Cyde Weys wrote:
> David Maynard wrote:
>
>
>>It could be anything, including the CPU. After all, memory can't be read
>>without the CPU being involved.
>
>
> Is it even possible for a CPU to malfunction in such a specific way
> while still letting everything else work? It's not having problems
> whatsoever booting and running memtest86 ... it's just when you try to
> load an OS (which needs memory) that you get the problems.

I don't understand your "not having problems" running memtest description
when you then list the problems memtest reports.

Based on the description below, though, the processor wouldn't be high on
my guess list as processor problems are usually more random, if it runs, or
catastrophic preventing it from running (for very long).


>>But with persistent global errors of that type I'd tend to guess a BIOS
>>setting before hardware failure. Try 'safe' defaults and, in particular,
>>the slowest memory speed settings.
>
>
> It doesn't seem to matter. I get the same errors. I paid attention
> more time and it's more like 00XX0000 the first few seconds of
> memtest86 where XX are random hex, but then it settles down to Err-Bits
> 00ff0000 over and over and over.

That sounds like a bus problem, such as weak driver, bad bus terminator,
dirty/bent contacts, warped memory connector, etc., on those specific bits
or some kind of contamination on the motherboard affecting the traces.
Could even be cracked traces from over-flexing (although I don't know what
would have caused the flexing). Flexing can also break the solder joints
under the BGAs as can overheating, like say the northbridge.

I'd concentrate first on a thorough cleaning of the motherboard, the memory
connectors and the memory module contacts mainly because it's doable and
there's not much you can do about cracked traces or the BGA solder.