News Samsung Foundry renames '3nm' process technology to '2nm' production node, following industry trends: Report

[Admin]

Samsung has reportedly changed the name of its process technologies, and resigned contracts with customers. The former '3nm' SF3 node has been renamed to the '2nm' SF2 production node, and we assume further name changes will be forthcoming.

Samsung Foundry renames '3nm' process technology to '2nm' production node, following industry trends: Report : Read more

 

[nogames]

Isn't that cheating? But "following industry trends" are they all cheating with this then?

 

[Sluggotg]

The naming of these modern very small nodes does remind me a bit of Stereo Amplifiers for cars in the 70's and 80's. The power ratings became meaningless. They would claim 200watts but could only put out like 80watts at the most, if your lucky.

 

[Giroro]

Isn't that cheating? But "following industry trends" are they all cheating with this then?

Yes, and yes.
Especially Intel.

 

[kjfatl]

Thanks to Intel, everything is now retarded.

Don't blame Intel. They were the last holdout of real numbers and their stock price plummeted.
Intel recently aligned their process names with those at Samsung and TSMC. Tit for tat, next week TSMC will be naming their 3nm process 10a.
This is all about fooling investors and uninformed consumers.

 

[JarredWaltonGPU]

Don't blame Intel. They were the last holdout of real numbers and their stock price plummeted.
Intel recently aligned their process names with those at Samsung and TSMC. Tit for tat, next week TSMC will be naming their 3nm process 10a.
This is all about fooling investors and uninformed consumers.

Technically, I think Samsung was the one that started the "marketing process names" trend. Its "22nm," "14nm," and "10nm" nodes weren't equal to the competition's similarly named nodes IIRC.

 

[MBOO7]

So wait, let me sum thiis up. You are saying, Intel is the honest one of them all?
🤔
🤭
🤣🤣🤣🤣🤣🤣🤣

 

[hotaru251]

Isn't that cheating? But "following industry trends" are they all cheating with this then?

no as the entire "nm" naming has no real guidelines anymore.

hence why amd & intel's 10 & 7 were differnt nm, but their technically in same class.

you could name it gitgud nm & be fine as the naming hasnt meant anything for liek a decade.

 

[Notton]

If Samsung were some unscrupulous seller on amazon, they would claim 0.002nm, knowing full well no ordinary customer could prove it.
Who has a one million dollar electron microscope they are willing to lend for free?

oh, and even if you do prove it, they'll probably just knife you.

 

[vertuallinsanity]

Just...no! Extra steps.. Same as Intel, Someone is laughably trying to sew seeds of confusion.

Neither company would muddle their nomenclature to create added work for their customer's understanding of process, nodes or fabs.

Re-signing contracts involving additional internal and external legal resources?

hahhHhahhhahAaAAAAaaa! No..

 

[Alvar "Miles" Udell]

Honestly I'm surprised they all haven't changed to the "simple" naming systems like the WiFi Consortium and Qualcomm did (Wifi 6, 7, and Snapdragon 8 Gen 1, 2, 3, 4), or maybe they're all waiting for the Angstrom Era and we'll see Samsung Angstrom Gen 1 or Intel Advanced Process 1st Generation.

 

[ingtar33]

Isn't that cheating? But "following industry trends" are they all cheating with this then?

yes, the original "process" size has long since been replaced by whatever the foundry decides to call it, based on whatever random measurement they want to use. this happened way back when AMD still had foundries, both AMD and Intel started measuring their process size by different values from the original industry standard. this has continued to this day with all foundries doing it.

