Volker Armin Hemmann <volkerarmin <at> googlemail.com> writes:


> > I let memtest86+ run overnight and it found zero memory errors.

> Maybe it really operated in the range of allowed tolerances - and those
> were never crossed with memtest as a very light system load.

Manufacturers and designers of ram have sophiticated and expensive 
systems to test ram. When and where to set the "cut off" for acceptable
and garbage memory, is a fluid call at most companies. Paying the high
price for the best ram does not ensure an outcome, but statistically
it moves up the odds favorably.

> But with an OS booted, the CPU, graphics solution, harddisks all sucking
> power like mad, your mainboard or PSU might not be able to deliver as
> stable currents as the specifications demand. Some memory is more
> tolerant than other.

Along with increases in power draw comes differential heating. Heat is the
enemy of all electronics. Add to this the stealth phenomenon of 
"wait-states" [1] and it is hard to predict ram behavior, let alone what
is acceptable and what "fails". We, as users, merely judge the ram
by the fact that the system keeps running and does not crash as OK.
That's not really how processors, busses and SS memory works, but it
is all we have to judge memory performance. 

Memtest is great, but it's really like using sharpened animal bones as
surgical scalpels. Version 5 [2] is better, but still it is only a very
modest tool to a very complex and transient problem; which is very sensitive
to the quality of the volage/current, as Volker pointed out,
as well as the differential heating effects inside your PC chasis. That's
why the gaming systems today create turbulent air flow inside the chasis.
I use water coolers, sandwiched inside (2) fans to cool my 4GHz amd-8350
CPUS. The coolder the better for electonics to about few degrees above
freezing. But, also humidity plays a roll on cooling electronics. 


hth,
James


[1] http://en.wikipedia.org/wiki/Wait_state

[2]
http://forum.canardpc.com/threads/84663-Memtest86-is-freezing-while-running-test-7


Reply via email to