The IBM PC text mode buffer was mapped into memory, so you could quickly update any part of the screen and easily do things like double-buffering (for rock-solid screen updates), composing the screen out of floating windows held in separate memory buffers, etc.
The Unix terminal is derived from line printers (teletype), a fundamentally different kind of TUI.
I don’t know how much this realistically affects today’s terminals since we have extremely fast connections. But it’s the original reason why those DOS text UIs had such a distinct snappy feel.
There are some *nix programs with complex TUI's, where a "snappy feel" is sought after (e.g. the `aptitude` TUI for package management in Debian and its derivatives) and they're just as snappy as any old DOS app. The overhead from using a terminal-based API is just too low to matter.
As it ought to be since the lowliest Raspberry Pi is tens of thousands of times faster than the IBM PC. Filling a screen shouldn’t be an issue regardless of the protocol in the middle.
But design decisions made in the 1980s still linger in a lot of Unix software.
The Unix terminal is derived from line printers (teletype), a fundamentally different kind of TUI.
I don’t know how much this realistically affects today’s terminals since we have extremely fast connections. But it’s the original reason why those DOS text UIs had such a distinct snappy feel.