Intel Confirms 8th Gen Core will be 14nm

AHAHAHAHAHAHAHA

anandtech.com/show/11115/intel-confirms-8th-gen-core-on-14nm-data-center-first-to-new-nodes

Intel on ABSOLUTE SUICIDE WATCH

No 10nm for you Jew enablers.

Other urls found in this thread:

youtube.com/watch?v=tZzFUmROFLY
wccftech.com/intel-losing-process-lead-analysis-7nm-2022/
seekingalpha.com/article/4042701-intel-forfeiting-process-leadership
twitter.com/SFWRedditGifs

Hahahaahaha Sup Forums on suicide watch

Missing a 3rd optimization for Basin Falls.

Arrandale is still the best microarchitecture for Intel Core I-series processors.

Stagnation is a happening

AYYYYYMD IS FINISHED AND BANKRUPT

youtube.com/watch?v=tZzFUmROFLY

so he lied

Everything till late 2018 will be 14nm. Ryzen won't be better but it's gonna expose insane Intel tax on more cores and overclocking. Intel won't lower current prices because that would pretty much mean admitting that they were selling their hardware at a 2000% profit margin so they will release 8th gen with same IPC and 100MHz higher stock clocks while talking some nonsense about "reducing manufacture prices due to new technologies" which allowed them to cut their prices in half.

Intel House Fires coming to a neighborhood near you!

Stagnation is good for software. All poojeets and shit tier coders will have to learn what optimization is or lose their jobs.

Process > Architecture > Optimization > Panic

And we need 10nm because....

Fucking this. Please let it happen.

Because otherwise Intel are assholes for letting us stay on the same chipset or only changing pins. I mean, I am an autist that loves to spend money on useless benefits, what is your problem?

>Stagnation is good for software. All poojeets and shit tier coders will have to learn what optimization is or lose their jobs.
At the condition that jewltel doesn't release new CPU but helps developper optimize their code...

is this just further confirmation that coffee lake will be 14nm (known like almost a year ago), or something beyond that?

Power effeciency, general faster processors, moar caors, etc

Need? No
Want? Yes

A 35w quad core with the performance equivilent of an lga 1150 i5/i7, thats the 10nm dream

Intel is afraid of 1.7% yields, kek.

WAKE ME UP

>be intel
>release new CPU every year
>performs slightly worse
>slightly more wattage usage

this is fucking pathetic, why not just do it every 2 years instead? Every motherboard manufacturer has like 60 boards for each fucking CPU already

nooooooooooooooo... DELETE THIS RIGHT NOW YOU GODDAM AMD SHILL

gotta sell you a new socket

Maybe this time they'll get the fucking TIM right so people won't have to void their warranty just to get the chip at a average temperature

These aren't desktop SKUs are they?

Cant really be compared to the 7700k and 7600k that just released.

Literally who the fuck cares. Moore's law is long dead, 2006 CPUs are still more than enough for most people.

Why don't they just tweak 14nm a little and call it 10nm like Samsung and TSMC do?

Its uncertain exactly what their "8th gen" arch is. Coffee Lake was supposed to be slated for 2018 instead of 2H 2017. They could have pushed it to market sooner I guess, but their internal Sysmark bullshit shows virtually no performance gains. A 15% uplift in that synthetic benchmark could be nothing but slightly higher clocks and faster memory.

Shitposter go home

Pro tip: Circuit features on Samsung 10nm is much larger than Intel's 10nm targets. Samsung 14nm features are also much larger than Intel 14 nm.

>Everything till late 2018 will be 14nm.
Nope, Snapdragon 835 and Apple A11 are going to be 10nm.

Fake 10nm sure

web devs are making sure that even 5-year old CPUs now have trouble running their javascript web applications

Shitposter, go home.
BEOL does not practically matter for anything but areal scaling, impacting how many potential candidates fit per wafer.

Snapdildos are already well into production actually.

The only ones doing fake nm counts have been Intel.
Their 22nm was more like 27nm, their 14nm process is a 17nm equivalent by everyone else.

Samsung's 14-nm and TSMC's 16-nm node are equivalent to Intel's 22-nm node. Samsung's new 10-nm node is equivalent to Intel's 14-nm node. When Intel launches its 10-nm node, it will be equivalent to Samsung and TSMC's 7-nm node.

>"optimization"
>"optimization"

Why don't you just claim Samsung's 10nm node is equivalent to Intel's 90nm node at this point?

...

>Tick/Tock abandoned
>Tick/Tick/Tick/SPROING implemented

This shit is 2004 all over again.

