If my memory serves me correctly, the PS3 for comparison had a custom Cell CPU which was used for supercomputer clusters. https://en.wikipedia.org/wiki/PlayStation_3_cluster So they must have had some kind of edge compared to normal PCs.
It didn't work too well from my understanding, difficult to develop anything that could make use of those streaming cores that was worth it over a traditional multicore chip with more cores.
Useful for scientific community though.
If anyone knows what the spe cores were used for in games, I'd be very interested to know.
The SPU cores were basically used heavily for any background data processing that might align well with what GPU compute is used for today.
Naughty Dog - Uncharted - Pages 33 - 39, 49, 54 - https://github.com/hww/game_programmer_reading/raw/master/nd...
Gorilla Games - KillZone 2 SPU Case Study - All of it - https://ubm-twvideo01.s3.amazonaws.com/o1/vault/gdc09/slides...
Insomniac Games - Insomniac’s SPU Best Practices Hard-won lessons we’re applying to a 3rd generation PS3 title Slides 46 - 48 https://www.slideserve.com/aldona/insomniac-s-spu-best-pract...
They did, but the Cell SPUs were hard to program and that edge quickly eroded as the Cell proved to be a dead end and x86 CPUs continued to evolve.
That's not quite true. Every architecture iteration brought an increase in IPC, while every process improvement allowed for higher clocks and lower power dissipation. It's true that GPUs evolved a lot more - but that's mostly by getting bigger, wider, and having wider, faster pipes to its own RAM and the shared PCIe bus, as well as bigger heatsinks and fans.