Thursday, September 24th 2020

TSMC Begins Construction of 2 nm Manufacturing Facility

TSMC, the leading semiconductor foundry in the world, has reportedly begun construction of its 2 nm manufacturing facility. According to a DigiTimes report, translated by @chiakokhua on Twitter, besides the construction of 2 nm R&D center, TSMC has also started the construction of the manufacturing facility for that node, so it will be ready in time. Please do note that the node name doesn't represent the size of the transistor, so it will not actually be 2 nm wide. The new facilities will be located near TSMC's headquarters in Hsinchu Science Park, Taiwan. The report also confirms the first details about the node, specifically that it will use Gate-All-Around (GAA) technology. And there is also another interesting piece of information regarding even smaller node, the planning for 1 nm node has begun according to the source.

Besides advanced nodes, TSMC also laid out clear plans to accelerate the push of advanced packaging technology. That includes SoIC, InFO, CoWoS, and WoW. All of these technologies are classified as "3D Fabric" by the company, even though some are 2.5D. These technologies will be mass-produced at "ZhuNan" and "NanKe" facilities starting in the second half of 2021, and are expected to significantly contribute to the company's profits. It is also reported that the competing foundry, Samsung, has a 3D packaging technology of its own called X-cube, however, it is attracting customers a lot slower than TSMC due to the high costs of the new technology.
Sources: @chiakokhua (Twitter), DigiTimes.tw
Add your own comment

17 Comments on TSMC Begins Construction of 2 nm Manufacturing Facility

#1
Prima.Vera
So what exactly this 2nm means anyways? In the article 14nm vs 7nm the size of gates were almost the same, seriously.
Posted on Reply
#2
AleksandarK
Prima.Vera
So what exactly this 2nm means anyways? In the article 14nm vs 7nm the size of gates were almost the same, seriously.
Until TSMC publishes further information we do not know.
Posted on Reply
#3
m4dn355
Prima.Vera
So what exactly this 2nm means anyways? In the article 14nm vs 7nm the size of gates were almost the same, seriously.
It means higher performance and lower power consumption. If intel achieves the same thing using 14nm and 14 pluses it is great leap both ways. Win - win no matter how low you nickname your gates.
Posted on Reply
#4
gottimw
Prima.Vera
So what exactly this 2nm means anyways? In the article 14nm vs 7nm the size of gates were almost the same, seriously.
Its pure marketing. From pure physics point of view at 5nm distance electrons should spontaneously jump that distance (read teleport), no matter the state of gate.
Its quantum physics territory where stuff works slightly different. Note this is not related to Quantum Computing as that is using different principles and does use standard gates as in conventional cpu.

Effectively it means that its impossible to create cheaper/faster product since, your gates need double checking - if their behavior is desired or quantum interferenced.
Posted on Reply
#5
ratirt
Prima.Vera
So what exactly this 2nm means anyways? In the article 14nm vs 7nm the size of gates were almost the same, seriously.
It is not all about the gates. There are things to consider.
Posted on Reply
#6
hardcore_gamer
Prima.Vera
So what exactly this 2nm means anyways? In the article 14nm vs 7nm the size of gates were almost the same, seriously.
Logic density should continue to scale (Yup, moore's law is not dead)



But don't expect signficantly higher clocks or performance per Hz / gates because Dennard scaling, on the other hand, is almost dead.
Posted on Reply
#7
Sandbo
gottimw
Its pure marketing. From pure physics point of view at 5nm distance electrons should spontaneously jump that distance (read teleport), no matter the state of gate.
Its quantum physics territory where stuff works slightly different. Note this is not related to Quantum Computing as that is using different principles and does use standard gates as in conventional cpu.

Effectively it means that its impossible to create cheaper/faster product since, your gates need double checking - if their behavior is desired or quantum interferenced.
I think it's better called quantum tunneling
Posted on Reply
#8
medi01
7nm => 22nm transistors
2nm => 22-5 = 17nm transistors

Right?
Posted on Reply
#9
Luminescent
Even though they "shrank" the silicon every year or two, cpu's and GPU's consume more and more power, 700 to 1000W power supply will be the normal for a gaming PC in a few years.
Posted on Reply
#10
RandallFlagg
hardcore_gamer
Logic density should continue to scale (Yup, moore's law is not dead)



But don't expect signficantly higher clocks or performance per Hz / gates because Dennard scaling, on the other hand, is almost dead.
Just to illustrate the naming convention fallacy :



Oh and Samsung's 5nm is only 127 MTr/mm2
Posted on Reply
#11
R-T-B
Luminescent
Even though they "shrank" the silicon every year or two, cpu's and GPU's consume more and more power, 700 to 1000W power supply will be the normal for a gaming PC in a few years.
Not really. We used to have dual-gpu cards and some really hot running Fermis. This is nothing new.
Posted on Reply
#12
Caring1
When you think about it, they aren't going to fit many workers in a 2nm manufacturing plant are they, not even the tiniest of ants would fit. :laugh:
Posted on Reply
#13
kiriakost
I will deliver this as simplified explanation of what is happening by shrinking the paint brush which this is used over a silicon canvas.

