Software emulation of a different processor is always going to be
considerably slower than native code. Even if you had the perfect 1:1
translation (impossible for 680x0 emulating 80x86 at least) for every
possible instruction sequence and situation you still have to do the
translation at some point.
Speed test programs have a hard enough time giving a good estimate of the
speed of a real PC and are much worse with software emulation. Most of
them test a very limited number of instructions and don't get anywhere near
real time use. That is why results vary so much between different tests.
Certain instructions fly in 80x86 emulation like the rep prefix ones that
can be used for memory copying comparing etc and others are slow
comparitively because of different flag manipulations etc. Disk access and
BIOS functions are very fast of course because they are always done
natively (in PC-Task's case anyway). Also emulation (and especially any
pre-translation etc) can give falsely high results by specifically
recognizing speed test code or manipulating the clock, of course PC-Task
does not try to do this.
So the only real way to test a software emulation is with a stop watch from
the time the you start the application to the time you have completed a
normal job with it. Compare the result with a real PC machine and you have
a good idea about the relative speed
If you want a very rough idea of PC-Task's Dynamic CPU emulation top speed
it equates to something like 68060/50=~80486, 68040/25=~80386,
68030/25=~80286.
Here are some benchmarks, many are real programs and hence more accurate,
as you can see PC-Task_Dynamic is clear winner (higher number is better
except for the Windows boot times):-
--------------------------------------------------------------------------
Details | PCX 1.1 | PCX 1.1 | PC-Task 4.2 | PC-Task
| | 1Meg | | Medium
| | Transcription | Interpretive | Dynamic
--------------------------------------------------------------------------
| | | |
Old command line si | | | |
A4000 68060/50/AGA | 12.0 | 13.2 | 12.5 | 68.9
| | | |
LandMark 2.00 AT MHZ | | | |
A4000 68060/50/AGA | 16.5 | 20.92 | 16.0 | 127
| | | |
3dBench FPS superscape | | | |
A4000 68060/50/AGA | 2.0 | 2.0 | 2.5 | 5.4
| | | |
BBench Sieve | | | |
A4000 68060/50/AGA | 12.21 | 12.67 | 11.36 | 86.87
| | | |
BBench Sort | | | |
A4000 68060/50/AGA | 1.048 | 1.269 | .9160 | 9.021
| | | |
BBench IntMath | | | |
A4000 68060/50/AGA | 1.799 | 4.621 | 1.726 | 6.676
| | | |
ByteCPU ByteMark 32DOS | | | |
Numeric Sort | | | |
A4000 68060/50/AGA | 0.01147 | 0.011744 | n/a | 0.143864
| | | |
ByteCPU ByteMark 32DOS | | | |
Bitfield | | | |
A4000 68060/50/AGA | 0.05572 | 0.059821 | n/a | 0.101563
| | | |
ByteCPU ByteMark 32DOS | | | |
String Sort | | | |
A4000 68060/50/AGA | 0.01484 | 0.016317 | n/a | 0.306317
| | | |
Doom TinyWin/LowDet FPS| | | |
-timedemo -demo3 | | | |
A4000 68060/50/AGA | 1.58 | 1.54 | 2.54 | 7.67
| | | |
DoomBench Full-1/Hi FPS| | | |
-timedemo -demo3 | | | |
A4000 68060/50/AGA | 0.446 | 0.439 | 0.714 | 2.09
| | | |
QuakeBench Full FPS | | | |
timedemo demo2 | | | |
A4000 68060/50/AGA | * n/a | * n/a | n/a | 0.4
| | | |
QuakeBench Tiny FPS | | | |
timedemo demo2 | | | |
A4000 68060/50/AGA | * n/a | * n/a | n/a | 0.7
| | | |
Win standard boot | | | |
A4000 68060/50/AGA | 24 | 24 | 19 | 29
| | | |
Win 386 enhanced boot | | | |
A4000 68060/50/AGA | * n/a | * n/a | n/a | 69
A4000 68060/50/AGA | * n/a | * n/a | n/a | 44(huge)
| | | |
| | | |
| | | |
| | | |
* PCx (Full 1.1 and Demo 2.1) does not appear to run software which
requires paging (eg Quake, Descent, Windows in 386 enhanced etc).
Home.
Web page created by C Hames Copyright 1997, All rights reserved.