Intel to outsource entry-level processor, chipset production

keller is only on the mobile shit, he isn't touching core.

it is a legitimate concern because currently amd has the process and architecture advantage, but if intel comes in and pays tsmc 2-3 times what amd is willing to pay, and pay for breach of contract, they can effectively shut amd out of 7nm entirely, or more likely, what ever comes after 7nm.

Thats not how a foundry works, pal. TSMC has something like 90 design wins on their two 7nm lines, the one currently running and their upcoming limited EUV line.
Taking on another client doesn't just bump someone else off. That in fact is not a concern at all.

You don't just take an agnostic design to any foundry and tell them to produce it. Complex ICs are tailored to their process, you have to follow strict design rules in routing, etc. It takes a very long time to port a design from one process to another. Even if intel wanted to they couldn't pay TSMC enough to magically make their Gayname Lake i7 enter production on their 7nm line over night. That takes about a year to port something like that, sometimes longer.

Dell played the offshore game.
> Hey, based Taiwanese company Asus, make this small component for us.
> OK. By the way, can we make your motherboards for you too? Great.
> Can we make the entire PC for you too? Great.
> Can we manage your entire supply chain for you too? Great.
> What's that? Your process engineers and other knowledge workers need to be around live production in order to make improvements? Great, send them all here.

Eventually Asus unleashes its own product and Dell has to fall off the stock market and try to go from being a sticker to a real company again.

When your manufacturing goes, your entire company follows.

>They didn't announce who they've chosen to provide more volume for their chipsets and Atom parts. Maybe its Global Foundries. Maybe its UMC.

It's gonna be the fab where the memory controller for AMD Epyc Rome is being made to try and slow down rollout due to shortages at said fab

>AMD shills actually believe that a billion dollar company will have troubles porting over their design over someone else's node
AHAHAHAHAHAHAHAHAHAHA

If you're an AMD shill you want Intel to stick to their own fabs and have them fumble with 10nm/"""10nm"""

It'll be Core 2 Duo and Core 2 Quad asskicking again if Intel sneakily manages to get 7nm+ or 5nm

>being this ill-informed
Intel's fabs will never die since they still produce other shit for other companies.

What will happen (is happening) is that Intel will move desktop CPUs and low end parts to TSMC while everything Intel does for mobile, SoC and Xeons will still be made by them.

They have no roadmap after 9th gen, just the promise of 10nm on winter 2019 which I doubt they'd truly make it. If anything it will be paper launch again with skyrocketing prices. What is more likely going to happen is that they'll refresh Coffee Lake-R again by making it more cooler.

>muh 6 gorillion dollar company
where's 10nm though?

>When your manufacturing goes, your entire company follows.
explain Apple then

God dammit, I'm an idiot. I should have bought AMD stock when it dipped down to ~16.50/share, instead of thinking it'd go lower so I can get more.

OH SHIT: businesswire.com/news/home/20181030006066/en/GSA-Expands-Vision-New-Leadership-Multiple-Market

>first chair woman of the Global Semiconductor Alliance

GSA Board of Directors: gsaglobal.org/about-us/board-of-directors/

ALL THOSE BIG FUCKING SILICON TECH COMPANY NAMES

Attached: 54465461596415941.jpg (368x245, 7K)

>Also, this implies you might get a lower quality chip that won't OC for shit/heat more.
Intel Atom CPUs never overclocked so that's a non-issue. How much heat they will output and how that affects "boost" is a bigger question. I have a notebook with a N4200. It's got a base-clock of 1100 MHz and a "turbo" of 2500 MHz. That's a pretty huge difference. In reality it's a 1100 MHz chip unless you're briefly using one core.

>once Intel sends their i7s and i9s to TSMC
Seems unlikely. It's interesting to note that Intel are selling CPUs even after the 50% price-hike which have given us a situation where the Intel i5-8400 is at a price-point where it competes with either the Ryzen 2600 or the Ryzen 1700X depending on where you look. From the article,
>Despite the CPU shortages, Intel still performed strongly in the third quarter with revenues rising 19% on year to reach US$19.2 billion and net profits growing 42% on year to arrive at US$6.4 billion.
..kind of makes me wonder if those CPU "shortages" is nothing more than Intel's way of probing the market to see just how high margins they are able to get away with.

Just more evidence CPUs hit the tech dividend wall and are become a mature product/commodity production item. If intel is smart will move across to some new technology development niche. If not they ded. Sad there will be no further CPU speed improvement.

