🚀 go-pugleaf

>>Now can someone tell me which is better in the next round; the 68040
>>or the 486?
>Sure, that's easy. The 29000 is better

Er, uh, yeah, that's the right answer!  :-)

Seriously, they'll be somewhat close again.  The 486 is a damn good
implementation of a damn bad architecture.  When data and instructions
are in the on-chip cache and when the instruction stream is organized
properly and uses the right subset of instructions, the 486 will have
excellent performance.  Yes, that's right, the simple instructions go
fast on this machine and there is some advantage to ordering the
instructions according to certain rules.  Must be a new discovery of
some sort!  :-)  It will run old code find, just not as fast as code
that adheres to the new rules.  Of course, it still has eight sometimes-
special-purpose registers and other warts.  This just goes to show what
can be done with 1.2 million transistors and a huge die.  Think what a
MIPS, SPARC, 29000, 88000, etc. could do with the same implementation
technology!  (yeah, it might end up something like the i860.)

The 040 will be very similar in spirit.  The simple instructions will go
fast, there will be on-chip caches and floating point.  The die will be
big, and it will have lots of transistors.  The 040 might gain something
from more registers.  It might also gain from separate instruction and
data caches (but this can also be a lose in some pathological situations
because the caches will have to be somewhat small).

So which is better?  The one that costs the least and runs the software
you like.  If you like Unix, then a cheap RISC (29000?) is probably better.
If you like MS-DOS, the 486 is the clear choice.  The 486 is going to make
LOTS of money if Intel can get a decent yeild.  We might even see significant
numbers of new UNIX machines using it.