Thats a professional opinion, and fact. Keep sucking gook dick, their process nodes are always behind the jews.

can't believe he thought it was worth politicizing his company like this. he has to be the worst ceo the company ever had.

that's AMD

Newfag detected

Stop talking about things you're too stupid to understand, little Sup Forums kid.

Intel's 22nm Trigate is very far behind TSMC and Samsung's 20nm nodes in BEOL metrics and SRAM cell size.

CPP :
Intel's 22nm Trigate is 90nm
intel's 14nm Trigate is 70nm
Samsung's 14nm FinFET is 78nm
TSMC's 16nm FinFET is 90nm

Metal pitch
intel's 22nm Trigate is 80nm
intels 14nm Trigate is 52 for their mobile line, 60nm for normal power chips
Samsung's 14nm FinFET is 64nm
TSMC's 16nm FinFET is 64nm

High Density SRAM cell area
Intel's 22nm Trigate is .1080um2
intel's 14nm Trigate is .0588um2
Samsung's 14nm FinFET is .0645um2
TSMC's 16nm FinFET is .0700µm2

Fin pitch
Intel's 22nm Trigate is 60nm
intel's 14nm Trigate is 42nm
Samsung's 14nm FinFET is 48nm
TSMC's 16nm FinFET is 48nm

You don't know what you're talking about.
Area scaling does not influence transistor characteristics.

Literally every professional analyst disagrees with most of your points

Why would you be salty about this? It means I don't have to buy a new motherboard if I want to upgrade my ram

Nice false appeal to authority, you 15 year old shit for brains. Everything I posted is strictly factual.
The BEOL determines how far apart structures on the FEOL can be placed, it doesn't directly influence the electrostatic characteristics of a transistor. Gate performance is entirely a factor of the gate itself, not how the contacts are carried to source and drain wells.

reddit spacing

silly goy forgetting the inevitable socket change. haven't had one since 2015, it's overdue already!

You are an absolute nobody that knows absolutely less than you think you do.
Gooktech is further behind Intel.
Garliceaters just name their way to parity with Intel with respect to process nodes.
You can regurgitate buzzwords all you want, everyone knows gooks are dishonest.


Fact.

wccftech.com/intel-losing-process-lead-analysis-7nm-2022/

seekingalpha.com/article/4042701-intel-forfeiting-process-leadership

>Argues his point is fact
>"Industry professionals"
>Posts WCCF

Confirmed for low effort troll.

>gooks are dishonest
>cites currytech

You're a dumb little Sup Forums kid desperately trying to argue with someone actually in EE.

Process nodes have public facing names that reflect minimum feature size, minimum feature size just so happens to line up with ASML guidelines for node names. This is the same for everyone. No one lies about anything. You simply don't understand this concept because you're a retard.
Neither of those links have anything to do with the topic at hand. BEOL scaling does not impact the switching characteristics of a transistor on the FEOL.

> numbers
> other people don't agree with your numbers!!1
Ok.
You could at least try and link some of your professional sources.

So refute it.
It should be exceedingly apparent with a bare minimum of research that gooks are less than honest with naming nodes. Intel will actually shrink everything, when gooks just shrink part of the node.
Gook names are literal marketing buzzwords. Im sorry you fell for the lies.

>Not smart enough to argue
>Posts meme pics instead

They literally just make up names. They took a 20nm and stuck new transistors on the end and called it 16nm. And your probably that fag who cried about appealing to authority? You are literally nobody.

No, you retarded little kid. They don't make up names.
TSMC and Samsung both leveraged their existing 20nm BEOL, and developed a new FEOL. These are their transistional FinFET nodes. Each respective foundry named their process for the minimum feature size that they can etch. Samsung could expose and etch a 14nm long line, TSMC could expose and etch a 16nm long line. That is the limit of the lith mask resolution and light source employed by those lines, that is what determines minimum feature size, and that is what their marketing names are based off of.

Again:
1.) BEOL scaling does not impact the switching characteristics of a transistor on the FEOL
2.) No one lies about anything when it comes to a public facing process node name. Not intel, not TSMC, not Samsung, not Global Foundries, not Chinese UMC, not any other foundry.
You're simply too much of an idiot to understand this simple concept.

>being this much of an avatarfagging sissy faggot

>Samsung's 14-nm and TSMC's 16-nm node are equivalent to Intel's 22-nm node. Samsung's new 10-nm node is equivalent to Intel's 14-nm node. When Intel launches its 10-nm node, it will be equivalent to Samsung and TSMC's 7-nm node.
Like it or not, the garlic eaters are still behind Intels node advantage

Salty Carmel is the color of your skin, correct?

yep
you must like them brown boys

Ayy the only way for Intel to compete with Ayymd is moarcores moarGhz housefires. History literally repeats itself.