I for one, welcome Intel's new 5nm CPUs

Attached: 1510568522046.png (1827x1147, 2.28M)

No they can't. Apple will always have first dibs over anyone, Intel included.

Intel had stock problems since Kaby Lake, I think they've taken the same playbook RAM makers have which is to manufacture an artificial scarcity to drive up initial prices before stabilizing it months down the line.

delid dis

Attached: 1515221586890.png (682x792, 339K)

>can't make their own 10nm work
>refreshes a refresh since sky kek
you don't say.

Design is still in the US.

>intel
>allowing overclocking on low end chips
good one

Attached: 13978348.jpg (203x211, 53K)

You do realize that what gave Intel its strength are its foundries, right? If they don't trust their own foundries, then neither will external chipmakers

>Look at me! Im fabricator now.

Attached: CHING CHONG CHING CHONG.jpg (1500x1000, 96K)

I am unsurprised that Intel shills are this stupid.

They could not pay them to port it to 7nm fullstop. The Intel architecture is the issue here. Not the shrink method itself. You cannot just shrink it and pray that it works.

So APOOLOO lake will be made by TSMC? Why is this newsworthy?

No, only all newer processors that will be used inside Apple products will made by TSMC.

I started hating Intel back when I bought a 1090t. It took them way too long to create a hexcore chip, let alone eight cores.

My question is, why are you so loyal to Intel? How much do you they pay you?

What is the deal with this constant bashing and praise war between these two rival companies here? What does it matter?

All you had to do was to believe in her

Attached: su_bae.jpg (2880x1920, 899K)

So Appoo? Oh newsworthy.

That is still fairly major. Why would they keep their Xeons on their 10nanomeme which is totally up in the air if they can even get it to work?

>fairly major
>when Xeons make up close to 60% of their revenue while desktop chips are 8% and less
>when they make more shekels from NAND, modems and other integrated chips than desktop CPUs

so pentiums and celerons will be cheaper, where is the issue exactly?

>intel shoves their shit onto TSMC in an attempt to slow down the capacity for AMD's Zen 2 architecture manufacturing which TSMC is the sole fab

pathetic. Intel kys yourselsvs

Attached: Lisa Su.jpg (757x627, 109K)

>her

TSMC is a semiconductor fabrication company, they're not on AyyyMD's team, not on Nvidia's team nor even on Intel's side.

They make shit for everyone who pays up.

ebyn

Yeah but Intel will pretty much suck TSMC cock for the rest of eternity and their fabs are 14nm++++++++++++++++ forever.

Apple "manufactures" software in the US.
If it starts outsourcing to Pajeet, it will be the end of that.

Intel makes potato chips. Send chip making overseas. Drive nails in own coffin.

>inb4 chink backdoor in the nsa backdoor

>That's the entire reason why porting the current one to 10nm is such a clusterfuck that hasn't worked out.
No, that's because their 10nm process is crap.

What design?

Is this the age of Pajeet chips?

India will never be a semiconductor producer of any relevance. The only EE talent they produce is few and far between.

It's literally because of the process. They tied to use EUV, which is unproven at that scale and sophistication, and they failed.

Intel's 10nm node wasn't an EUV one, it was immersion with quad patterning. Their biggest issues likely came from the metal stack they implemented on the node. That coupled with complexity is what did them in. Their 10nm node used more masks than non EUV 7nm nodes from the rest of the industry.
That many masks and high use of triple and quad patterning with immersion litho spells absolutely abysmal yields.

Intel contracting out to GloFo would be hilarious.

14 LPP would be perfect for small core parts given its high yields and relatively good electrostatics with mild clocks. For a platform chipset and a dinky little Atom it'd be perfect to shit them out by the millions.
All of intel's parts are built around 3D fins though, with a really high fin and tight fin pitch. No other comparable process has fin metrics like intel utilizes, so I'm not sure how they'd handle porting that over.

They could likely come close with their 12nm line with some tweaks to improve BEOL scaling a bit more. Global Foundries definitely has the capacity in terms of wafers per month.

>take on low margin business to better utilize fabs
>fab capacity drops because new process problems
>outsource low margin chips
>fix fabs
>bring back outsourced production to own fabs
>???
>profit
Why do you always have to explain business 101 on Jow Forums ?

could also be that capacity didn't drop and the demand is just that fucking high

that's leather jacket man in drag.

Intel is prioritizing modems for Apple

They can't even supply Xeons and their 9th gen is a paper launch

But that's what I'm saying.

