[comp.unix.i386] AMI motherboards design flaws

jr@frog.UUCP (John Richardson) (08/15/89)

In article <329@csense.UUCP>, bote@csense.UUCP (John Boteler) writes:
> I may be hasty, but at least I am honest.
> 
> I just got word from a technician at Equinox, the makers
> of that tricked-out 24-port serial board, who indicated
> some design problems in the AMI 386 motherboards, both
> Mark I and Mark II designs, which leave me with serious
> reservations about using these in a multi-user environment
> such as UNIX. This after exclaiming the wonder of these
> products!
> 
> He confirmed what I already knew: that the cache sticks
> its nose into address space left alone by many other
> m/bs, causing havoc with network cards, smart serial
> boards addressed up high, etc. That's the least of it; it
> also retrieves incorrect data infrequently, but
> occasionally, and without detection under very heavy
> memory processing loads! That scares me. 
> 

   Which addresses do they cache?

  We just decided to use these motherboards and I have not run into `
problems yet. I will be using a 32K WD 8003 Ethernet controller board,
and of course I do not need cache problems. Also did they forget the A31
convention?

				JR