Im having a variety of strange hardware bugs only with a VIA FIC-503a
motherboard (AMD K6-2/400). (No overclocking at present.)

The first and most obvious is after X starts, and even after X quits,
any large amount of text screen activity will cause bursts of static
to come *if* the sound device is playing. The sound card is a generic
one, operating in MSS mode, the video card is a Generic S3Virge. This
video card also will not operate properly in any horizontal resolution
other than 80 characters, until X has started (it works even after X
has quit). This isnt due to mclk settings or anything else specific to
X that I can find.

Also, this motherboard completely fails to run at the 112MHz setting
in Linux - it fails to even boot(!), although it tested totally stable
with a variety of intensive tasks (prime95, running Quake demos
constantly overnight, etc) in Win95 at 112x4 (overclocked, 2.6v) as
well as 112x3.5 with standard 2.4v. The motherboard even booted at
124MHz FSB in Windows, although the L2 cache on the motherboard isnt
_quite_ fast enough to make it stable.

This is the same unit that hosed a test filesystem with hdparam when I
activated "32bit" mode, by getting the order of the bytes incorrect.
Using DMA on this motherboard results in occasional errors to the
filesystem (a full file check every system) although the DMA mode
worked in Windows, near as I could tell (Windows crashes often enough
and nonspecifically enough that its difficult to place blame,
especially with the lousy error messages it gives). I dont have any
problems with drives as long as I dont use DMA or hdparam.

The text mode problem requires intense activity to be audiable at
normal 80x50 resolutions, however it becomes annoyingly obvious at
higher text resolutions (I prefer 160x60). What sort of IO takes place
in normal text screen usage (I did all my machine level stuff as a kid
on Apple 2s... I can write good 65c02 assembler code, including the
time-dependant routines to use the Apple 2' floppy disk drives (think
software modems are bad? Imagine if the CODEC for the drives magnetic
head was done in software! it was, in the Apple 2) ... but dont have
much knowledge of the extreme low-level IO software in the PC
architecture)

Since this is an extremely common SS7 motherboard, and since DMA
functions worked as well as anything else does in Windows, I sort of
feel the problem lies somewhere within my (misconfiguration of?)
Linux. dmesg gives me the following (among other things):

PCI: 00:38 [1106/0586]: Work around ISA DMA hangs (00)
Activating ISA DMA hang workarounds.

Where does this come from? I definitly have some ISA DMA going on
(sound card), so this is a hint, although it doesnt explain the
non-ISA DMA problems I have had (the drives, the total failure of the
112MHz FSB)


-- 
jkenner @ mindspring . com__
I Support Linux:           _> _  _ |_  _  _     _|
Working Together To       <__(_||_)| )| `(_|(_)(_|
To Build A Better Future.       |                   <s>