> > How come that the 1K-sector is so much slower on output, but about
> > the same on input???
> Maybe this is due to the write cache being smaller? Maybe the cache
> is not optimized for larger blocks?

It's even conceivable that the cache is entirely disabled for any setup
other than what they format the drives at the factory with.  Given that
relatively few (no?) recent disk drives will format to anything other than
512 byte sectors, it's pretty clear that the disk manufacturers aren't
interested in providing any more flexibility than MS-DOS absolutely requires;
and I could easily envision someone at the drive manufacturer figuring that
it was easier to shut off the caching for untested configurations than to
extensively test (and debug!) in highly unusual configurations.