Mode 0x13 is not that slow and Mode X is not that fast.
The differences are:
- For 4:3 aspect ratio, mode X pixels are perfect squares (320/240 = 4/3) - 320x200 isn't.
- In Mode X you have to work with planes, each plane has 1 pixel (in sequence): Plano 0: pixels 0, 4, 8, 12..., Plane 1: 1, 5, 9, 13; Plane 2: 2, 6, 10, 14 and plane 3: 3, 7, 11, 15...
So, (320/4)*240 = 19200 bytes, which gives us 3 "pages" per plane in VRAM and 7936 extra bytes.
On old machines Video Memory (physical addresses from 0xA0000 to 0xBFFFF) were VERY slow. On those old machines it pays off if you keep an offscreen buffer in System memory. Since some SuperVGA cards, 0 wait state video memory was available.
Old MS-DOS games avoid spending too much time updating VRAM with techniques like "dirty rectangles", for example.
I really recommend, if you want to play with those old hardwares, you search for "Zen of Graphics Programming Black Book", by Michael Abrash.