What are the odds of buying bad RAM, truly?

Factual Questions
1

About two and a half years ago I bought some RAM that turned out to be bad. No biggie, just sent it back to Amazon and got a replacement.

14 months ago or so, I did a massive upgrade on my machine - new mobo, CPU, and RAM. For about a year I experienced crashes on a regular basis - my machine would simply reboot without warning. I had Windows set to not automatically restart on failure, but I never once got a BSOD. Both Memtest86 and MS Memory Diagnostics assured me, on several occasions each, that my RAM was fine.

At first I was only getting crashes several times a week, so I could live with it; but by a few months ago it was happening several times a day, so I finally rolled up my sleeves and attempted to solve the problem. With the help of some kind people here, I finally determined that my two RAM sticks weren’t doing well together, even though they were the same make and model. That is, if I pulled one out and ran with only 2 GB, everything was fine; it didn’t seem to matter which stick I used.

I thought my problem was solved, but after a couple of weeks I began getting crashes again, though they were intermittent - a few times a week at most. Knowing I’d be buying new RAM when Windows 7 was released, I lived with it.

A month ago, in preparation for Win 7, I bought 8 GB of RAM (2 packages of 2 2GB 1333 Corsair sticks), and put two of the sticks in my system (was running Vista 32, planning to install 7 64). Never had another crash. Go figure… the problem was my RAM all along, even though both diagnostic tools repeatedly told me it wasn’t.

This past Friday, I got Windows 7, put a new hard drive and the remaining four GB of RAM in my machine, and tried to install. I kept getting errors. I read somewhere on the 'Net to try installing with only 2 GB of RAM, and this worked. I then put the rest of the RAM back in my system, fired it up… and within 10 minutes got a BSOD.

“Great OS,” I thought. I rebooted, and in the next hour it crashed on me four more times. Of course, I knew in my gut what the problem was. I’d been keeping track, so I knew that the second 4 GB I’d put in where both from the same package; I yoinked them, and my system has run fine since.

So, to sum up:

Number of times I’ve bought RAM in the past three years: 3.
Number of times I’ve bought bad RAM in the past three years: 3.

F’in unbelievable.

Is it really this common for RAM to be bad, or have I just been incredibly unlucky?

2

I’ve had 2-3 sticks of ram go bad on the 20-30 systems I’ve had in the past 18 years. so 3-5% chance using that data.

3

I find that very disconcerting. Can some of the pro computer techs on board chime in as to what kinds of memory problems Memtest86 would fail to report?

4

IAAPCT, none that I know of, there are some combinations of boards and RAM that cause problems. We just had an engineering workstation, cant remember the board off the top of my head but 2GB crucial DDR3 sticks will not boot in it, period. Any other brand was fine.

5

Yeah, a couple of years ago, I had a RAM incompatibility that took four months to steep, and then my computer wouldn’t boot at all. It was speed or CAS or latency or some such issue between the original 1GB sticks that came with the PC and the 2GB I upgraded to. I replaced the 1GB with 1GB that matched the upgrade in every way and have been golden since. I guess I’m wondering if Memtest86 would have told me about the mismatch in any meaningful way that would have let me know a problem was brewing.

6

There’s the case of totally bad ram and then there’s RAM that don’t work with your motherboard.

I’ve had a couple bad sticks over the last twenty years. I can’t give you a percentage. I’ve had more that were funky with my motherboard alone.

7

Been wondering, have you tried just the other 4GB of RAM in the computer? (removing the 4GB you have that works). Maybe the problem isn’t the RAM, but the slots on the MB? I’ve seen bad RAM slots on motherboards before.

8

Bad DIMMs are very often our number one field failure mode. I don’t know the fail rate during board assembly and system test (and couldn’t say even if I looked it up) but I suspect it is non-trivial. There are always going to be defects that only show up when you put a chip on a board, and other defects sensitive to specific configurations, so reasonably high failure rates are to be expected.

I looked up the algorithm used by memtest86, and it looks like a good compromise between speed, coverage, and limitations inherent in testing a memory chip in a system.

When memories are tested after fabrication, the quality of the test will depend strongly on the expected market. Memory tests are anywhere from 5n to 17n and above, where n is the number of bits in the memory. You’re going to be on the low end of that when price is driving production. They will also be a bit more careful about testing parts meant for large customers who track fails and yell at them.

There is also an increasing problem with single event upsets - one bit flipping due to cosmic rays and the like. Highly reliable systems have various mechanisms to correct single bit errors and detect double bit ones. I’m including this for completeness, because it is unlikely to happen all that often for small amounts of memory, but it will happen more often in Denver than in Death Valley.

There is also a lot of gray market RAM out there. It is even possible that some of this never got tested.

9

No, that didn’t occur to me. It’d probably be good to try that before sending the RAM back. :slight_smile:

10

Does your motherboard provide a list of approved memory vendors and part numbers? Sometimes a board is only guaranteed to work with specific memory.