Samuli wrote:
> Yes, a non-idle bp6 machine won't lockup, at least my didn't. When running
> a rc5/other number crunching app I have achieved 30+ day uptimes. Make
> the machine a 95% idle one and I'll give it a week, max. Tested with
> all available BIOSes and the BX-trick.
I've been working on the BP6 instability issue for some months now.
First I developed `burnP6` to test CPU's and cooling. But machines
still locked up, and the MS-Windows crowd reported CRC unzipping errors.
So I developed `burnBX` to test RAM and the controller. Success!
I can generate RAM errors and occasional lockups. Needless to say,
RAM errors are extremely serious, and will certainly cause a crash
or lockup if they occur during a kernel instruction fetch. Since
the errors are distributed across many addresses, and even in a
few different bits, I don't think the SDRAM is actually causing it.
Most likely it is a problem with the 3.3V power supply. Also possible
is a fundamental weakness in the Celeron IO drivers not being designed
for the extra capacitance. Less likely, it is a problem with the BX
chip [they've been overclocked to 133 MHz] or mobo trace routing [too
simple to follow the rules].
An idle system really isn't. Especially with X active, there will be
screen redraws that will approach `burnBX` in severity of RAM loading.
The power supply may not be able to handle the load transient well,
and the voltage sags too much. I'm a little unhappy with my 300W PS,
because I've measured transients down to 3.16V, and statistically it's
likely that I get below 3V. Nonetheless, I'm happy with my BP6 and
it's week+ stable (loaded or not) at 2 * 5.5 * 97 MHz.
I've just released the "stable" :) version of `burnBX`. It is a very
intense RAM tester with max bandwidth asm block-move instructions
and an evil data pattern to rapidly provoke errors. I still use
and recommend Chris Brady's memtest-86 as a general memory tester.
Most importantly, it tests kernel-space RAM, which `burnBX` cannot.
I would invite people with stability problems to run `burnBX`. I'd
be very surprised (and want to hear about) anybody that can run
`burnBX` for 1+ hour, yet still gets lockups.
> Face it, the board has flaws.
The only "flaw" I can see is it uses powersupply 3.3V, instead of
regulating it from 5V as some Asus boards do. But that shouldn't
be a problem if a good powersupply is chosen. I'd use the AMD Athlon
list at http://www1.amd.com/athlon/power .
-- Robert Redelmeier author `cpuburn` http://users.ev1.net/~redelm
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Mar 31 2000 - 21:00:17 EST