RAM Test Issue

[KachiWachi]

I have a bad stick of RAM, but the test result doesn't make sense.

I'm testing with Gold Memory v5.07. The stick in question is a 128MB DIMM that shows as 129,984 KB in the program.

When I test in one slot, the failing address is 0222998C (35,821,964 Bytes).

When I test in the other slot, the failing address is 023C9E0C (37,527,052 Bytes).

In either case, it is the 4th bit that is stuck ( 32 bit test, ---x---------------------------- ).

Another known good module tests fine in both slots (no errors).

When I use MemTest 3.1a, the same addresses show failing as well (8 bit test, first bit is stuck - good is FFFFFFFF, fails as EFFFFFFF).

Why would the tests show different failing addresses, yet the same bit stuck??

[edwin]

good question. Maybe the addressing is a bit different between the slots? Different test address pattern used?

[Ritchie]

I think probably a bad stick of RAM, as you suspect. Would be interesting to test that same (assumed) bad stick of RAM in another board, again trying different slots, to see if you get similar results. If not, the chip may just have an incompatibility with your board or possibly the connectors may just be dirty and making bad contact.

[KachiWachi]

I would try that Ritchie, but I don't have another board that accepts DIMMS of that size.

I also upgraded to GoldMemory 6.64...same result, plus a few more intermittant ones.

The stick tested fine when it was first installed, so it must have developed this error over time for some reason. The other 128 MB stick I have in there tests fine in both slots, so it isn't a dirty slot issue.

I'm leaning towards edwin's thought, though I don't know how I could prove that, or why that would be.

[Ritchie]

Probably not a dirty slot issue as you say, but possibly still a dirty RAM stick issue. Pity you cannot test in another system, even if it would run at PC66/100. But I have had RAM sticks that have failed over time, as you suspect.