Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

So the way I understand it is gpus are really just a lot of really dumb CPUs and they toss out all but the simplest interlocking. So Nvidia gets to skip a lot of legacy complexity and focus on building chips that just have lots of copies of the same thing. The complexity is pushed downstream. Stuff also has to be rewritten for gpus. And this is for people chasing speed so they're more likely to rewrite. These people couldn't do what they wanted without this new tech.

CPUs need to make legacy code up to 40 years old run faster so there is a ton of complexity in the hardware. They are chasing people who want modest speed bumps without large changes.

Kinda like how apple was able to pull off a performant phone/tablet after Microsoft failed a bunch... Because they got people to rewrite apps (or create) for their platform instead of shoehorning windows apps into a different form factor. Much bloat was cut, usability was redone. It's an analogy so don't go silly over the differences.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: