if you were on a 200mhz pci powermac and didn't have too much background things (extra extensions especially) loaded, would the emulation performance perhaps be similar to a 120mhz pentium or close?
the guest os would had been windows 95, running simple desktop apps or 'slower' games (I mean - for a quick example from my head now lets say yes to railroad tycoon 2 but no to quake 2) just for notes
I have SoftWindows 3.0 on my PB 1400c/166 running Windows 3.1 (or 3.11 I forget which) in 386 mode. It feels like a 33MHz 386DX to 33MHz 486DX, but I haven't bench tested it as such. I mostly ran Borland Turbo C++, ClarisWorks 1.0 (for Windows) and the IAR H8 compiler. It was... usable to OK. It was faster than the 100MHz 486DX4 I originally had been given at DAC, but it was chronically slow because it only had 4MB of RAM. It's certainly slower than the 66MHz 486DX2 I got them to swap it with.
So, as a rough estimate, an interpretive emulator would run about 100x slower than the host. That's because the host would need to emulate the MMU and a whole host of environmental factors.
A JIT (or DRE) emulator is faster, maybe merely 10x to 20x slower than a host. This is because there's a trade-off between the compilation effort and emulator performance; and the JIT cache, but also because again, mismatches between the host and target architecture add up, particularly with flags. e.g. a PCI PowerMac is big-endian, emulating a little-endian computer.
The Gary Davidian 68K emulator achieved much better than a 100x slowdown, roughly 12x slower, because it didn't really have to emulate the system architecture (only the CPU); Apple allowed some apps to fail ("Fix your 68K app guys!") and critical system components were native (QuickDraw, Memory Manager and a few other things).
So, a 200MHz 604e would be like a 50MHz to 100MHz Pentium (maybe) with a DRE emulator and a 20MHz Pentium (40MHz 486DX) with an interpretive emulator. So,
@obsolete 's estimate of a 75MHz Pentium sounds more credible.