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

2 years behind Intel, as they release 20A with backside power delivery in 2024.


Intel 20A is scheduled for mass production in 2H24 and TSMC N2 in 2H25, so 1 year according to current roadmaps.

However, fabs get judged on what they ship, not what the promise. TSMC is shipping N3 in volume, and Intel 4 is still not shipping, and sounds like it will be a cut down limited volume release (https://www.tomshardware.com/news/intel-meteor-lake-cpus-for...) most likely because the yields are not there.

Based on prior delivery it is highly unlikely that Intel will ship meaningful volume on 20A before TSMC does on N2, but miracles happen.


>Intel 20A is scheduled for mass production in 2H24 and TSMC N2 in 2H25, so 1 year according to current roadmaps.

TSMC N2P ( Backside Power Delivery ) is 2026. So that is what OP meant by 2 years.


you are right. but the claim that tsmc is "2 year behind intel" based on those roadmaps is assuming a lot.

intel has consistently failed to deliver on their prior roadmaps and is currently behind tsmc in what they are shipping, so it requires a bit of magical thinking to believe that they will suddenly get a 2 year lead on tsmc.

intel is still not shipping 4, which is their first part to use euv. based on what they are shipping, they cant do it at volume yet. the idea that they are going to scale euv and then jump to gaafet and bspd in the next 18 months is kind of beyond belief.

when intel was failing to deliver 10nm (now called 7) in volume, they were saying the exact same thing, that 10nm was delayed but 7nm (now called 4) was on schedule (https://www.anandtech.com/show/15217/intels-manufacturing-ro...). back then intel 4 was scheduled for 2021. the magic leap didnt happen and it is 2 years late so really no reason to think that everything else wont get pushed back the same this time.

tsmc has already been shipping euv products in volume for years and they have a roadmap for introducing gaafet and bspd incrementally which looks a lot more realistic.


We'll see if that -

1) Is actually released in 2024 2) In volume 3) Plays a meaningful role in getting Intel back into striking distance to the performance crown in perf per mm per watt. Intel has pulled every shortcut lever (larger dies, crank up the power) to continue to be relevant. They're running out of shortcuts.


Intel's scheduled release dates tend to be... optimistic.


TSMC was never about being first, despite somehow they are now the leading edge Fab.

TSMC has always been conservative with feature set. GAA was supposed to be 3nm but ultimately pushed to 2nm. Same with Backside delivery.

TSMC is very clear about what their date does and means. Intel's 2024 ( and it is late 2024 ) will only be starting to producing and shipping to customer in 2025.

The volume difference between Intel's 20A and TSMC is also not comparable. This is a similar case with Samsung where they announce their latest nm but you dont see actual product shipping with it.

That being said I still think Intel would catch up by 2025. The question is whether they could find enough customers and volume.


What makes you think they'll catch up so soon?




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

Search: