undefined | Better HN

0 pointsjayw_lead5mo ago0 comments

Yes, both software improvements and tailored hardware will continue to pay dividends (huge gains from TPUs, chips built specifically for inference, etc, even if the underlying process node is unchanged).

Slowing transistor scaling just gives us one less domain through which to depend on for improvements - the others are all still valid, and will probably be something we come to invest more effort into.

0 comments

jopsen5mo ago

More over there are probably lots of avenues we haven't even attempted because when the node scales down we get quadratic benefits (right?).

Where as tailored hardware, software improvements are unlikely to continue yielding such payoff again and again.

So the argument that cost of improvements will go up is not wrong. And maybe improvements will be more linear than exponential.

We also don't know that current semi tech stack is the best. But it's fair to argue that the cost of moving off a local optimum to a completely different technology stack would be wild.

j / k navigate · click thread line to collapse