Re: Wozniak vs Jobs

[dempson@actrix.gen.nz (David Empson)]

Supertimer <supertimer@aol.com> wrote:

> doug.cotton@cmdweb.com (Doug Cotton) wrote:
> 
> >regnirps@aol.com (Regnirps) wrote:
> >
> >> Jobs also killed an Apple II running the ARM (32 bit, 6 MHz at the time)
> that
> >> outperformed the Mac.
> >
> >According to an interview with Bill Mensch, even the 65816 at 4 MHz (which
> >it was capable of at the time it was engineered into the GS) could
> >outperform the original Mac, and 'Apple' purposely kept the clock speed
> >down because of this.
> 
> Mensch is probably right...at 4Mhz, the 65816 is doing 8 MIPS,

Ah, the good old "meaningless indication of processor speed".

Each 65816 instruction takes a minimum of 2 CPU cycles, so a 4 MHz 65816
does a maximum of 2 MIPS, and they aren't very useful instructions (NOP,
CLC, SEC, 8-bit immediate loads, etc.)

The average number of cycles per instruction used in typical programs is
more like 3 to 3.5, so a 4 MHz 65816 does about 1.2 MIPS.

By comparison, the 68000 (not later models - just the 68K) requires a
minimum of two cycles for each memory access, several more per
instruction.  This means that an 8 MHz 68000 and 4 MHz 65816 are neck
and neck for number of memory accesses per second, but the 68000 is
transferring twice as much data in this time (16 bit data bus), and
needs longer to execute each instruction.

In the long run, the 68000 will win for heavy duty calculations, since
it has more and larger registers, while the 65816 has to do more memory
accesses to achieve the same calculations.

The 80286 is similar to the 68000, but the 8086/8088 is even worse (four
CPU cycles per memory access, assuming no wait states).

-- 
David Empson
dempson@actrix.gen.nz
Snail mail: P.O. Box 27-103, Wellington, New Zealand