I decided to find the cause of the machine check reported in:

http://mail-index.netbsd.org/port-alpha/2004/04/05/0000.html

I made a modification to sys/arch/alpha/alpha/interrupt.c 
causing the machine check to become non-fatal, SIGBUSing curproc.
gdb revealed the problem lies in the int10 emulation.

Following the guide at http://www.caseydoodle.com/index1.html
and adding Option "NoInt10" to Section "Screen" I got an
ATI Rage IIc PCI running at 1280x1024 16bpp. :)

By the way, what condition is it that mcheck code 0x211 signals?

-Tobias