>>Samsung's 14-nm and TSMC's 16-nm node are equivalent to Intel's 22-nm node
This is simply not true, and I've already provided ample proof: Intel's 22nm Trigate is very far behind the 20nm planar nodes in SRAM cell size, metal pitch, only matching TSMC's 20nm back end in connected poly pitch, Samsung has an advantage here over both of them.
>Samsung's new 10-nm node is equivalent to Intel's 14-nm node. When Intel launches its 10-nm node, it will be equivalent to Samsung and TSMC's 7-nm node.
Yet again, BEOL scaling does not impact switching performance of a transistor on the FEOL. Samsung and TSMC haven't detailed what gate topology their 7nm nodes will use. All we know is that they'll employ EUV. We also know that intel's 10nm node is their 3rd iteration of their Trigates. Samsung or TSMC could bring a GAA process to market, Samsung already has ample experience in fabbing GAA structures, their 3D VNAND is a GAA device.


Stop trying to talk out of your ass, you know nothing about this topic. You're parroting currytech clickbait with is entirely devoid of any factual information.

Thanks for confirming gooks nodes are marketing terms that have little basis in reality

Nice damage control. And Ryzen is not even released.

>I try to argue with an informed and educated person, get absolutely blown the fuck out, and resort to nothing but childish shitposting

Top tier bait

wasn't amd saying they were going to jump 10nm and go directly to 7nm?

either way, intel 6xxx, 7xxx and 8xxx series are basically the same

AMD has no fabs, but yes, GloFo will skip 10nm in favor of IBM's 7nm SOI FinFET process.

AMD didn't say they were going to skip a 10nm process, though they likely will.
Its Global Foundries who will not offer a 10nm process, they're bringing a 7nm SOI based FinFET node online. AMD amended their wafer supply agreement with Global Foundries to have access to the upcoming 7nm node, and have only agreed to buy 14nm and 7nm wafers from them.
They'll likely have 7nm parts on the market sometime in 2019. The key takeaway here isn't the feature size, its that the process is advanced SOI instead of bulk silicon. Its going to be the first of its kind on the market.

IBM selling off their foundry business was a godsend for Global Foundries.

INTEL IS FUCKING DEAD

ARM WILL TAKE OVER 100%

Im mobile now, but Ill start a thread this weekend.

His points are marketing only. Gooks shrink part of a previous node, and call it a new node. At least Intel had the decency to shrink the whole node. Guess who has better density and profit margins? Intel. The gooks are jewing you guys, and you are more than willing to suck their marketing cocks because you absolutely hate Intel. Its ok, though Intel could literally rename their existing, and future nodes, and still be in spec, and still be ahead of the gooks in feature size, density, and profit margin.

You what?

yeah but a 2006 mobo will shaft performace up the dingus
>DDR2
>sata 2 if you're lucky
>not even PCI-E 2 let alone 3
if all you're doing is watching youtube and playing flash games then sure

You're laughably uninformed, flat out a low IQ little fanboy.
Intel's 14nm node was their worst in recent history for cost, yields, and time to refine it. Desktop Broadwell was shitcanned because intel was losing money on every die sold. That was after it was delayed heavily. They rushed out Skylake, and even that took an inordinate amount of time for volume of high clocking desktop parts to actually reach retailers. After it was officially released it was nowhere to be found for near two months, all because intel couldn't deliver. They're having the exact same issues with their 10nm process.

Samsung never had any such issues with their 14nm LPE and LPP. TSMC had their 16nm online a bit later, but yields were strong right off the bat, and had no problem improving.
Intel's advantage is in back end of line, and nothing more. They will keep selling 14nm parts into 2018 because they need to recoup all the money they've lost trouble shooting their buggy process. Similarly they will keep on 10nm for 3-4 years as well, because their foundry business is faltering.

Yet again: Back end scaling does not impact performance of the FEOL.
Also again: No one is lying about the public facing marketing name of a process. No one. You're grasping at straws, little kid.

ARM still has a long way to go. A very long way.
The A73 is incredibly impressive though. Pretty solid clocks, short pipeline, very low power.
I'd like to see a wider ARM design like this implement SMT, like a MIPS Warrior core.

Quite a long way since the Cortex A9 cores in my Galaxy S3.

>7nm SOI based FinFET node

ex-IBM IP? IIRC IBM had a real hard on for SOI

You betcha. IBM sold its foundry business to GloFo.

>but yields were strong right off the bat
No, TSMC had yield issues from the start. Don't you remember the GoyworksThermi 2.0: The Re-7% Yielding last year?

