Intel is losing the race

Intel will split from their fabs and start ordering production from TSMC soon enough.

TSMC is in full steamroller mode, delivering a new node per year. They can do so because the entire industry buys from them, except Intel and Samsung.

A single costumer can't support the moving to leading edge processes, not now when each new node doubles the cost of its predecessor. The same way that AMD couldn't support the move to 7nm at GlobalFoundries, Intel won't be able to support the move to leading edge nodes at their own fabs. We have seen it at 10nm, and from now on it will be worst for every new node.

Volume is the key, it allows you to afford the spectacular investment for new nodes, and volume favors massively TSMC.

But once TSMC becomes the sole source; won't they just go "full greedy juden" and fuck everyone raw?

Of course it is, it doesn't support Israel because it didn't provide fabrication opportunities to them first! This is just another example of WASPs trying to keep the Jew down! It's Walt Disney all over again. It's all across the media, 'Intel hasn't got 7nm yet', WE WEREN'T GIVEN A FAIR CHANCE! We deserved the opportunity to have the same fabrication schedule as AMD. We will not stand for this. We're assembling a crack team of lawyers, and with the help of corporate lobbying we're going to form pressure groups to enact legislation that guarantees equality in lithography through controlled distribution. If AMD can't release 7nm until 2022, it's all fair. A controlled release schedule means that nobody is on equal footing.

You might be joking, but Intel is so vile and scummy they might do something like this, throwing legal tantrums over TSMC having more money and thus better nodes.

We still have Samsung dude, they have a LOT more money than Intel. But yeah, Intel will be FORCED to spin off their fabs before Samsung.

Samsung fabs 99% print out DRAM and NAND, they're for a different purpose.

Wasn't GloFo's 14nm Samsung's IP?

>AMD goes TSMC
>Nvidia is TSMC
>Apple is TSMC
How long until intlel's fabs go bankrupt?

Denser doesn't mean everything is better

Sounds like you're coping, kike.

It honestly wouldn't surprise me user. The rules never apply to jews.

IP doesn't necessarily mean Samsung can make it, it could just be a white paper design / patent that they licensed to GloFo.

This is not really a surprise for those who follow semiconductor market. Only a few companies (which you can count in a single hand) are able to spend billions of dollars in R&D and in fab construction/renewal.

Adding on top of that, if you are in the foundry business, TSMC has virtually the monopoly of the market, accounting with about 90% of the market share alone!
In terms of customers of bleeding edge manufacturing nodes, also, only a handful of companies: Apple, Nvidia, AMD, Qualcomm (which are, by the way, all TSMC customers).

Global Foundries ran Samsung's 14nm LPP after canceling their own internally developed 14XM process around 2013 or so.
In exchange for the process IP that Global Foundries received as part of their cross license agreement they also serve as overflow capacity for Samsung. The 14nm line at Fab8 in NY is 100% the same as the 14nm lines in Samsung's own facilities. Samsung has engineers station in Fab8 who periodically run test wafers to ensure uniformity across all 14nm facilities.

If Samsung suddenly needed more 14nm wafers they would get priority at Fab8 to have them produced.

Cool, thanks user.

They're going to license GloFo's IP and buy their machinery

GloFo's justification for abandoning their leading edge FinFET node was lack of customers and cost to get volume ramp up. They stated that into the 2020s most of their customers would still be on 12nm and above so they weren't really giving up any market share, and didn't believe they'd recoup costs with 7nm sales.

They did mention that they'd be offering specialized variants of processes that were cost effective. Thats FDX, thin box super fancy SOI. Pretty much no one in any sort of performance oriented market is going to adopt this when all the conventional competitors are offering FinFET nodes. All ARM chip vendors are going to be pushing out FinFET SoCs. SOI is still mostly used for RF and other random things.
Global Foundries basically just killed themselves over a short term cost projection.

I bet you GloFlo was bribed by Intel, and before you know it, TSMC will start have "production issues" long enough for Intel to shit out something by the end of next year.

>it's not profitable for us to produce a product which has only one company competing
Sure as hell fucking isn't, I can hear the silver rattling from here. What the fuck do they think is going to happen when TSMC starts producing 10-7nm versions of their 12nm products?

Given intel's history is likely. Its just as likely that the Saudi investors in GloFo decided to stop handing them free cash constantly.
Long term keeping 7nm running would have been a sound move. Eventually commodity tier chip producers would adopt it, and they wouldn't have to move to a totally different design methodology to utilize ETSOI body biasing.
An additional $2-4 billion wouldn't have been that big of an investment to get the process ramped up to volume production for the performance it offered. AMD will be increasing their chip shipments per quarter. If they had brought their process online in a more timely fashion they might have gotten orders from Qualcomm. In a couple years MediaTek may have been a huge 7nm customer for them.

