Intel Foundry Customer Bails Out: 10nm Process Not Commercially Viable

Megalith

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
SemiAccurate is reporting that one of Intel’s marquee foundry customers has just jumped ship: while their identity appears to be paywalled, the loss is due to Intel’s supposed failure in developing a viable 10nm process. These products were supposed to be out more than two years ago, and it still remains unclear when they will truly debut.

…Intel can’t make 10nm chips, has a customer list that can be counted on one hand, and the ones of those with financially significant volumes can be counted on one finger. Progress is being made though, and customers should be happy with the timelines Intel is promising. And delivering. Page 11 notwithstanding. Since the 2013/2014 announcements, several others, a few big names included, have officially signed on. This ‘wave’ was set into motion by ARM announcing they would be working with Intel as a foundry. Big names followed. One of the largest just left.
 
Well that's going to make things interesting for post coffee-lake. 14nm +++? LOL

I wonder who the major vendor was? And I thought Intel HORDED all new process node production capabilities for their latest and greatest chips.
 
Just throwing this out there, but you might want to be careful quoting Semi Accurate. They're basically an AMD fan blog. They've been writing stories on the demise of Intel and NV for years.

This story could be true, but you might want to try and find a different source.
 
akaliel said:
Just throwing this out there, but you might want to be careful quoting Semi Accurate. They're basically an AMD fan blog. They've been writing stories on the demise of Intel and NV for years.

This story could be true, but you might want to try and find a different source.
Click to expand...

Years ago Intel was boasting that a 10nm cpu was coming any time now. And that still has not materialized. Doesn't that tell you enough ?

Btw. the full article is behind the pay wall so not that interesting to begin with.
 
I hope it's Apple. AMD's APUs would be very nice in 15-28W if that's even on the cards.

#wishfulthinking
 
Pieter3dnow said:
Years ago Intel was boasting that a 10nm cpu was coming any time now. And that still has not materialized. Doesn't that tell you enough ?

Btw. the full article is behind the pay wall so not that interesting to begin with.
Click to expand...

I never said the story wasn't true, just that they might want another source. It's like seeing an article about how Google's new Pixel phone sucks from 9to5mac.com
 
How advantageous is the transition from 14nm to 10nm going to be, realistically? 28nm to 14nm for GPUs was pretty major in terms of both IPC gain and energy savings. But 22nm Haswell to 14nm Broadwell seemed less impressive...? How much of a difference can 14nm to 10nm make? Is this more relevant to mobile processing?
 
10nm has been a nightmare for Intel and from the sounds of it, is a continuing nightmare for them. Would not be surprised if they scrap 10nm and go to 7nm instead at this point.
 
DigitalGriffin said:
Well that's going to make things interesting for post coffee-lake. 14nm +++? LOL

I wonder who the major vendor was? And I thought Intel HORDED all new process node production capabilities for their latest and greatest chips.
Click to expand...

Where i believe you are right that Intel usually keeps the newest process node to themselves my guess is that it is getting so expensive Intel wants to recoup their money faster.
 
Factum said:
I think you have grosely misunderstood "Moore's law".
It's dosent talk about performance, but the number of transistor/price....

https://en.wikipedia.org/wiki/Moore's_law
Click to expand...

I think you have grossly misunderstood a joke as a technical understanding...:rolleyes:


Also from you own goddamned link:
"The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months"
Hence the saying is from INTEL aka the company of this post, not a quote of mores law.

But i guess you where to busy trying to score cheap internet points rather than reading your own "evidence"


maxresdefault.jpg
 
Last edited:
And yet reality confirms my point:
https://upload.wikimedia.org/wikipedia/commons/0/00/Transistor_Count_and_Moore's_Law_-_2011.svg

You should stick to posting pics...arguments are not your strong suit...not when you have to cherry quote like that, let take the FULL quote:


The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and the transistors being faster).[9]
Click to expand...


Is a "prediction" about Moore's Law:
Moore's law is the observation that the number of transistors in a dense integrated circuit doubles approximately every two years. The observation is named after Gordon Moore, the co-founder of Fairchild Semiconductor and Intel, whose 1965 paper described a doubling every year in the number of components per integrated circuit,[2] and projected this rate of growth would continue for at least another decade.[3] In 1975,[4] looking forward to the next decade,[5] he revised the forecast to doubling every two years.
Click to expand...

Anyone can try and "predict" stuff...doesn't alter anything.

I will refrain from pictures, as I am over 40...ball in your court.
 
Factum said:
And yet reality confirms my point:
.
Click to expand...

You point is not the debate here but nice trying to move the goal post . You tried to play it smuck and correct my quotation of David ( name not known at the time though).
But because you didn't understand/know off the quote you instead referredd to more laws which was no the one i was referring to.

In short: You correction was wrong as it was not what my post was about not matter how much you are right about mores law.

But i guess it will be hard time before you can admit you correction was indeed invaliad


I will refrain from pictures, as I am over 40...ball in your court.
That's sad though, maybe try to get a bit of humor, maybe you would have understood the joke to begin with.
 
Gigantopithecus said:
How advantageous is the transition from 14nm to 10nm going to be, realistically? 28nm to 14nm for GPUs was pretty major in terms of both IPC gain and energy savings. But 22nm Haswell to 14nm Broadwell seemed less impressive...?
Click to expand...
Throwing this out there --> with no high-end competition at the time from AMD I don't think the "less impressive" result from Intel going from 22 to 14 has anything to do with the process. Like you said in the GPU space it was huge going from 28 to 14.
 
