On Wed, 10 Jun 1998, Mike Montour wrote:
> A warning for anyone running Linux 2.0.xx on an AMD K6 processor:
> It appears that there is a new bug in the K6, that can cause instability
> and hard lockups with Linux kernels such as 2.0.34 (the -pre shipped
> with RedHat 5.1, and the released version). My K6-233 (serial number C
> 9815DPDW) system has been extremely unstable since I upgraded to RH5.1
> (less stable than a typical Windoze box), and it looks like this is the
> reason. Has anyone else had similar experiences?
>
> >From the linux-kernel mailing list archive:
> http://www.tux.org/hypermail/linux-kernel/1998week23/0093.html
Mike, are you sure this is the cause for your extreme instability?
a bug that has existed in the K6 since its inception, at least if
I understand it all correctly. It was only uncovered by extensive
testing with the "crashme" program, and then the bug was isolated
by checking to see which illegal instructions were being executed
and crashing the kernel. My understanding is that this crashing
is related to a CPU erratum that AMD had already discovered and which
will be fixed in future steppings of the K6. See the message concerning
the erratum published by AMD:
http://www.tux.org/hypermail/linux-kernel/1998week23/0324.html
and the reply which claims (correctly???) that this erratum is not
a problem because Linux doesn't use SMM.
Also look at:
http://www.tux.org/hypermail/linux-kernel/1998week23/0214.html
This says that kernels >= 2.1.43 have different kernel memory
protecion, and this works around the bug for Linux. And:
http://www.tux.org/hypermail/linux-kernel/1998week23/0129.html
which documents that 2.0.xx have the problem, but not >= 2.1.43.
Hopefully the same fix will be applied to the next 2.0.xx
kernel release(?).
BTW, which type of K6 are you running, stepping number, etc.?
I just bought a new K6-266 a few weeks ago (before the big price
drop -- D'oh!) and haven't had any problems with it. It may be
of a later stepping, although I'm running 2.0.30 from Debian 1.3.1.
This chip overclocks nicely to 300MHz at a 75MHz bus x 4.0 multiplier
setting, and runs stably at 2.1V, below the 2.2V rating. I can't get it
to run any higher (83.3MHz x 4, 75MHz x 4.5) without cranking the CPU
voltage up to 2.5V, so I'm content at 300MHz :^) I'm using an Asus
P/I-P55T2P4 Rev. 3.1 motherboard -- the overclocking champ, but still an
83.3MHz board, not one of the new Super Socket 7 boards that goes to
>100MHz.
HTH.
Clemmitt Sigler
Va. Tech Physics Dept.