22 Oct
1998
22 Oct
'98
5:54 a.m.
You guys are really raining on my parade....
Sorry, but most is just physics.
I have looked at Z180's, Z380's and etc....
They are SLOW!!!!!! The
only availabe packaging for the Z380 is un-useable to me as a hobbist
(without a lot of headache.)
Anything you need is a socket adaptor.
So, If some company would take the Z-bus (but with an
(input/output) data
size set of bus control lines), Z-80 instruction set and match a DEC Alpha
performance (at a cheap price of-course), I would worship that company.
(Oh yes, I want a 64bit data bus and a 128bit address bus). Is this asking
for too much? Why can't we all just get along?
Why not just take an Alpha if you want Alpha features ?
And now tell me where you need 128 Bit address ? Just in
case, even to fill a 64 Bit memory you need 4 GIG of
mem thats just 4 grand ... and 128 Bit memory used ...
oh unly 16,000,000,000,000 Dollar ... gee rich man
Since this is a dream:
Just listen to Allison, she (now I know) already
pointed it out: physics.
My thoughts are this: If I can find some REALLY
SIMPLE mirco-controllers
that do just the basic microprocessor functions,
Lets just assume you need only 4 cycles on your
micro-controller to do the equivalent of an Z80
cycle (after all they are software controlled).
this means a 400 MHz micro controller equals
a 100 MHz Z80 with an sustaind rate of around
25 MIPS - thats just double to tripple the rate
of an Z180 - not a big deal - and anything faster
would require Memory with less than 10 nS access
access time - and if you remember SDRAM design,
there are some quirks to resolve - your simple
Bus concept is dead.
And speaking of timing - at 1GHz, a signal
traveles less than 30 cm (or less than one
foot if you like the ancient way). If you take
just 500 MHz and lets say 1.4 ns of switching
time, wireing at hobist level is impossible.
just forget it - these speeds are only possible
inside of chips - gee even chip designers have
problem with the path inside the chip, and you
want rant about using special socets ?
I can parallel them to make them read-in and intrepret
Z-80 code.
Paralell them - interleaved (pipelined) or even
out of order execution ? Have you ever been involved ?
O3 is the crudest thig one could imagine - you need
5 to 15 times the logic of the active components (ALU
etc.) just to coordinate the concurent components.
I cannot see why massively populated microprocessors
(like the PPC, Intel,
and DEC Alpha) can reach clock speeds of 600 Mcyc and a really simple (one
accumulator, bare instruction set) microcontroller can't exceed those speeds!
Because they are ONE pice of silicon ?
If I got some of these micro-controllers and had two
or three of them
reading in instructions ahead of execution (looking for branches) I could
do half of the job and speed up the through-put.
What about taking an extreme simple CPU and try to think
again on a simplified base - I would suggest the 1802 -
one of the simpelest designs - less than 300 gate function
as I remember. No complex math just addition no fancy
OPs just minimal needs. And now try to think about
O3 and parallel concepts at this simple base.
Then use several more u-controllers acting in parallel
to actually do the
instruction execution.
See above.
For math, a bunch of stinking fast memory locations
acting as look-up tables.
A one cycle ALU is still faster - and when using 64 or
128 bit math you need more ROM than available on earth.
Any ideas?
Using hyperspace ?
Gruss
H.
PS.: pleas notice, I got it on topic (1802) <g>
--
Ich denke, also bin ich, also gut
HRK