@Piru
Piru wrote:
Yeah. 16 colour hires is painfully slow, too :-D
Well, it's SOMEWHAT faster than 16-color hires, actually. The DMA bus contention is exactly the same as 16-color hires, but for a given graphical operation the CPU is writing only half as much data.
Basically the speed is exactly same as opening a 16-color hires screen but only writing to 2 of the 4 bitplanes.
So in interactive GUI use a 4-color Productivity/DblPAL screen feels about as fast as an 8-color 15Khz Hires screen, or 33% slower than the equivalent 4-color 15Khz Hires screen. (Double Denise DMA traffic but Blitter DMA and PIO traffic stays the same at same color depth).
Edit: This is an oversimplification, I should point out. In OCS/ECS the CPU gets the same amount of bandwidth to chipmem no matter what. CPU writes to chip memory on a 2-bitplane screen will always go twice as fast as a 4-bitplane screen provided there is no blitter activity. The blitter, however, gets extra bandwidth from unused Denise DMA cycles, so things like moving a window and text rendering, as well as other typical blitter-accelerated GUI operations, will go faster when you have less Denise DMA traffic.
Generally though I think for the way most GUI operations "feel", the (ECS) 4-color 35ns DblPAL 640x512 == 8-color 70ns PAL Hires 640x512 is a good comparison, but if you use a patch to replace blitter functions with CPU block moves, then all the 4-color modes will actually feel exactly the same (although theoretically you're wasting the few DMA cycles that the blitter could be using that the CPU can't get at).