They just pissed money away for short term cost savings.

Thanks for the info

fucking dumb glofo faggots

Real men have fabs

Useless if you don't have the money

>"7nm"
So it's not really 7nm?

12FDX+ > 12FDX > 22FDX > 12LP
And there's plans for 7FDX later

If you think Epyc is going to be widely adopted you are delusional. Vendors look for consistency and support. Not one time deals like AMD has been known for.

It'll take some time before they stop going to Intel. Especially with the performance drop sometimes being 1/2 to 1/3rd of the original after the security patches.

Epyc will not make dent in the market, China's Epyc Clone will undermine Epyc's own marketshare with the chinks selling for even less.

there is no industry standart, so yeah

>somebody outside of CHYNA buying this spying tool
nah

TSMC is literaly cheap chink knock-off sweatshop full of retards. They literally get viruses regularly.

I trust the expertise and experience of GloFo

And yet TSMC has more dosh than both GloFo and Intel combined. This is the truth.

Having more money doesn't necessarily mean good, I mean look at Apple and their Macbook disasters this year.

> amd
> massacre anything Intel has to offer
imagine being this fucking clueless? i cant even begin to.

Tsmc already had them
And still has them

In this case having more money matters a lot, since this is the main reason why GloFo dropped 7nm, not enough dosh from customers.

Oh yeah, because Xeons are totally not defective aborted garbage.

slowpoke.jpg

Gonna take 10 years before AMD do anything with that architecture

Only 22fdx is in production, it'll be who knows how long before 12fdx is ready for volume. GloFo says they're doubling down on it, but they have a history of missing the mark and shouldn't be trusted.

It would be great to have cheaper wafers, literally half of the masks of a new FinFET node, and get performance on par with them. If I had to put money on the line I'd bet on no high performance chip maker adopting it.
Nvidia won't use it. AMD won't use it. IBM won't use it. No ARM vendor will use it.
It'll be used for chips that handle RF and analog signaling primarily.

Why do they call it that then? Isn't that false advertisment or something?

No, it isn't false advertising. Spergs just don't understand anything about lith to know why processes are named as they are.
The tooling used in the process can etch a 7nm line with one exposure, so a process node on the absolute bleeding edge of what these brand new machines are capable of is called a 7nm class node.
Back end metal and poly pitch has literally never been used to describe a process.

Is this image showing the actual circuit distances?
If yes, why is this not shown more often, instead of those opaque 14nm, 10nm, 7nm?

What is being shown is the pitch of one of the back end metal layers and the connected poly pitch. None of this is front end feature size.

>So it's not really 7nm?

afaik the 7nm is the thickness of the insulating layer. This is fundamentally tied to moors law which was originally a statement about the thickness of the insulating layer, the electrical field strength and the resulting power consumption, which is ultimately limiting the transistor density and frequency.

What are you smoking?

I thought the nm referred to the size of a transistor.

In the early 90s some process nodes were named for their targeted gate length, but it didn't last. Occasionally the gate length still aligns somewhat with the process name, but the process isn't named for that.

TSMC's 28nm HKMG bulk process for example produced 33nm wide gates.
Global Foundries 28nm process produced 25nm wide gates.
This was just an artifact of trying to cut their boxes as small as possible and utilizing different methods to achieve it.

Attached: figure-3-stratixV.jpg (500x500, 129K)

>implying one of America's massive semiconductor fabs will go under just so AyyyMDrones can wank off with Chinkshit
(lol

techpowerup.com/247191/intel-micron-qlc-nand-yields-less-than-50-a-prelude-to-global-ssd-price-hikes

Global Foundries has only been staying alive because wealthy Arabs have been giving them cash bailouts. The biggest one has only been giving them cash bailouts because he was a big AMD investor, and GloFo was providing AMD with high performance(relatively) process nodes.

That arab has been selling his shares in AMD, and GloFo now no longer has a process suitable to large die high performance chips. FDX is cannot compete with fins for something like a high end CPU, let alone a GPU.
Global Foundries has literally no future has a high performance foundry if they abandon their 7nm node.
With no dedicated high performance customer, no cash bailouts, they're left only with the automotive market keeping them alive. They're now a commodity fab. Maybe Rockchip will toss them a bone.

IT'S OVER INTLEL IS FINISHED

Attached: 1501591438228.jpg (218x250, 7K)

techpowerup.com/247148/amd-chip-manufacturing-to-lay-solely-with-tsmc-on-after-7-nm-and-why-its-not-a-decision-but-a-necessity

wtf is gf doing?

