2D Video acceleration is about how fast the card can draw lines, complex curves, fill in colors, and move stuff on the screen. The CPU has the job of telling the video card what it wants, then hands over the process to the video chip to accelerate it. Early 2D cards were just frame buffers meaning the CPU did ALL the work and then dumped it to the video card RAM and the card then just dumped that to the screen. The best framebuffer was the one that had the fastest RAM and interface to the CPU. All the early speed benchmarks just showed how fast the card could draw, color, and move things around.
By the later 90's pretty much all video cards accelerated the GUI of windows to the point where it could not get any faster (like scrolling down a spreadsheet so fast it was all a blur). So OS companies started doing more eye candy to bog things down again.
Getting back to games VESA modes came out in the 90's so video chip makers could accelerate specific tasks for games and programmers could just program generic tasks in specific VESA modes (resolution and color depths) and have them done faster. For Quake you have standard VGA modes, accelerated modes (needing a wrapper such as glide to do translations) and VESA high res modes which offered much higher resolutions and color but crappy speed. I played quake at 800x600 using a VESA accelerated video card (a Nvidia Riva 4MB PCI card) and it worked but was super slow. Using the same card in Direct3d mode made quake playable (and with the proper looks).
