On Wednesday, August 14, 2013 12:09:20 PM UTC-5, yoda wrote:
Maybe we should explore a better bus structure for these more powerful processor
The problem is the silicon and physical hardware/connectors for the fancier stuff is such a huge pain.
Right off the top of my head, reliving the 80s/90s, I don't think fiber-HIPPI is a starter. Would be really cool, but...
STD80 is co-incident with the Z80, totally appropriate, and its big brother STD32 is a perfectly good 32 bit bus design. Hardware chips, connectors, and mechanics to be officially STD32 compatible, I donno about.
I believe PC/104 gives you 32 bits and is more or less contemporary with the era we're talking about. Not too much of a mechanical nightmare. Small cards...
SBUS from SUN is (obviously) perfect for 68000 series and I donno about its pinout or mechanical hardware, but if I recall correctly it had 96 pins exactly like the DIN41612 in the ECB project, in fact you could probably re-define the pins on a off the shelf ECB backplane to handle sbus? Maybe? Call it the ECB-SBUS?
Other than that, there's not much from the past that I can remember thats 32 bits? Like I remember VMEbus did something weird like each U stacked provided 16 bits or something so you needed multiple connectors? That would not be fun to fabricate.
Now if you're willing to "cheat" and do stuff wide but slow, maybe in no hurry, you could build a "unlimited" width bus out of serial I2C. I bet it would be scary slow.
If only there was a physical, off chip "standard" for opencores wishbone. I suppose we could make one up. Its about as universal as they get. How bout wishbone on DIN41612 connectors at 5 volt levels.