Supercomputers used to be defined as vector processors. They were SIMD (single instruction, multiple data). Effectively they were vector processors that could multiply arrays of numbers, essentially do matrix algebra. Modern video processors on graphic cards are pretty close to what a Cray from the 1990s used to be. I've lost contact with the field over the last 15 years so I'm sure I'm not up to speed on the subject anymore. Just my 2 cents.