TSMC 7nm can fit 7B transistors in a die under 100mm2 and HVM mass production

>TSMC 7nm can fit 7B transistors in a die under 100mm2 and HVM mass production

How did the Taiwanese do it when no one else can?

Attached: Tsmc.svg.png (1280x1008, 82K)

Other urls found in this thread:

chinatimes.com/newspapers/20140625000062-260202
extremetech.com/gaming/237835-nvidia-gtx-1050-ti-1050-outed-priced-and-built-by-samsung
gizchina.com/2018/08/16/tsmc-invests-19-bln-to-build-3nm-wafer-fab-mass-production-in-2022/
twitter.com/NSFWRedditGif

chink intuition

They paid for it. TSMC gets massive customers, and their biggest customers are prepaying for chips. That means they have boatloads of cash to throw at things.

Its not that Global Foundries had any technical issues with their 7nm process, they just didn't want to pay to bring it up to volume when their only customer would have been AMD who had previously committed to at least dual sourcing 7nm parts.
TSMC just won the race and got the node online first.

>They paid for it.
absolutely this

all that money from apple and qualcomm (and nvidia, etc) had to go somewhere

glofo could, but their arabian overlords wanted more short term money

Wasn't the 7nm process from IBM?

Global Foundries, Samsung, and IBM engineers collaboratively worked on a 7nm and 5nm GAA process. They're the last of the Common Platform Alliance.
TSMC developed their own internally.

Xilinx too. People forget the billion dollar FPGA industry wants more transistors per area and less power consumption more than anyone.

Interesting. I wonder if Apple will one day decide to own their own fabs, they can provide sufficient volume and they like to bring things in house.

oh fuck oath, i never think of those cunts

but then cpus are what i order around for a living, fpgas are black fucking magic to me still

Fujitsu was also developing their own 7nm architecture but bankruptcy happened

TIny dicks, tiny transistors.

TSMC raises their R&D spending by 15% every year. Might have something to do with it

What about Intel? Who had more money to throw at things, TSMC or Intel?
>where were you when the jews got BTFO by chinks

here's why?

why would they? the industry reached a wall and getting in now means massive R&D costs with no tangible benefits in the near future. Unless they want to spearhead graphene

Pure play foundries like TSMC could channel all their money on fabrication technology and give fuck all to architecture

because they have the legitimate chinese genes from 5000 years ago.

Taiwan is superior
They are the Japan of today

No they couldn't, and the Arab owner was sick of investing billions without only future prospects of breaking even.

No

Even with Japan being under slump for the last 30 or so years, they still have (literally) cutting edge companies like disco corp and fujitsu which is one of the few relevant IDMs left in the world

Attached: 1535414059680.jpg (1280x1741, 530K)

>spearhead graphene
hopefully someone does

>Global Foundries, Samsung, and IBM
How can three combined companies get btfo this hard?

Everyone but Samsung team were pajeets

Samsung is not cancelling their 7nm and is on track to go into mass production in Q4, in time to fab new Exynos for S10. Samsung's 7nm is supposed to be denser than TSMC's, but we will see once the chips actually go on sale and someone decaps them and put them under electron microscope.

Think TSMC might? Especially if they increase yearly R&D by 15% like a poster above says.

>Its not that Global Foundries had any technical issues with their 7nm process
Didn't Nvidia have a few cheap gpus on glofo 14nm?

Doesn't qualcomm use samsung as their fab?

No such thing, Nvidia was never a GloFo customer

24/7 Research and development

chinatimes.com/newspapers/20140625000062-260202

Yes, a couple of the low end 1000 cards are 14nm GloFo parts

Don't talk out of your ass, child.

It was samsung
extremetech.com/gaming/237835-nvidia-gtx-1050-ti-1050-outed-priced-and-built-by-samsung

No, stop talking out of your ass

>“We utilize industry-leading suppliers, such as Taiwan Semiconductor Manufacturing Company Limited and Samsung Electronics Co. Ltd, to produce our semiconductor wafers,” a statement in the company’s latest 10K filing reads.

Nvidia has never used GloFo, period

Same line due to the cross licensing, some dies are fabbed in New York, thats GloFo Fab8.

