>>One thing we should do on these macines.  Use SCSI for everything!  And
>>while we are at it, use the Qbus.  I think there is enough documentation
>>out there that one can figure out how it works.  :-)  What's the
>>speed/bandwidth limitations on the Qbus?  

About 3MB/s or so.

>> Might have to change it, make it
>>SuperQ or something.

Well, you could start by demultiplexing the address and data busses. Of
course, this will take more pins, so you might have to trim the address
range down a bit (512KB with a scatter/gather map might be enough, for
instance; you'll only be using it for I/O) to save some pins. Then you
could do a bit of optimization by (say) allowing the next DMA master to
arbitrate for the bus while the current transfers are going on...

> SCSI yes.  Q-bus no.  Get real.  No one is going to produce cards for this
> beast since the market is so small.  You are going to have to use a commodity
> bus such as PCI to get peripherals.  If Compaq were to make a new VAX, you
> can be use it would be PCI based with no ISA.  

Actually, if Compaq were to make a new VAX, it would be based around the
CVAX pin-bus. They have so many ASICs for that bus (SGEC, CQBIC, and on and
on and on) that it would be less costly for them to go that route than to
build a PCI interface.

(I actually had this discussion with Digital at one time. The most they
would offer is a baby-AT formfactor NVAX system with multiple CVAX pin-busses
on it, despite my demonstration that PCI could be done using strictly
off-the-shelf parts)

Roger Ivie
ivie@cc.usu.edu