Sunday, October 23, 2005

I was right! Phew!

I'd been having problems with (computer) system stability and finally decided that it was something besides memory, even though memtest86+ showed lots of errors. My suspicions were raised when I put the memory in my wife's computer and memtest86+ thought it was fine there. Unfortunately, my wife's computer also ran the memory with very conservative timings which, in addition to reducing memory performance by at least 1/3rd, may have also allowed the memory to work without error. I'm not willing to run my system crippled in that fashion, so I gambled that the memory actually was okay and ordered a new processor and motherboard.

They arrived on Tuesday and I finally had the time to install them on Friday. The first stick of memory, the newest, showed one error at a single address, but only about once every three passes or so. Running at full speed. This same stick had shown hundreds of errors every pass at a wide range of addresses, when run in my old system.

The second stick tested perfect at full speed over 14 hours (many passes through memtest86+'s test suite). Likewise the third, and oldest, stick.

So my gamble was right! There was something wrong with my old cpu or motherboard. I just wish I knew which it was. If I could find an AMD Athlon XP processor slow enough, and cheap enough, I'd like to give it a try. Unfortunately, cheap enough would have to be pretty dog-gone cheap to justify the effort because the problem could be the northbridge chip on the motherboard, and that requires a new motherboard.

Anyway, the new processor and motherboard are doing fine. I'm compiling Linux right now (writing this on the kids' computer) and am looking forward to a newly stable system.

Oh, and I have to get that memory stick replaced. It has a lifetime warranty and the manufacturer has a reputation for honoring it.

No comments: