Fighting with K7VZA (via KT133A/686B): Any news?

Rob Landley (landley@trommello.org)
Tue, 25 Sep 2001 23:01:15 -0400


Clearing bit 7 of register 0x55 didn't help: it wasn't set on this
motherboard. Compiling the kernel for "pentium classic" didn't help.
Swapping the tulip network card for an ne2k-pci managed to get RedHat 7.1
installed, but it still paniced trying to untar the 2.4.10 kernal tarball.

Background: A friend of mine got a shiny new motherboard, and put a
thunderbird 950 processor and 256 megs of pc133 ram into it. It can run
memtest86 like a champ, overnight even. Just don't try to access the PCI bus
or you're playing russian roulette.

It hangs. It panics. It's not overclocked. I know there are veterans out
there who have fought with this before: has anybody actually found a way to
nail it to the wall?

Sometimes it goes for half an hour (at the command prompt with nobody
touching it, although it's been known to hang there too, but that could be
cron or some such), sometimes it dies in thirty seconds. Trying to recompile
the linux kernel reliably either panics or hangs. Untarring the kernel
tarball has a better than even chance of dying too.

Compiled 2.2.10 on another box, booted that from a floppy. This time the
tarball untarred, but it hung (and, on a second attempt, paniced) trying to
recompile it.

Twiddling every bios setting there is (and upgrading the bios to get more
settings to twiddle) just wastes time. It's not a heat problem, we've
confirmed this not just with the built-in thermometers but by TOUCHING the
heat sinks.

Compiling a kernel for "pentium classic" didn't help. I heard athlon
optimizations offended Via's bureaucracy, but removing them didn't appease it.

I tried the patch that wandered by last week for the 686. Did nothing. It
doesn't even trigger: bit 0x80 of register 0x55 this box's register set
doesn't seem to be on. The "686 Via southbridge" fix -IS- triggering on
every boot (vialatency), but it's not solving the problem.

Stuck in a printk to see if bit 80 of register 0x55 is set in this box's
register set. Doesn't seem to be, the value appears to be 0x0b.

I've searched the lists and found a lot of theories. The last concrete
information I heard was that Alan Cox had a cold, but unfortunately was not
in a position to give it to VIA in person.

Is there some way we can help? (Perhaps arranging for fed-ex to mail used
tissues from Alan's house to the Via developers, perhaps?)

Does the best course of action at this point involve attempting to get money
back, and if so, which motherboard can actually accept ~gigahertz athlon and
PC133 memory and possibly even do something useful with it under any known or
expected 2.4 kernel variant?

Thanks for your time. (How much did Tom's Hardware get paid to recommend
this motherboard as "Rock Solid"? What, Talc? Pumice? Sandstone? Perhaps
a particularly flaky vein of mica?)

Rob
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/