Yep. Part of the deal with taking control of IBM's foundry business was getting all their process IP, tooling, and engineering staff. In exchange Global Foundries has to produce chips for IBM, but they basically got handed a golden ticket.
All the R&D that IBM did for their own POWER chips is now at their disposal.

They have a 14nm SOI FinFET process online right now called 14 HP, but its uncertain if its actually being used. No one has gotten confirmation on what process the new line of POWER9 chips is made on.

I'm still sad that IBM's CEO sold off one of their most interesting business units to hire more pajeets.

I know they were making a loss but they didn't seem to push to get any wins from chip makers on their processes.

At least they sold it to the right dudes that actually make chips.

No, 14HP is not used. GloFo uses licensed 14nm LPP from Samsung, and yields are pretty good based on poolaris line.

Dumb question, but what are the benefits of SOI vs bulk CMOS and why is SOI relatively unpopular?

Lower yields on large parts is normal, they didn't have any issues with 100mm2~ dies off the bat.
Intel's 14nm node first launched with Broadwell CoreM, and yields were poor despite the small die size.
Thats a pretty big difference.

I wouldn't say that Global Foundries is a great business, but with IBM IP and engineering talent being brought into the fold they might become one. Its one hell of a stimulus. Enough to potentially change the entire future of the company.


SOI as the name implies has a layer of insulating oxide on top of the base silicon. This increases the raw cost of the wafer, usually by a few thousand dollars compared to a bulk one.
The benefits are in better channel isolation, structures are simplified leading to less mask layers being needed, far less dopants are needed which leads to greater uniformity, and lower defect density, etc. There are tons of benefits. The main negative is just the increased starting price of the wafer.
SOI based processes have been pretty common in the past for CPUs, they only fell out of favor in recent history, but the complexity of short channel devices is tipping the scales back in their favor.

Historically IBM and AMD both leveraged SOI processes for their chips.

Thanks for your detailed explanation user. Nice to know that Sup Forums isn't totally devoid of any interesting technical discussions.

>structures are simplified leading to less mask layers being needed

So would this avoid the need for multiple patterning? -> quicker cycle times

>So would this avoid the need for multiple patterning?

No because multiple patterning is about the wavelength of the light source and how well it etches the silicon in one pass. Some layers will need more than one exposure no matter what the starting wafer is. When I say it reduces complexity its because that insulating oxide is already there. Certain structures would need channels cut and an insulator to be laid down before they could be built up if you were using bulk. With an SOI wafer you can remove this step since you already have isolation. Less steps adding material, removing material, adding material again to build the same types of structures.
STMicro has some decent side by side comparisons somewhere on their website.

Do you consider GP104 dies to be "big"?

Intel has been stagnating since the Sandy bridge.

Incremental die shrinks and marginal performance boosts between generations is their death. AMD couldn't compete and at least tried their hands at different market sectors before coming back with Ryzen

314mm2~ is pretty big for a 16nm part, especially on a brand new node.

>marginal performance boosts
Most people only care about the better efficiency, faster eGPU, and better hardware features. The processors are already ridiculously fast.

That's fine, they'll throw in a couple of extra cores for each model and crush AMD once again.

It'll be Presshot 2.0 with moar cores instead of longer pipeline.

remember to buy Netburst :')

Underrated post

They are not yet panicking. Real panic will start the moment they'll oficially announce new goyim-friendly SKUs.

>>Tick/Tick/Tick/SPROING

Please come see me for an offer you can't refuse. Our Marketing Department needs people like you.

>new goyim-friendly SKUs
Like what?

AMD has yet to beat Intel yet. People are so confident Ryzen is going to blow Intel CPUs out of the water. It's going to be funny as fuck if Ryfail comes out and people find out it was nothing but unfulfilled promises from AMD yet again.

i5 with hyper-threadan, buuuut it's on x299. Be a good goy and buy it.

>intel rep on stage
>cart rolls out full of processors
>huge angry voice: "10NM????"
>n-no
>arms fall off, instantly dies, stage implodes, just silence

>3 years later, new stage, new rep
>cart fucking rolls out again
>stares at CPUs
>"10NM!?!?!??!?!"
>.....n-no
>arms, dies, stage, etc, silence

>23 years later
>cart rolls out
>they're all labeled "AMD"
>"HAHAHAHA NO"
>Rep doesn't say shit, as arms dies stage silence etc.

>150 years later, new stage and rep of course
>cart rolls out yet again
>"HI THERE"
>y-yes
>"SO, THE CPUS"
>y-yes
>"ARE THEY 10NM"
>momentary pause
>produces sledgehammer and starts crushing cpus
>"they are now, faggot!!!!!!!!!!!!"
>"KAY"
>cart slowly rolls back
>8th Gen Core on Sale Now