[rescue] Amiga videos online for your enjoyment

[Charles Shannon Hendrix]

Sridhar Ayengar wrote:
> der Mouse wrote:
>>> if you wanted to rewrite memtest86 in forth, you could patch the sun
>>> obp...
>> One of my "someday" projects is to build something memtest86ish for
>> SPARCs.  But that's going to require a good deal more knowledge than I
>> currently have about their memory interfaces....
> 
> Are they well-documented?  I could see why some companies would want to 
> keep certain things about their memory architecture hidden.

memtest uses enough varied and different test methods that it would probably
tickle almost any problem a Sun would have.

That's the whole point of memtest: it runs on machines with wildly varying
memory interfaces and wildly varying potential problems, and almost always
finds the problems.

The most common symptom of memory damage is increased capicitor drainage, and
naive memory tests generally hide this by refreshing so fast the problem never
occurs.  (or it was a few years ago anyway, and obviously not static ram and
cache).

You actually want to *NOT* drive memory hard when testing, because that very
act makes it work more reliably, as long as heat isn't an issue.

Another good test for most CPUs and memory is prime95.  The algorithms for
that project so far push the heat up more than most any other application, and
also have the benefit of quickly showing FPU and memory errors because they
can be set to generate known data for comparison.



-- 
shannon           | There is no such thing as security.  Life is either
                  | bold adventure, or it is nothing.
                  |         -- Helen Keller