Samsung speed-boosting 3nm GAAFET transistors aim to boost chip performance

Well damn. That's insane we got to 3nm this fast. I figured it would have stalled around 5nm.

It'll be super expensive too. I could see AMD going to them for Epyc to try to get even more market share. That performance increase alone would be quite something paired with 64 cores.
 
Well damn. That's insane we got to 3nm this fast. I figured it would have stalled around 5nm.

It'll be super expensive too. I could see AMD going to them for Epyc to try to get even more market share. That performance increase alone would be quite something paired with 64 cores.

It is worth noting that these node names are mostly marketing.

Area-wise Samsung's 45% reduction from 7nm isn't the most impressive thing in the world, especially when they have 5nm entering volume production in 2020. My guess is that the area reduction won't be that much when compared to 5nm, but ultimately its the performance targets that make 3nm GAAFET impressive.

EUV becoming viable has allowed the industry to blast past some old limitations. The question now is what is the next hurdle/stumbling block, and how much further things can be pushed before we hit it.
 
Yeah the next few generations of EUV nodes will essentially bring the area reductions just by utilising EUV on more layers & transistor types so more types of unit scale down(Rather than scaling all existing units down further), we've seen pretty non-linear scaling across various blocks as we've become more reliant on multi-patterning, it's partly why 14nm is still better at some things than 7nm, or why some types of processor/unit scales down much better than others, though TSMC 7nm+(Or just 7nm for Samsung's first EUV node) should slowly start to change this to bring things more uniform.
 
Back
Top