Silicon this is a raw material as is for example the butter.
To cut down butter in small portions you need something smaller than an axe.

Carl Zeiss the optics manufacturer he has advance in the regard of making fine optics them able to focus laser beam (this does the cutting) and to focus the rays in to a smaller diameter.
It’s time that the optics industry this advances then we gets a new generation of Chips.

Such a knife it can cut silicon and also engrave silicon as required for the creation of electronic components, they are named as transistors.
Imagine transistors as a number of horses them pulling a wagon. ( many horses = faster speed = less strain to its horse)

Additionally the silicon canvas this is named as wafer.

In conclusion, the size of the electronic parts which are available to us, this is defined by the progress made in the industry of optics design.
Posted on Reply
#14
deu
Luminescent
Even though they "shrank" the silicon every year or two, cpu's and GPU's consume more and more power, 700 to 1000W power supply will be the normal for a gaming PC in a few years.
The powerusage fluxuate from lifespan of an architecture; At end of architecture the manifactures pressure the gpu's to the maximum with fully fledged boards, and then a new optimized comes along and reduce TDP of the board. :) We might have gotten used to 1000-series power efficientcy, but wait for AMDs new GPU, I think we might see AMD pushing in the correct direction (and NVIDIA will follow). But remember; noone remember second place i performance at half the power usage; only king of the hill is victorious (hence why gpu's get clocked beyond efficient clocks :)
Posted on Reply
#15
kiriakost
deu
We might have gotten used to 1000-series power efficientcy, but wait for AMDs new GPU, I think we might see AMD pushing in the correct direction (and NVIDIA will follow).
Those words coming from some one owning an ASUS STRIX 1080Ti OC ? :eek:
You seem itching to upgrade you VGA by something that this is not here yet.
It is wiser of you to contact AMD by email for getting a hint of their upcoming road-map.

NVIDIA released all ready their fastest VGA RTX 3080/3090 which this is not further tweak-able by the users.
Any one reasonable thinking he will never tweak a racing car which this is factory optimized to win all known opponents at September 2020.

AMD this is well aware that at high speed racing cars, the slightest mistake will be drive the car out of the road.
They are the ones which will be forced to add power limit circuit too and follow of what product design tweaks NVIDIA did.
Posted on Reply
#16
deu
kiriakost
Those words coming from some one owning an ASUS STRIX 1080Ti OC ? :eek:
You seem itching to upgrade you VGA by something that this is not here yet.
It is wiser of you to contact AMD by email for getting a hint of their upcoming road-map.

NVIDIA released all ready their fastest VGA RTX 3080/3090 which this is not further tweak-able by the users.
Any one reasonable thinking he will never tweak a racing car which this is factory optimized to win all known opponents at September 2020.

AMD this is well aware that at high speed racing cars, the slightest mistake will be drive the car out of the road.
They are the ones which will be forced to add power limit circuit too and follow of what product design tweaks NVIDIA did.
Im not planning to upgrade anything anytime soon, but I can form a rational opion on a hardwaresite eventhough I am not going to buy the stuff ;). I think you might be picking bits out and miss the context: 'The powerusage fluxuate from lifespan of an architecture; ' 1080Ti is precisely what I am talking about: a max out architecture (hence not efficient); If you look at a 1070 you have 145W during gaming. (which was the first 1000 series card) It is the cycle of architecture. Im sure the 3070 will be more efficient than the 3080 (it would not make sense if it where not) Also I am confident that NVIDIA have more to deliver in the 3000 series (look at the segmentation they have done in 2000 series all by themselves :) Competition is a good thing so when AMD preassure NVIDIA I am happy (eventhough im not buying anything right now (1-2 year might be when)
Posted on Reply
#17
kiriakost
deu
Im not planning to upgrade anything anytime soon, but I can form a rational opion on a hardwaresite eventhough I am not going to buy the stuff ;). I think you might be picking bits out and miss the context: 'The powerusage fluxuate from lifespan of an architecture; ' 1080Ti is precisely what I am talking about: a max out architecture (hence not efficient); If you look at a 1070 you have 145W during gaming. (which was the first 1000 series card) It is the cycle of architecture. Im sure the 3070 will be more efficient than the 3080 (it would not make sense if it where not) Also I am confident that NVIDIA have more to deliver in the 3000 series (look at the segmentation they have done in 2000 series all by themselves :) Competition is a good thing so when AMD preassure NVIDIA I am happy (eventhough im not buying anything right now (1-2 year might be when)
I will give you two valuable hints.
a) I care of the context when this is signed by non-anonymous people, such context has the name of white paper.
b) Power Usage this is of highest importance factor to me, entire Europe trying to educate their consumers to stop spending at power hungry devices.

I did my own research and made the discovery that GTS 1660 Super this is the latest ambassador of performance this bonded with wise use of power in watts.
I also started a topic that no one care to respond, I am thinking that terms as Low power Usage combined with the term Silent PC they are not popular in our days.

www.techpowerup.com/forums/threads/i-am-looking-feedback-for-asus-dual-gtx-1660-super-evo-oc-cooler-performance.268943/
Posted on Reply
Add your own comment