[plug] strangely crashing PPro [long]

[Brad Campbell]

Leon Blackwell wrote:
> On Thu, Dec 11, 2003 at 11:45:41AM +0800, Richard Meyer wrote:
> 
>>I have read that most memtest programs don't strain the memory as much as
>>doing real work does.
> 
> 
> I can't speak for "most", but memtest86 has detected at least three dud
> RAM SIMMs/DIMMs for me -- two of which passed all other (BIOS, etc)
> tests.  It also has the benefit of running in an OS-free environment, so
> things such as swap and multiple processes don't affect results.
> 
> Next to PSU issues, memory failures have been my second most common
> cause of hardware "misbehaviour".

I second that, also cpuburn is a great app for testing the effectiveness 
of your cooling and cpu/northbrige stability under high thermal loads.

As for Craigs problem, I had a similar one where the server would pass 
all tests with flying colours, then at some point in the middle of the 
night would retrieve a dud block from swap for some reason and do 
anything from a full on panic, to oopsing a process to just locking up a 
process.

Is the machine running UDMA on the hard disk?
Standard ATA has no CRC or error checking on the data being passed 
to/from the disk. UDMA modes do. If it's a dodgy controller/cable/disk 
you might never pick it up in routine testing.


Regards,
Brad