 

[ingtar33]

Technically, I think Samsung was the one that started the "marketing process names" trend. Its "22nm," "14nm," and "10nm" nodes weren't equal to the competition's similarly named nodes IIRC.

no, the whole industry stopped counting it according to industry standards back when AMD still was making it's own chip, around 32mn i think, both intel and amd were both counting their 32nm process differently from each other, and both were not counting it according to how it used to be counted (the industry standard). (I don't recall how they were being counted, but I seem to remember thinking that AMD's standard made the most sense as chip design had changed since the standard was established, and theirs seemed to honor the concept of it best, where as intel's was "closer" to the original in the details and appearance/letter of the law though it felt a bit arbitrary, but then that might be personal bias and a faulty memory)

I won't blame it on either of them, because i don't know who did it first, I remember back during piledriver someone was saying AMD's 32mn wasn't true 32nm, but then someone else was saying nor was intels, and both sides were pointing fingers at each other. I'm not sure who started it, because both admitted that thier 32nm process wasn't being measured like the original standard, and both insisted their new measurement was the "true" and "valid" way to measure it.

one thing i know for sure is intel isn't measuring it according to how they measured it at 32nm, and since amd isn't making their own chips anymore they aren't either. so both of their standards have changed multiple times since 32nm.

I think i saw an article talking about TSMC's 7nm, and said it basically was 12nm by the "original" measurement, where as intel's 7nm was basically the same as the original 14nm. so they're both majorly skewing the truth. (don't quote this, this is based on a foggy memory of something I read way back when intel was struggling to release chips on 7mn, so several years ago, those numbers could be completely wrong. but that's what that dusty old memory recalls from it)

Now the industry standard is whatever goes.

 

[watzupken]

This sounded like the industry practice no? TSMC's 4nm for example, isn't it 5nm (and we can be sure the transistors are not actually 5nm)? Ultimately, it really depends on the performance of the end product, and less so of what they call it. For example, they can call it 2nm, but performs like competitors' 5nm, they won't be able to charge as much anyway.

 

[vern72]

I'll rename the $0.87 in my bank account to $87 million. I'm rich!

 

[ekio]

Just like a brat that turns their B into a A…

Easy the life at Samsung, just cheating the axioms to get their equation working.

 

[jkflipflop98]

So wait, let me sum thiis up. You are saying, Intel is the honest one of them all?
🤔
🤭
🤣🤣🤣🤣🤣🤣🤣

It's true. Intel was the last company that still matched node names with actual structure size. contact-to-contact measurement. Everyone else in the industry has just named their stuff all willy-nilly for many years.

 

[purposelycryptic]

Honestly I'm surprised they all haven't changed to the "simple" naming systems like the WiFi Consortium and Qualcomm did (Wifi 6, 7, and Snapdragon 8 Gen 1, 2, 3, 4), or maybe they're all waiting for the Angstrom Era and we'll see Samsung Angstrom Gen 1 or Intel Advanced Process 1st Generation.

Isn't Intel tacitly pretending that era has already begun, with it's 20A, 10A, etc nodes? They aren't actually calling them Angstrom, but they sure are implying it heavily.

 

[Alvar "Miles" Udell]

Isn't Intel tacitly pretending that era has already begun, with it's 20A, 10A, etc nodes? They aren't actually calling them Angstrom, but they sure are implying it heavily.

True, but they started that before with Intel 4 (10nm) and Intel 3 (7nm), and they claim that 20A is the start of the Angstrom Era, and is a better naming convention unless Samsung and others will go directly from 2nm to 1nm, which is doubtful since Intel has an 18A planned, and is far better than TSMC's daft naming system.

 

[JarredWaltonGPU]

Isn't Intel tacitly pretending that era has already begun, with it's 20A, 10A, etc nodes? They aren't actually calling them Angstrom, but they sure are implying it heavily.

Umm... the A after the numbers is definitely for "Angstrom" and Intel has repeatedly used the phrase "Angstrom era" when discussing these new nodes:

https://www.intel.com/content/www/u...=angstrom:~:text=As we enter the Angstrom Era

Process and Packaging: Intel 6 Pillars of Technology Innovation

Intel is powering the future with advances in transistors, packaging, and chip design. No other company has our in-house R&D capabilities.

www.intel.com

https://www.intel.com/content/www/u... angstrom-era process technology advancements

Are they actually "Angstrom" features? No, they're not. Just like TSMC "3nm" nodes don't actually have any 3nm features. We basically left all of that behind in the post-180nm node names. Intel's 45nm for example (from WikiChip):

For a time, the process size was the average half-pitch (i.e., half the distance between identical features) of a memory cell at a given size (e.g. 45nm). However, the process node names have diverged from actual features sizes starting in the late 90s.

