Samsung Foundry renames 3nm process to 2nm to better compete with Intel 20A

DragonSlayer101

Posts: 372   +2
Staff
What just happened? Amid increasing competition with Intel, Samsung Foundry has reportedly decided to rebrand its second-generation 3nm-class fabrication technology, previously known as SF3. According to the report, the upcoming node has been renamed as SF2 to denote the 2nm-class manufacturing process. Samsung Foundry, however, has yet to officially confirm the renaming.

The news comes from the Korean version of ZDNet (via Tom's Hardware) which claims that one of its sources was informed by Samsung Electronics of the change earlier this year. According to the source, Samsung Electronics had signed a contract with Samsung Foundry for the 2nd-generation 3nm production last year, but the contract was subsequently revised to reflect the change.

The renaming is said to be part of Samsung's plans to simplify its process nomenclature, but it is also likely to help the South Korean tech giant better compete with Intel Foundry, whose 2nm production node, dubbed Intel 20A, is expected to arrive later this year.

Samsung announced its process technology roadmap back in 2022, listing a slew of nodes to be rolled out through 2027. According to the roadmap, it will target mass production of 2nm process technology (SF2P) by 2025 and 1.4nm (SF1.4) by 2027. The company also said that it was planning to expand its production capacity for the advanced nodes by more than 3x by 2027.

However, if the report is to be believed, things changed earlier this year, when Samsung notified its customers about the changes to its roadmap and the renaming of its SF3 process to SF2. As of now, there's no word on whether there are any real design changes with the new node despite the rebranding, meaning customers will very likely get the exact same technology as the erstwhile SF3.

Samsung is expected to roll out its new SF2 process node in the latter half of this year. From what we know, SF2 uses gate-all-around (GAA) transistors, or Multi-Bridge-Channel Field Effect Transistors (MBCFET), but does not feature a backside power delivery network (BSPDN) unlike Intel 20A. BSPDN is said to improve transistor performance, lower power consumption, increase logic transistor density, and eliminate interference between the data and power wires, potentially giving Intel 20A a significant advantage.

Permalink to story.

 
Silly Samsung!

Everyone knows 3 is bigger than 2, therefore it is better. If not, then ask the 1/3 pounder why it lost to the 1/4 pounder.
 
This whole entire mess is Intel's fault. They cannot compete with TSMC's process directly so they decided to beat them with naming schemes. Intel 4 for example is supposed to mean 4nm but it's actually closer to 7nm.

This is what we can expect from Intel going forward. Rebranding with buzz words like Extreme, Hyper, Ultra, etc. that don't actually mean anything. Deceptive marketing campaigns where they try to convince the masses that less is more.

"Buy our processors, they're 2% faster in certain games compared to AMD and forget the fact it has to draw 2-3x more power to achieve that."

Shady PR tactics, biased reviews, etc. Intel will fight tooth and nail to claw back market share from AMD and process leadership from TSMC.
 
Nanometer is nothing but a marketing term now.
Very true. The node and their and its technical internals began to diverge all the way back at the 90nm node. To show just how far they're currently disconnected from reality, a chip with an average feature size of 7nm should be able to pack 20 *billion* transistors/mm-1. But Intel's 7N node is around 100 million/mm-1, and TSMC not much higher.
This whole entire mess is Intel's fault. They cannot compete with TSMC's process directly so they decided to beat them with naming schemes.
Utter nonsense. TSMC is just as guilty as Intel is here, and in fact historically speaking, TSMC's naming convention first began diverging more radically from actual geometry than did Intel.
 
This whole entire mess is Intel's fault. They cannot compete with TSMC's process directly so they decided to beat them with naming schemes. Intel 4 for example is supposed to mean 4nm but it's actually closer to 7nm.

This is what we can expect from Intel going forward. Rebranding with buzz words like Extreme, Hyper, Ultra, etc. that don't actually mean anything. Deceptive marketing campaigns where they try to convince the masses that less is more.

"Buy our processors, they're 2% faster in certain games compared to AMD and forget the fact it has to draw 2-3x more power to achieve that."

Shady PR tactics, biased reviews, etc. Intel will fight tooth and nail to claw back market share from AMD and process leadership from TSMC.
I heard a different theory. That is, TSMCs 4 or 5 nm is in size close to intel's 7 nm.
But ultimately, if either is faster, then their nm is more accurate.
 
This whole entire mess is Intel's fault. They cannot compete with TSMC's process directly so they decided to beat them with naming schemes. Intel 4 for example is supposed to mean 4nm but it's actually closer to 7nm

Just plain wrong. Actually TSMC is the one who started this and Samsung is not better.

Samsung 8nm is like 10-12nm TSMC.

TSMC 4nm is just optimized 5nm, has nothing to do with 4nm just like TSMC 6nm is just 7nm optimized.
 
Back