Re: The best way to upgrade your IIGS

[GUDATH@EZINFO.VMSMAIL.ETHZ.CH (Henrik 'Ratte' Gudat)]

Haven't been here for quite a while, and saw this discussion about emulation,
performance, etc. Thought I might provide some facts.

It is true that running coprocessors at full load requires a fast(er)
Macintosh. How fast is to be negotiated, but around 150Mhz (and this is *very*
ballpark) you're on the safe side.

Extrapoliting emulation speed is very difficult because:
- more Mhz do not mean a proportional real-world speed gain (on the Mac)
- more Mhz do mean a slightly more than proprotional real-world speed gain on
the emulation layer. That's because fixed-frequency maintenance tasks are
causing less overhead. The effect of this mystic acceleration flattens out at
some point, of course, but it is a very noticeable effect when benchmarking
low-end systems.

Emulation is very similar to the real hardware in terms of split bus and CPU
speeds. Bus speed is typically what you get with coprocessor emulation, and CPU
what you get when benchmarking CPU instructions that do not access
peripherals/other silicon.

I must correct Supertimer when he said that Bernie's speed is usually
overrated, in fact it is mostly underrated. That's because there are
differences in screen refreshing (30fps vs. 60 fps) and mouse handling that
make it look more jerky than it actually is.

Ok, now a few benchmarks. On a 601/66 system there's a emulated speed of about
5 Mhz. On a 603e/200, performance equals 18-20Mhz. On a G3/266, bus
speed is around 20 Mhz and CPU around 45 Mhz (Dhrystone), totaling for a 
performance of 34 Mhz. These figures are really approximate and measured while 
not running a CPU monster like Diversi-Tune or so.

One last personal comment, compatibility is a side effect of what an emulator
is for, and not the main issue. An emulator is supposed to let "home users"
have an enjoyable time playing their games or whatever and, that's what I'm
interested in mostly, enable Apple II-based business to be as productive as
possible. Focusing on theoretical incompatibilities that do not contribute to
usefullness is a non-issue to me. It is also getting difficult to discuss
performance when a AppleWorks macro takes 4 seconds to execute and a huge
optimization would make that 3 seconds.
The above is not implying that there's nothing left to fix in Bernie. Not at
all. We are working on that.:)

- henrik