Re: 65020 CPU? Look this...

["Ranando King" <rk@magictouchcorp.com>]

Bryan. Consider something. Theoretically, a 2Ghz pentium processor should be
able to emulate about 250 Apple II machines without slowing any of them down
given enough memory (barring disk access). Run any Apple II emulator on
Linux or Windows with no other applications running and see how many
concurrent II machines you can run on a II ghz machine. If you've got at
least 64meg, then you've got enough RAM. You won't get anywhere near that
250 processes. Why?

Counting cycles doesn't tell you how efficient a processor will be in
performing a certain task when there's no direct correlation between
processor speed and instructions executed per second. If you were to make a
65xx chip having only 1 cycle/instruction, even at 2Ghz, you wouldn't be
able to compete with a 2Ghz x86 processor. The reason is simple. 1
instruction from an x86 chip would take several instructions on a 65xx chip
in most cases.

Don't let that bother you though, because there's still a chance that a
computer based on such a 65xx chip would perform better than a x86 machine.
Even though you can accomplish more in a single instruction on x86
processors, that doesn't imply that a chain of x86 instructions would be the
most efficient way of performing a given task. It could be that there exists
a series of 65xx instructions that produces the same result faster than an
equivalent sequence on x86, but it will depend heavily on the task. Floating
point will definitely be a slow down on 65xx based machines. But, with a
simplified hardware architecture, it is definitely possible that running
software on a 65xx based machine could compete at those speeds given
super-efficient and creative coding.

R.


"Bryan Parkoff" <nospam@nospam.com> wrote in message
news:5mFVd.60450$iC4.13703@newssvr30.news.prodigy.com...
>
> "Eric Smith" <eric@brouhaha.com> wrote in message
> qhacpl5y5u.fsf@ruckus.brouhaha.com">news:qhacpl5y5u.fsf@ruckus.brouhaha.com...
> > "Bryan Parkoff" <nospam@nospam.com> writes:
> >> I would predict that 6502 CPU can run at 2GHz
> >> which it would be much faster than Intel Pentium 2GHz and other
> >> processors
> >> because 6502 is only focused in the cycle performance.  I would claim
> >> that
> >> 6502 CPU is the best design with clever optimization however Intel
always
> >> does not do their good job for CPU design.
> >
> > The Pentium 4 and Pentium M execute on average more than one instruction
> > per clock cycle.  How is your hypothetical 2 GHz 6502 going to even
> > come close to that?
> >
> > Intel's processors don't do nearly as much work per clock cycle as AMD
> > processors, but they run at higher clock rates, so the actual system
> > performance is similar, though AMD has a slight edge for some things and
> > Intel for others.  Although I personally prefer AMD, and it does appear
> > that Intel had some missteps last year, on the whole I think you're very
> > much mistaken in your claim that Intel does not do a good job of CPU
> > design.
> >
> > I've been a 6502 user and fan since 1976 (though I like the 6809 even
> > better), but stretching it to 32 bits and running it faster is NOT going
> > to put it into the same league as a Pentium 4 or Athlon 64 by any
> > stretch of the imagination.
> Eric,
>
>     Please remember that I am against 2 or more instructions per clock
> cycle.  It is because of many mispredict branch and other mispredict
> instructions that can reduce or degrade cycle performance.  I never agree
> with the manufacturer CPU design however I have to stick with 6502 CPU
which
> I support one instruction per one cycle.  The problem is that original
6502
> CPU has to execute one instruction per two to seven cycles, but it can be
> changed to improve one instruction per one cycle.
>     You have to prove your theory that 6502 CPU with 16 Bits and 32 Bits
may
> be much faster with 2GHz than Intel Pentium 2GHz because it always avoid
> mispredict instructions.  Try to test the theory.
>
> Bryan Parkoff
>
>