08-10-07 - 1

A message of warning : My RAM went bad. In the past RAM errors were relatively rare, but they've become more and more common. In the drive for low cost and high speed, pretty much all DRAM these days is non-parity and non-ECC. The RAM error rate is also roughly proportional to size, so the bigger chips are more likely to have errors. They're also cramming more chips on each stick and everything is running blazing hot these days which increases the chance of error.

The scary thing is that I was getting random crashes and didn't know WTF going on, so my first thought was a hard drive corruption, so I ran chkdsk. Well of course if you run chkdsk when your RAM is bad, you can actually *create* horrible hard disk corruption. I got lucky in that my RAM had bad sectors way up in the high memory addresses and a chkdsk that runs from a DOS boot doesn't ever use that memory. I finally figured it out and got this nice free program Memtest86+ which found the error and isolated to one stick, so I'm currently running safe with half the RAM. In the future running a RAM check will be one of the first things I do when I get super bizarro random crashes, definitely before I do a chkdsk.

No comments:

old rants