[Info-vax] DLT drives and antique SCSI compatibility

[Terry Kennedy]

On Thursday, July 16, 2020 at 3:37:02 PM UTC-4, Dave Froble wrote:
> Things change, and what's rather good at one time just isn't so 
> important at a later time.
> 
> The 11/780 CPU was I believe 4 boards.  Large boards.  Maybe 16-18 
> inches on each side.  Yu wanted multiple CPUs, you got multiple large 
> cabinets.  (Not counting the 11/782.)  You got the star coupler to join 
> multiple boxes to the same storage.

More like 12 to 16 boards (too tired to look up the exact number). Heck, the FP780 floating point accelerator was more than 4 boards all by itself). And then there is the memory backplate, the Unibus adapter backplate, the CI adapter backplane.

That was all fine in the 780 era, but large parts of DEC upper management felt that the only way to get higher and higher performance boxes was to go to increasingly complex giant designs. That worked for the 8600, except it was late to market and didn't meet its performance targets (the 8650 ran at the original 8600 target speed and was only a 2 board upgrade from the 8600 which shows how close they got to their goal before they were told to ship it "or else").

The 86xx group was also responsible for the 9000, which was also late and didn't meet its performance targets either. Ken O and others kept pouring time and money into it because they were apparently incapable of believing that the VAX chips coming out of their semiconductor fabs were at least as fast.