You are thinking write glitches - this is what double buffering solves. Read glitches happen when two reads collide, and without any delay mechanism you need t

You are thinking write glitches - this is what double buffering solves. Read glitches happen when two reads collide, and without any delay mechanism you need to choose which one to drop on the floor. What you do to solve this is implementing busy wait on external bus (/DTACK on amiga), you can also read ahead video output data in chunks (think filling cache line using SDRAM Burst mode) giving you more memory controller time in between. This is why back in the day different graphic cards delivered different Doom1/2/quake1 fps, while intuitively there should be no difference games being software renderers. here is a sweet little article By Michael Abrash about vga isa bus hardware optimizations http://www.drdobbs.com/ramblings-in-real-time/184410037 http://www.drdobbs.com/ramblings-in-real-time/184410037