Samsung Still on Track for 3nm Chip Production, but Mass Production Will Start in H1 2022:
A while back, Samsung announced that it would be mass producing 3GAE (3nm Gate-All-Around Early) and 3GAP (3nm Gate-All-Around Plus) nodes, resulting in incredible performance and power-efficiency gains. Unfortunately, not everything groundbreaking will have a positive start, and while the Korean giant aims to commence its mass production plans, those will happen in the first half of 2022.
[...] Instead of pushing forward just to get ahead or match TSMC’s progress, we believe that Samsung has made the right call in delaying its 3nm technology. This will allow the Korean manufacturer to establish a firm base, get over the experimental process earlier, and churn out a higher number of wafers at a faster pace to various clients. Though Samsung claims that its 3nm technology will offer a 35 percent performance jump and 50 percent power savings compared to its 7nm LPP nodes, it is not confirmed how it will fare against TSMC’s own 3nm offerings.
[...] However, Samsung continues to exude that ‘never give up’ attitude as we previously reported that the Korean giant is about to finalize its $17 billion chip plant in Texas. Perhaps that location might also serve various clientele for 3nm orders. It looks like we will find out next year.
(Score: 2) by takyon on Friday October 08 2021, @10:07AM
Samsung Foundry’s New 17nm Node: 17LPV brings FinFET to 28nm [anandtech.com]
17LPV could be the good new cheap node.
Samsung Foundry: 2nm Silicon in 2025 [anandtech.com]
Samsung Foundry to Almost Double Output by 2026 [anandtech.com]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Friday October 08 2021, @10:43AM (2 children)
Are they using engineering units or marketing units?
(Score: 2) by takyon on Friday October 08 2021, @10:46AM
Always marketing units.
The significance of Samsung's "3nm" is that they are switching to gate-all-around transistors [wikipedia.org], while TSMC will stick with FinFETs for one more generation.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Friday October 08 2021, @08:02PM
Seethe more Intcel shill, every business in the industry uses "marketing units."
(Score: 0) by Anonymous Coward on Friday October 08 2021, @11:14AM (1 child)
if only one could integrate the dc-ac conversion into solarpanels ...
alas, it seems semiconductor lithografics has no way to make "inductors", (classically a coil of wire) which i think is a essential element in dc-ac conversion?
oh, wait, maybe that's just for changing voltage (without a resistor)?
well nevermind: more complex silicon so we can remotly buy more stuff for the landfill(tm) project!
(Score: 2) by Username on Saturday October 09 2021, @10:33PM
120v AC is just 60v DC. Most appliances don't use the negative half of the wave. You could run right off the panel by using caps and voltage regulators. The thing is, it will only work when the sun is out....
(Score: 3, Interesting) by bzipitidoo on Friday October 08 2021, @12:29PM (3 children)
It's really amazing that the Moore's Law rate of improvements has been maintained all these decades. I keep thinking the end has to be near, but then they find another direction to take things. When they couldn't push raw gigahertz much any more, they turned to multiple cores.
I don't recall just when die size became the measure (and then the marketing lie) of performance improvement, and now, there can't be much more to wring from that. After 2nm, 1.8 then 1.666 nm? To go from 5 to 3 to 2, people may be expecting 1.5 next, as if it's a geometric progression converging on a physical limit of 1nm. I'm guessing the next area to mine is what Apple did with their M1 chip: more hardware implementations of complex software.
(Score: 2) by takyon on Friday October 08 2021, @12:45PM
Names like 2.5nm and 1.5nm [semiengineering.com] have been proposed. Maybe Intel's 20A, 18A, etc. will catch on.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Username on Saturday October 09 2021, @10:51PM (1 child)
Speed is improved, but how durable will it be? Closer the density the less voltage it can tolerate, more heat it holds, and the more susceptible to interference it becomes. Eventually there will be a line in performance vs longevity. Maybe some wearable electronics that you'll toss out next week will be x10 faster and smaller process than your phone, and your phone x10 faster and smaller than your pc, and your pc x10 faster and smaller than spacecraft electronics.
(Score: 2) by takyon on Saturday October 09 2021, @11:31PM
I've heard some talk of smaller nodes having less longevity.
https://semiengineering.com/aging-problems-at-5nm-and-below/ [semiengineering.com]
If we continued down our current path with no big changes, maybe CPUs would start dying after 10 years of use and that would be undesirable but tolerable. If it's as low as 2-3 years to cover only the average life cycle of smartphones, then people should be concerned and angry about it.
I think the ways around it will involve lowering the power consumption and heat dramatically to accommodate (and as a result of) 3D designs, and maybe using new materials.
It's possible that PC and spacecraft electronics can converge if a new transistor type is very power efficient and radiation resistant, such as those nanoscale vacuum tube transistors. Or the segments can go their separate ways. While it would be nicer to use COTS hardware in smallsats and spacecraft, if the space economy expands, they will get the chips they need.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]