DOSBox is a 32-bit application, so it probably wouldn't be too much higher
Speed Test results on Celeron 300 were significantly higher in XP (with core=normal: +30% in Doom, >50% in speedtst). P3 500 had much smaller difference. P200 MMX did not work correctly with DOSBox in Win9x at all (0.3 XT). Seems there are DOSBox 0.74 issues with small cache CPUs in Win9x.
OK, I saw on 1st page there are also real PC results for comparison.
It's additional info for comparision, not part of the 2 testing sections. There is already close P200.
While P4 and later machines are post DOS machines and their performance is not so interesting. Other thing - their hardware was not designed for DOS and may give big difference on same CPUs, - video card or something else affect this, - it's hard to say is the result near max or was bottlenecked.
And I really don't see a sense of running Dosbox on such slow sub-GHz machines
XT games may be played, I suppose. May be useful for booters and CPU sensitive ones. Without data it's still unknown where is practical limit.
which cannot offer playable emulation speed (the quake demo would take eternity)
You may do 1st subsection only. Speedtst and Doom, all 4 measures in sum probably take a couple of hours on your CPU.
When I upgraded from E8600 to i7-2600k I was quite disappointed by Dosbox performance gain: if compared at the same freq. 4GHz I got only +4% speed gain
There is a little of progress. According to the chart, i7 4xxx with same frequency is 40% faster than C2D.
So it would be nice if Dosbox will be rewriten to use SMP/multithread and VTx. I belive it could be paralelized
DOSBox on modern CPUs works fast for any DOS stuff or close to this. The only multithreading wich is needed is Voodoo emulation wich is not officially supported still.