#Commodore Liquidation
02-May-94 22:51:10
Sb: #142175-#Commodore Liquidation
Fm: Andrew Carol 71350,3646
To: Harry Manolopoulos 75410,322
> Although RISC cpu's are "faster", people are over looking that a
> CISC cpu takes less CODE than a RISC cpu with software doing
> the missing instructions.
You seem to missunderstand the basic differences between CISC and RISC.
A CISC CPU brings in complex instructions and _interprets_ them using
micro-code. Most instructions requires a micro-code routine to be run.
RISC says "Why can't that micro-code be my actual machine code?". This cuts a
_lot_ of overhead out of the CPU.
I spent three years of my recent life _very_ deep in the microcode internals of
several large computers and I belive I have very good grounds to stand on.
Dollar for dollar, executing the same source code, a RISC machine will almost
always walk all over a CISC machine. i.e. a 66Mhz PowerPC601 chip is only $20
more than a 50Mhz 68060 yet is lots faster. And RISC chips are usually much
easier to scale to very high clocks and effective speeds.
You are 100% correct about OS efficiency. But, please don't forget that same
efficient OS could be modifed to run on a RISC machine.
The OS and the CPU are two different issues! Of course an good OS can make a
poor CPU look okay, and a pig OS can saddle a RISC CPU. But all things being
equal, a RISC computer will eat a CISC alive!
RISC _is_ the future for at least the rest of this decade until something
better comes along. (Which it will!) I can only hope that the OS's don't
become pigs.
—– Andrew