>is a 12 year old kid talking out of his ass

Nvidia had some midrange GPUs fabbed by Samsung, but never at GloFo.

You're still talking out of your ass eventhough you have been DEBUNKED HARD by official SEC 10K filing

Nvidia has never used GloFo, period

Is it just me who sees loss.jpg in the bad dies on the TSMC logo?

> wonder if Apple will one day decide to own their own fabs
That would be fucking dumb, do you think Tim Cook is what? A Elon Musk?
A company having it's own foundry is cool until you don't have the best and latest tech and you nobody will want to have business with you (like Intel).

>hurrr durrr I'm a drooling child
Global Foundries Fab8 is overflow capacity for Samsung, thats why they had total parity across their facilities. Nvidia has parts that are produced by Global Foundries.

Go back to /v/ kid

Hey fag, you sound stupid cuz you ain't know nothing
Japan was so popular because they delivered superb products for less money then their Western counterparts
Now every manufacturer migrated to chink land but guess what, they manufacture their high end products in Taiwan, be it tools or electronics

Nvidia does not use GloFo, Samsung's fab has more than enough capacity for Nvidia

Fuck off you ignorant clueless moron, SEC 10K filing pretty much laughs in your face

Super expensive electric rocks

[citation needed]

but eventually global foundries will have had some people requesting 7nm
so hopefully they didn't actually stop their 7nm adventures and have it all sitting there ready

those faggot fruitfuckers are rich.
But not "I'm gonna own a FAB foundry" rich

Fuck that's terrible journalism. One page of nothing. State your thesis, support your thesis, stop meandering, dumbfuck.

I dunno if this is intentional or not.
But you sound stuupid

Attached: tumblr_inline_n4s5a8bXC31r0j6cl.png (500x281, 115K)

They are more then content of researching their own shit and making someone manufacture for them, not like they are doing a nuclear reactor and ordering something externally will nearly always be cheaper. Apple is not a conglomerate like Samsung.

Real men own fabs

Intel's fabs are like toys compared to TSMC

TSMC is purely a fab plant company, Intel is pretty much diversified and not only limited to processors

>Intel's fabs are like toys compared to TSMC
TSMC dare I say it btfo

Attached: Mark-Bohr-22FFL-2017_006.jpg (2667x1500, 223K)

Attached: Mark-Bohr-22FFL-2017_004.jpg (2667x1500, 269K)

TSMC doesn't give a shit about architectures and processor design

They just literally make stuff and throw shekels for bleeding edge shit due to all the shekels from the companies that fund them

IT'S OVER INTLEL IS FINISHED

TSMC Graphene when?

No I see it too.

They are kinda late, huh? Also, ATOM was a disaster, who the fuck would want business with Intel after that?

Never. we will be using light pulse tech before graphene semiconductors.

well they can also fit a lot more dick into g-strings than any other country's people

I bet we'll stuck at 7nm for years like 28nm

Attached: mYAS1E2S_400x400.jpg (400x400, 25K)

Show some sauce

Samsung beat AMD to 7nm and already has plans for 5nm well ahead of AMD.

Get fucked kid.

5nm literally in 2019, TSMC really doesn't fuck around.

What's after 5nm? Or 3nm? Aren't we at the limit of what Silicon can offer?

AMD doesn't have any fabs.

3nm is presumably the limit, after 2022 things will get really interesting

Quantum effects will start influencing transistors at scales smaller than 5nm

gizchina.com/2018/08/16/tsmc-invests-19-bln-to-build-3nm-wafer-fab-mass-production-in-2022/

Yes, we're reaching the end of Process improvements.

after 5nm, or 3nm, we will have to change fabrication process, and therefore, to an extent, CPU architectures to compensate.

Current CPU, and operating design will be akin to pre and post DOS or Pentium

Completely incompatible. Everything will have to be recreated.

Imagine China stole secret from taiwan and build 10 of those Fab

Stop regurgitating this stupid bullshit you don't understand.
Current leakage happens. Always. On every process node. It doesn't just magically pop up below 7nm or 5nm. The short channel effect has always been a thing.
FinFETs and GAAs exist explicitly to combat it.

Silicon substrates will be around for years to come.