Just as another datapoint, I've been soaking a bunch of machines which
contain SGS Thompson devices (a rebadged Cyrix), they all get a similar
problem. Either a sig11 or a kernel panic as above.
The intended use of these systems is for a turnkey control/EPOS system
running a different OS (FlexOS). We were experiencing a number of system
panics and unexplained behaviour. This occured around the time that we
switched to the SGS devices. However these failures wouldn't happen with
any sort of pattern except that memory appeared to be getting corrupted,
so I set up the kit to run Linux as a test.
The soak test simply consisted of a continuous loop of recompiling the
kernel:
while make clean && make zImage && make modules && make modules_install; do
echo .version >/dev/tty8; done
These would usually die after 3 to 5 complete loops. The hardware itself
was known to be good, and the memory system was running at the lowest
possible settings. Replacing the ST device (DX2-66) with an alternative
suppliers (sometime a 40 MHz bus device) would allow the test to run without
problems. The longest I left if for was 27 complete loops.
These tests were tried out on 3 different makes of motherboards. Eventually
the SGS devices would always fail.
As a result of this, we've now stopped using these SGS Thompson devices
(or any Cyrix varient), and are now specifying AMD devices (120 and/or 133).
DF
-- Derek Fawcus (G7FVS) df@eyrie.demon.co.uk