Post #2068681
2026-05-03 10:04 UTC
Replies (3)
-
@Fissile@mastodon.art 2026-05-03 11:44
@Kiki@chaos.social Afaik they have been massively parallel for a while. Because you need to compute shaders for all pixels as fast as possible
-
@sandzwerg@chaos.social 2026-05-03 14:23
@Kiki@chaos.social afaik GPU were always/for a long time more parallel than a CPU. That allowed them to process many graphics in parallel. But they had a much simpler instruction set and were less complex. That's all from memory though, I'd need to look up a source for that. I might have picked it up in university or a c't article or something.
-
@picklish@weirder.earth 2026-05-03 18:15
@Kiki@chaos.social I am not an expert, but my informal understanding of graphics card hardware progression over time: * 1980's: 2d graphics support via cga -> ega -> vga progression; hardware so that your computer could display pictures that weren't a terminal; as more memory was available then it could switch from low color/palettes to 256 colors? * SNES Mode 7 Graphics is sort of an inbetween; hardware support to matrix transform a 2d buffer to make it appear 3d * opengl / 3dfx voodoo cards: framebuffer support (for depth buffered polygon drawings) and texture mapping units to perspective transform 2d textures onto polygons; these early cards didn't support 2d graphics either so you'd also need one of those; this eventually led into fixed-function pipelines, where there's a set of vertex/pixel shader transforms that you can control knobs for and have set variables you can tweak, but aren't programmatic. The Nintendo Wii is in this category as an advanced example.