Where’s the 0xbeef?
TSMC have delivered consistently. Intel have failed to year after year. Why should we believe the Intel roadmap now?
TSMC said it won't start production at its 2nm node until the second half of 2025 or possibly the end of that year, which could signal a shift in the competitive landscape. The Taiwanese chip foundry revealed the timeline for its 2nm node, known officially as N2, during a conference call [PDF] last week for its first-quarter …
I don't wish Intel any ill will...But as someone that grew up near a 1 mile thoroughbred race track, it's hard to see how a formerly champion horse that was renamed to "14++...++" could suddenly reemerge as a winner.
Even the supply chain dynamics would seem to make that challenging. No matter how many $300M+ checks one writes to ASML, high NA EUV machines aren't showing up the next day even if one is a Prime member.
There is a lot going on here. It sounds just like numbers going down, but the passage to 20A and beyond also involves a new transistor architecture, power vias, new metals for via filling, ... Reliability will be somewhat of an unknown until they hit the market, plus the main questions right now seem to be related to architecture, not scaling (something which Intel is seemingly also well aware of, see Loihi and multi-ISA).
TSMC seems to be playing it not just safe but smart. They have a significant advantage in process capabilities (see chip-on-wafer and wafer-on-wafer for example). I mean, also look at what they are able to do with Cerebras.
Intel might not have the luxury to be too strategic, we'll see if they are stretching out too thin (pun not intended) or whether they really will have some ground to stand on. The answer might also depend on how their business will evolve in these 3 years (foundries, GPUs, how much market they will lose to competitors).
Last but not least, I am not even sure the comparison between Intel and TSMC is actually meaningful. It is not entirely clear whether the <20A nodes will be available to Intel's foundry customers. At a minimum, they might be exclusive to Intel products for x years before they are added to the foundry offering. In this case TSMC is again, playing it smart.
Marketing tripe might sound good to other marketers, but it really pisses off most technical people.
Either name your process for the size of an actual, physical feature, or stop pretending it's a length and use a codename instead, like for the products.
Biting the hand that feeds IT © 1998–2022