>One more thing: On the W98 machine I've got Linux running on
>the second hard drive, totally separated from W98. I'm doing
>dual boots, but branching from a diskette in the A drive so as
>to keep everything separate. The point is that Linux sees all
>128Mb of the RAM. So I think it's not a BIOS problem.
>

Linux seeing all 128MB is not the issue; with the TX chioset, only the first 64MB of RAM could be cached in the board's L2 cache, regardless of operating system. You should be able to see all 128MB of RAM from Win9x as well.

The issue is cache coherency; assumptions made about the location referenced in L2 cache when the cache gets 'dirty'; IOW the data held in cache is altered, forcing it to be rewritten to main memory. If the motherboard tried caching data from both the first and second 64MB, you could run into a situation where the data from the high-order 64MB was written back to the low-order 64MB, which could result in all sorts of interesting and fatal errors. This is conjecture; if it isn't a TX chipset, then there's probably nothing to worry about.

One thing - Linux allocates memory 'bottom up', while Windows allocates memory 'top down', so the operating system behaviors are significantly different, too.

>Just more fun stuff to wonder about when I should be setting
>up validations and transactions.
>
>-- Bob