5 ms·
> Booted in VirtualC64 Perhaps and possibly an unfair comparison given that you're already running in an emulated C64?
by teh_klev 7y ago
> Booted in VirtualC64
Perhaps and possibly an unfair comparison given that you're already running in an emulated C64?
- giovannibajo1 7y agoIt would be worrying if a PC can’t emulate a C64 much faster than the original system. And if the GP didn’t remove frame locking, the emulator ran at the same speed of a real C64
- anyfoo 7y agoNo, my computer, as most today, has plenty of power to emulate the C64 with cycle-accurate speed (at least close to), so it should be exactly the same performance as on a real C64. I even switched to unthrottled at times, which with my setup gives about what a ~9MHz C64 would be like. The real one is about 1MHz. Even that was very slow.
- teh_klev 7y agoFair enough, though I'd like to grab an actual C64 to see what the experience is really like with the C64's own hardware. But like all things here it's a hack, everyone's mileage may vary :)
- anyfoo 7y agoDo you mean because of subjective feeling of it independent of speed? Because as for performance, we are talking about emulating a fixed-frequency machine thousands of times slower. There was no "variable performance", CPU and video timings are pretty much exactly defined. This CP/M emulator was exactly as fast or slow on any stock C64[1], and, correctness assumed, any emulator worth its salt on any non-ancient PC (20 years?) will show no perceptible difference. The "thousands of times slower" is not entirely true, since a C64 emulator has to emulate much more than just its CPU, and the emulator task gets uncontrollably preempted. But there is no way that CP/M will feel any different on the real machine, in terms of performance. [1] The only significant exception was that NTSC machines were a tiny bit faster than PAL machines: A tad over 1MHz vs. a tad below 1MHz, to fit the respective video standard's timings.