙂 Going to get to it shortly.

 

[Tegeril]

Well, on to the single stick tests, Memtest froze at the exact point (test 5, 13%) that it did the first time I ran it.

Edit: This is going much faster with only 1GB to test 😉

 

[Tegeril]

Well..., each stick passed in the first slot. Testing sticks 2 and 4 in slots 2 and 4 now 🙁

Edit: that passed, testing those two sticks on ram in slots 1 and 3 together...and that passed too. Trying sticks 1 and 3 in slots 1 and 3... 🙁