3 ms·

The whole thing is incredibly frustrating to read because no one is providing any sort of data or graphs that are beyond something you would see in a commercial

The whole thing is incredibly frustrating to read because no one is providing any sort of data or graphs that are beyond something you would see in a commercial. > In terms of the overreaching Power, Performance, and Area (PPA), the paper states that the process delivers a 30% power improvement or 15% performance gain and >1.15x density versus the previous 3nm node. Since I know the nm numbers are meaningless anymore, it’s kind of an insult to the math of 3 to 2 to call these sorts of minor improvements that.

MathMonkeyMan 2y ago
I think that this graph sums it up pretty precisely: <https://semiwiki.com/wp-content/uploads/2025/02/Figure-1.jpg https://semiwiki.com/wp-content/uploads/2025/02/Figure-1.jpg> You see, the bars go up and to the right as indicated by the arrow. The thing at the top right, which is under discussion, is more than fifteen times better than the thing on the bottom left. This is a huge achievement.
- taspeotis 2y ago
  I know you’re joking but that graph is spurious: > We took the graph image, pulled it into Excel and created an Excel graph overlaying it with the 28nm bar normalized to 1 and then entering values for the other bars until they matched the graph. If we then build a set of bars starting at 28nm = 1 scaled up based on the TSMC announced node to node power improvements we get a total improvement of less than 9x. Nodes from N28 to N10 match well but from N7 on the bars on the graph show more improvement per node than TSMC has announced. Just the N3 to N2 bars on the graph show a 55% improvements versus the announced 30% improvement. > … > It isn’t clear what may be driving this difference, but it is a big disconnect. This may be why the graph was removed from the final paper.
  - Ragnarork 2y ago
    Not sure if if this is way overblown in my head but this society, domain, and economy don't take asymptotic progress for an ok answer, and always have the pressure to deliver ever increasing performances and results, even when the law of physics become harder and harder to bend or circumvent. It shouldn't limit innovation and trying to go past those limits one way or another, but at some point it feels like marginal (or below expectations) progress isn't acceptable at least for a while. That it then has to become coated in some marketing lie or half-truths to mask reality and oversell what it truly is... That's still progress! And it can slow down, accelerate, who knows? This should be acceptable.

MathMonkeyMan 2y ago

I think that this graph sums it up pretty precisely: <https://semiwiki.com/wp-content/uploads/2025/02/Figure-1.jpg https://semiwiki.com/wp-content/uploads/2025/02/Figure-1.jpg> You see, the bars go up and to the right as indicated by the arrow. The thing at the top right, which is under discussion, is more than fifteen times better than the thing on the bottom left. This is a huge achievement.

taspeotis 2y ago
I know you’re joking but that graph is spurious: > We took the graph image, pulled it into Excel and created an Excel graph overlaying it with the 28nm bar normalized to 1 and then entering values for the other bars until they matched the graph. If we then build a set of bars starting at 28nm = 1 scaled up based on the TSMC announced node to node power improvements we get a total improvement of less than 9x. Nodes from N28 to N10 match well but from N7 on the bars on the graph show more improvement per node than TSMC has announced. Just the N3 to N2 bars on the graph show a 55% improvements versus the announced 30% improvement. > … > It isn’t clear what may be driving this difference, but it is a big disconnect. This may be why the graph was removed from the final paper.
- Ragnarork 2y ago
  Not sure if if this is way overblown in my head but this society, domain, and economy don't take asymptotic progress for an ok answer, and always have the pressure to deliver ever increasing performances and results, even when the law of physics become harder and harder to bend or circumvent. It shouldn't limit innovation and trying to go past those limits one way or another, but at some point it feels like marginal (or below expectations) progress isn't acceptable at least for a while. That it then has to become coated in some marketing lie or half-truths to mask reality and oversell what it truly is... That's still progress! And it can slow down, accelerate, who knows? This should be acceptable.