From: Kenneth R. Koehler, PhD
Associate Professor of Physics
Raymond Walters College
University of Cincinnati
(koehlekr@ucrwcu.rwc.uc.edu or kenneth.koehler@uc.edu)
Re: AMD K6-2/400 and FIC VA-503+BM with Linux 2.0.36
First of all, I want to thank all those whose quick and helpful
responses have enabled me to make progress on my problem.
It seems (as many of you suggested) that the problem was mostly thermal.
Using a "Heat Sink Pen" (Radio Shack catalog # 64-4342, on sale
last week for $2.99), I remounted the hink sinks/fans to the cpus
after applying a more or less uniform coat to the cpu (and removing
cat hairs that were caught in the liquid). All cpus save one were
cured; the most dramatic cure was the cpu which I mistakenly tested
before treatment. It failed repeatedly within minutes; after treatment,
the test bed ran for 10 or 12 hours without a fault before I killed it.
Interestingly, the last cpu, which had always seemed to give me more
frequent faults than any other, worked fine when moved to a different
motherboard and equipped with a slightly larger heat sink.
After browsing through AMD's fan recommendations (found at their web
site at http://www1.amd.com/products/cpg/thermals/), I noticed that
only 9 out of some 60 fans tested were recommended for K6-2s at
400 MHz, while the majority were recommended at 380 or 350 MHz and
below. It seems that a sort of "thermal threshold" exists for the
K6-2 at 400 MHz; anybody contemplating running one should not scrimp
on their cpu fan!
Lastly, I wanted to comment on the suggestions that if the test ran
with the external cache turned off, the motherboard was faulty. All
cpus and all motherboards ran fine with the cache turned off. While
I am not qualified to design pipelined microprocessors, I do teach
about them on occasion, and believe that the above "cache test" is
not conclusive. By disabling the external cache, the pipeline timing
is altered dramatically, effecting everything from microcode pathways
to operating temperature (as I found out). It is also worth noting
that the test bed ran fine at 350 MHz with the cache enabled.
Thanks again - once more I find that the linux-kernel listserve users
provide some of the most timely and useful support in the computing
world (this coming from a former OS support analyst with Sperry) !
Ken
-
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/