berger escribió:
The 'real' difference may be minimal, but MAME should always try to emulate the hardware as much as possible, and this Fidelity hardware works this way... I can vouch for that!
Well, but not with these negative side-effects, at least in my opinion.
In the future I guess the driver could be optimized, or maybe the normal evolution of the PC hardware will stop this from being a problem...
In my tests I didn't see any problems in the PC hardware, there must be an other reason.
After Tibono has reported this problem, I've checked it on my notebook (Core i5/3GHz) with the Taskmanager
and one other CPU utility, and I didn't see any difference in the CPU load of my computer - it was around 30%
in both cases, with or without this CPU divider activated.
I only saw the problem when I switched on 'Show FPS' in the 'Options' menu: here the last value is the emulation speed,
and it should be 100% for a module running with the correct speed.
But for these Fidelity modules this value went down (sometimes even to 65%) although the computer CPU was still
showing not more than 30% CPU usage!?
Especially my overclocked modules (e.g. feasbu_8, the 8 MHz Elite A/S) was suffering extremly from this CPU divider,
after the program was out of book and started thinking the module's speech was totally distorted and stuttering.
So it seems that it's not the fault of the computer CPU (with only 30% CPU load), but rather the fault of MAME itself -
maybe this new CPU divider code is not working well in the old MAME version 0.175?
In CB-EmuX (which uses MAME 0.190) this problem is not as noticeable, but of course in CB-EmuX I've not added
such overclocked versions (or any code for chess pieces) for these Fidelity devices.
For me this problem (not 100% emulation speed despite only 30% CPU usage) is still a mystery ...
CBF