No I'm not doubting its viability, just the hilarious irony of intel using what was once AMD's fabs for production of their chips.

Attached: 0b31852bf249d946605e72a1b40b97d4.png (800x800, 639K)

So it's final then, pajeets and chinks have won.

> GloFlo ditches development of 7nm due to expenses and nobody wanting it
> focuses on 12/14nm stuff
> a few weeks later, Intel might contract them to do their 14nm shit

If it's true that Intel's production is strained since at least April, then the whole thing makes alot of sense.

Attached: intrigued.gif (730x413, 11K)

> that's leather jacket man in drag.

She's his niece, so the likeliness isn't just random.

Although it's fascinating when you think about it:
> Leather Jacket Man makes mad cash leading Nvidia
> his niece saves AMD from bankruptcy and brings them back on track

Just what the fuck is up with that family and semiconductors?

Attached: consider the following.jpg (600x600, 55K)

Don't forget that arabs own GloFo now and want to see a short term return on their investments. Abandoning 7nm will end up biting them in the ass eventually, unless they're working on leap frogging down to 5nm.

> Jews begging Arabs for help to make shekels

Attached: dis gon b gud.gif (350x191, 1M)

GloFo's 7nm process was, for the most part, fully production ready except they would have to invest billions into spining up 7nm fabs and/or cannibalizing their already running 14nm fabs
They are sticking with 14nm because most of the market doesn't need 7nm, most of the market is still 22nm or 32nm if not larger processes.
having good volume production with 14LPP will earn them more money than continuing to pursue the feature size pipe dream.

You're incredibly mistaken. Global Foundries stopped developing FinFET processes, but they didn't give up on high performance nodes. They're offering advanced SOI processes that have half the mask count of a comparable FinFET process while delivering similar performance. They're on track to have a few billion in net profit from 22FDX.
NExt year they'll bring 12FDX to market that is comparable in performance to industry 7nm nodes. In They keep up with R&D then in 2020 they should announce 7FDX which competes with leading edge EUV nodes while being less than half the price and having yields in the strong double digits off the bat.

12nm FDX -> 7nm FDX ( later )

12nm FDX sits between the 7nm FinFet and 12nm ( 14nm LPP / LPU )

7nm FDX > 7nm FinFet or something like that

Yes, in theory thats how it'll end up.
7FDX should have power consumption and fmax near 5nm EUV FinFET nodes. 12FDX already hit 90% of their target performance and competes with 7nm FinFET processes.

>You're incredibly mistaken.
Couldn't be happier to be wrong. What type of chips will 12FDX be on? x86 or ARM? GPUs? Network controllers? All the things?

A bunch of stuff on government contracts, solid state lidar sensors, radar sensors for the automotive market, 5g modems, misc RF stuff, some micro controllers, probably some misc ARM parts, nothing big that I know of. They have 52 clients on 22FDX already and they expect current clients to transition to 12FDX, as well as new clients to pick them up for the process.

SOI in recent years passed was skipped over in favor of bulk processes because the starting wafer was more expensive. Though as leading edge FinFET nodes became more and more complex SOI started to win out massively in terms of cost. Now the starting wafer is cheaper so right off the bat theres a cost savings, they use far less masks which means lower complexity and higher yields. They're double gate, FinFETs are basically the same electrically, but these advanced SOI processes can control their gating surface areas separately. They use one gate to alter the properties of the channel while the other controls the current passing through it. That lets them get extremely low power, or extremely high fmax, and they can switch on the fly.

The only caveat as of now is that the process isn't suitable to larger die sizes due to wire capacitance. FinFET processes still hold an advantage there for things like big GPUs. Though this could change as the processes are refined.
Global Foundries isn't going to be hurting at all in years to come.

This is a simple primer on the tech
youtube.com/watch?v=uvV7jcpQ7UY

They'll eventually do EUV with the FDX too at some point

They will, and it will only further decrease mask count, may get rid of double patterning all together if the light source intensity increases enough. Its keeping in line with the trend of being around half the complexity of a competing FinFET process.

Global Foundries isn't the only big company who kept developing high performance advanced SOI, Samsung did as well, though they call theirs FDS. Samsung is coming to market with an 18FDS pretty soon here.

Thanks, guy. You're probably the most helpful and polite person I've met on Jow Forums.

intel owns the patent for the x86 instruction set, TSMC couldn't release a CPU themselves without licensing it

>tfw Chinese
feels good man