7nm is too expensive, they're producing more 12nm chips to get more dosh

Thanks for the info.

idk but my gf is nowhere to be found

>Intel and Global Foundries team up
>Intel box will have a Made in USA stamp
Chinks BTFO

yeah right

The cost to bring 7nm FinFET up to volume would have been between $2 and $4 billion usd, thats after they've spent around $4 billion on it already. They didn't like the ROI over the period of the next few years, so they decided to put it on hold indefinitely.

They'll be offering 14/12nm FinFET, 22FDX, and 12FDX in coming years.
The biggest market segment they're looking at is automotive for carputer related things. Modems, solid state sensors, etc. Thats where they're making profits so they're giving up on trying to be high end. Its kind of ridiculous because 7nm would have followed the same trend as every other leading edge process. In a couple more years all the smaller ARM outfits would have leveraged it. They could have seen positive revenue from it over 5 years, but decided it wasn't worth it.

Nah

oy vey

America won't allow certain companies to just die ( they'll have investors give them money at some point... )

There is no saving Global Foundries. They're going to end all high end R&D investments, focus on processes for commodity chips, and they'll never be relevant to the PC market again.

PC market is slowly falling into irrelevancy

AM5 when

Zen 3 and DDR5

Zen5*

Zen13*

Zen88*

Zen 1488

Wrong.
>etch a 7nm line
Maybe in the absolute best theoretical scenario using the most redundant transistor design
>in the early 90s
>some
From the first inception of transistor lithography in the late 60s/70s up until the 350/300/250nm processes, BEOL and FEOL could be made the same sizes.
>he then talks about gate sizes at 28nm-class nodes
What you meant to say was
"They then used the (theoretically obtainable) half-pitch of a gate as the marketing nomenclature."
Meaning, the size of a complete transistor had been twice the node marketing-nanometers, and that's still pushing definitions because they measure only from the inside of the gate and exclude the thickness of the gate itself, sometimes (often) omitting any necessary structures in between the "half pitches"

And finally tl;dr
Since about the 90nm node foundries have been advertising -only- the blank spaces in between any or all physical structures which actually make a gate.
It's misleading in the best cases and outright bullshit in the worst.

>most
least. i meant the most simple transistor possible on the litho process.

fucks sake they even tell you their SRAM figures are from their own best-case test circuits you wouldnt put in an actual CPU cache.

Zen666

What do you think will happen after Intel spins off its fabs? Brainlet.

if they're already shitting the bed with 10nm 7nm will cost TWICE as much to get it running, intel 7nm is either fantasy or 10 years away

Meant for

AMD spinning off their fabs was the best decision, the costs to develop better nodes get ABSURDLY high, future nodes like 5nm or 3nm will probably go beyond $20b

Meanwhile in reality

Attached: lithot1.jpg (300x389, 21K)

>all these AMD shills here on Jow Forums are literally 50 centers wanting American companies like Intel and Global Foundries to die so that their low-quality Chinese factories will have a monopoly

Intel is too poor to compete with Samsung and TSMC user, this is reality. Intel dying is just a bonus.

tsmc.com/english/dedicatedFoundry/technology/5nm.htm

>TSMC’s 5nm Fin Field-Effect Transistor (FinFET) process technology is optimized for both mobile and high performance computing applications. It is scheduled to start risk production in the second half of 2019.

IT'S OVER INTLEL IS FINISHED

Global Foundries has two really big facilities in New York, and they employ a lot of people here, but the majority of the company's management and employees both are Asian or Arabs over seas. Its a stretch to call them American.

The only hope they have is to be bought out by Samsung so the entirety of the Common Platform Alliance braintrust is under one roof. Its a sin for all that IP to be wasted and unused.

So if TSMC gets the lead and GloFo IBM and Samsung become one, what's going to happen with Intel's fabs?

They'll be retooled to produce drones and smart dildos

thank you based intlel

Intel isn't going to do anything with their foundry business except maybe open it up to more 3rd party customers besides the small handful they're servicing now.
They'll keep sinking money into it until they kill their problems with a mountain of cash. Their OEM contracts will keep them a float no matter how badly they perform over the next 3 to 5 years.

What everyone should be looking out for is China's growing foundry industry.

>What everyone should be looking out for is China's growing foundry industry.
Why?

A couple foundries are rapidly reaching parity with the rest of the industry, and the Chinese government is never going to slap them down for stealing IP from Korea or Taiwan.
They also heavily believe SOI has a future and will be willing to work with Global Foundries on that front. Some of the biggest contributors to SOI conferences now are Chinese.

And?

A Chinese foundry could end up being peer competition in the market fabbing chips for AMD or Nvidia in a few years time.