At 05:59 PM 6/27/01 -0500, Scott Horton wrote:
>Okay - so here's the way I would do it, given resources, time, talent and so
>forth:
>
><talking through my hat>
>
>Start by duplicating the basic VAX processor in a single chip,

         Check

>as discussed
>here.  Build the Memory Management unit to run current-tech PC100 SDRAM, or
>an interface between the SDRAM and MMU.  This may go on the main chip, or a
>second chip.  Clock the whole thing at some multiple of 100Mhz.

         Ok.

>Second - build a MSCP-emulator in chip #2 that interfaces to SCSI.

         Check

>Third (optional) now that we understand MSCP, build another chip that
>interfaces MSCP to EIDE (because the drives are so cheap).

         Ok

>Fourth - 10/100BaseT interface - again, backward compatible to common DEC
>equipment (DEQNA?) (DELQNA?)

         Check

>Fifth, and last - build a chip for QBUS interface.

         Check


>Now take these four (five?) chips, drop them all on a QBus board with as
>many RAM slots as you can fit.
>
>Voi-LA!  A single-board upgrade for most (any?) QBus-based VAXen, with SCSI,
>Network, and high-speed RAM all unfettered by QBus limitations.  QBus is
>still there for all your older peripherals.
>
></talking through my hat>
>
>The sad part is, although I love the VAXen - I don't know enough to know if
>what I just lined out is even remotely practical, possible, or even if it
>makes any sense at all.

Well what you have described is basically the KA640 processor from DEC. 
With some differences. It has a single chip VAX (CVAX) and 4MB of memory on 
board, it has an ethernet interface (SGEC) and a chip that understands SCSI 
(SII) that is actually talking to a DSSI bus. Further is has a Qbus 
interface that talks to the card "fingers."

I expect that your additions (SDRAM and SCSI) to replace on board DRAM and 
DSSI would actually be a project at DEC if the technology was still alive. 
It is quite "practical."

 > Can you *imagine* NetBSD on something like that?

It would be very quick indeed...
--Chuck