 

[Pierce2623]

Yes, and yes.
Especially Intel.

Look I’m never one to defend Intel, but they were literally the last company using the real correct terminology on their 10nm node that was denser than tsmc 7nm. Then they changed the name to 7 to line up with the rest of the industry.

 

[Pierce2623]

This sounded like the industry practice no? TSMC's 4nm for example, isn't it 5nm (and we can be sure the transistors are not actually 5nm)? Ultimately, it really depends on the performance of the end product, and less so of what they call it. For example, they can call it 2nm, but performs like competitors' 5nm, they won't be able to charge as much anyway.

Yeah the bottom line is now nodes are just named in a sequential order of performance/density and the names only have a meaning relative to other nodes in the same family. Intel was probably the most honest by removing the “nm” then they went full on BS by adding an “A” to nodes with no features small enough to be measured in angstroms.

 

[sygreenblum]

Yeah the bottom line is now nodes are just named in a sequential order of performance/density and the names only have a meaning relative to other nodes in the same family. Intel was probably the most honest by removing the “nm” then they went full on BS by adding an “A” to nodes with no features small enough to be measured in angstroms.

Yeah, I agree. I'm just waiting for the industry to start using negative numbers. It's about as meaningless as any current naming scheme.

 

[thestryker]

FinFET is what ended any relation between node names and sizes which first happened with Intel's 22nm process. Literally everything using FinFET transistors is just a made up name which is why TSMC stopped calling the nodes "nm" anything upon the shift. Samsung I think had used "nm" during the ~14nm stage, but then they too dropped any pretense. Intel was certainly the last to it, and I think the only reason they changed is because they weren't the leading foundry node anymore. Names don't matter when you're number one, but if you're trying to sell to other people and have real competition there needs to be something at a glance.

They all still talk about measurements when discussing the nodes, and Intel is trying to take back the lead which likely explains the return to using measurement in the name.

 

[purposelycryptic]

Umm... the A after the numbers is definitely for "Angstrom" and Intel has repeatedly used the phrase "Angstrom era" when discussing these new nodes:

https://www.intel.com/content/www/us/en/newsroom/news/research-advancements-extend-moore-law.html#highlight=angstrom:~:text=As we enter the Angstrom Era

Process and Packaging: Intel 6 Pillars of Technology Innovation

Intel is powering the future with advances in transistors, packaging, and chip design. No other company has our in-house R&D capabilities.

www.intel.com

https://www.intel.com/content/www/us/en/newsroom/news/intel-foundry-arm-announce-multigeneration-collaboration-leading-edge-soc-design.html#:~:text=Intel angstrom-era process technology advancements

Are they actually "Angstrom" features? No, they're not. Just like TSMC "3nm" nodes don't actually have any 3nm features. We basically left all of that behind in the post-180nm node names. Intel's 45nm for example (from WikiChip):
View attachment 317
For a time, the process size was the average half-pitch (i.e., half the distance between identical features) of a memory cell at a given size (e.g. 45nm). However, the process node names have diverged from actual features sizes starting in the late 90s.

Fair point, I don't generally read the advertising propaganda pieces, and didn't realize they were arrogant enough to actually full out call them that.

For a good while, though (long after the 90s, at least), the advertised process size of nodes was still somewhat informative as a rough, relative measure between nodes from the same manufacturer. The actual number may not have meant what it should have, but you were able to get something from it.

Now, though... We have things like Samsung just straight up renaming their slightly updated "3nm" process to "2nm", and Intel's whole Angstrom spiel. The values are now well and truly pulled straight out of someone's rear end, and don't have any informative quality whatsoever, aside from "the newer one is likely to perform better than the older one to some extent, we would hope".

Intel could replace "Angstrom" with "Armstrong" in their advertising (Or "Super Ultra Sexy Fast Happy Time Surprise", for that matter), and it wouldn't really change anything.