Bet the suits at Intel are now wishing they listened to their engineers and invested way more into graphene and carbon nano-tubes R&D a long time ago. Hindsight and all...
 
If Samsung and tsmc are slated to get to 7nm as early as the end of 2018, I don't see the point of putting out a 10nm chip this late in the game.

The intel foundry edge is slipping more and more, they could count on being the big chip maker giving them the edge when it was mostly x86 processors, but with arm chips going in hundreds of millions of phones each year, if not billions, and memory that goes in almost all electronic devices, and graphics cards, there is too much demand the world over to have intel be the dominant fab player long term. Intel used to be at least a node ahead of other fabs. That's gone now, they reportedly still have the best 14nm process, but that's not an entire node ahead of everyone else.
 
SvenBent said:
You point is not the debate here but nice trying to move the goal post . You tried to play it smuck and correct my quotation of David ( name not known at the time though).
But because you didn't understand/know off the quote you instead referredd to more laws which was no the one i was referring to.

In short: You correction was wrong as it was not what my post was about not matter how much you are right about mores law.

But i guess it will be hard time before you can admit you correction was indeed invaliad


I will refrain from pictures, as I am over 40...ball in your court.
That's sad though, maybe try to get a bit of humor, maybe you would have understood the joke to begin with.
Click to expand...

I am talking about Moore's law.

You are talking about House's prediction about Moore's law...you must really like PR.
 
If freaking Intel is having issues with 10nm how realistic is 7 for Intel or anyone really? .. supposedly Vega 20 will be 7nm from TSMC .. I .mean i think we are going into the single digit atom numbers now.
 
Uvaman2 said:
If freaking Intel is having issues with 10nm how realistic is 7 for Intel or anyone really? .. supposedly Vega 20 will be 7nm from TSMC .. I .mean i think we are going into the single digit atom numbers now.
Click to expand...

We're a ways off from that small...the largest atoms have a diameter of around 0.4-0.5 nm.
 
Uvaman2 said:
If freaking Intel is having issues with 10nm how realistic is 7 for Intel or anyone really? .. supposedly Vega 20 will be 7nm from TSMC .. I .mean i think we are going into the single digit atom numbers now.
Click to expand...

tybert7 said:
If Samsung and tsmc are slated to get to 7nm as early as the end of 2018, I don't see the point of putting out a 10nm chip this late in the game.
Click to expand...

SA isn't a reliable source. Also be careful mixing nodes numbers across foundries because they cant be compared at all. Specially not after the node number inflation.

Also for anyone but Intel cost per transistor increases.

Cell-SizeComparison.png

Intel-vs-TSMC-10nm.png

10nm-Hyper-Scaling.png
 
Last edited:
Factum said:
I am talking about Moore's law.

You are talking about House's prediction about Moore's law...you must really like PR.
Click to expand...
You are talking about something he never mentioned so whom were you correcting really?
 
Shintai said:
SA isn't a reliable source. Also be careful mixing nodes numbers across foundries because they cant be compared at all. Specially not after the node number inflation.

Also for anyone but Intel cost per transistor increases.

Cell-SizeComparison.png

Intel-vs-TSMC-10nm.png

10nm-Hyper-Scaling.png
Click to expand...


And how does their 10nm compare to the 7nm target of the other fabs? As far as I can tell global foundries are targeting their own 7nm parts while intel is on their 10nm chips.


https://www.extremetech.com/computi...samsung-globalfoundries-intel-gear-7nm-beyond


So intel 10nm will likely be put head to head not against a potentially lower performing samsung 10nm part, but a 7nm part. We'll see how all that shakes out, I'm just saying intel does not have some clear cut domination of the space anymore, too much money, too many other chips and memory being made for them to stay that dominant. If I'm wrong, I'll be surprised.
 
tybert7 said:
And how does their 10nm compare to the 7nm target of the other fabs? As far as I can tell global foundries are targeting their own 7nm parts while intel is on their 10nm chips.


https://www.extremetech.com/computi...samsung-globalfoundries-intel-gear-7nm-beyond


So intel 10nm will likely be put head to head not against a potentially lower performing samsung 10nm part, but a 7nm part. We'll see how all that shakes out, I'm just saying intel does not have some clear cut domination of the space anymore, too much money, too many other chips and memory being made for them to stay that dominant. If I'm wrong, I'll be surprised.
Click to expand...

You are reading numbers when you should be weary that most of the non Intel fabs are referring to marketing smugness (AKA Nano Marketing).

From the site which is on the ball about this: https://semiaccurate.com/2016/09/26/globalfoundries-7nm-process-isnt-even-close-name/

Although games around process numbers have been going on for a while, much of this started out at the 14/16nm node. If you recall both TSMC and the Samsung/GF processes were nothing more than their 20nm planar nodes with FinFET transistors slapped in, the back-end of line (BEoL) was the same. This meant the shrink delivered by the process was something like zero. Rather than take the honest road and call it 20nm FinFET or something similar, they called it 14 to intone a full shrink and the technical merit that it carried.
Click to expand...
 
Shintai said:
SA isn't a reliable source. Also be careful mixing nodes numbers across foundries because they cant be compared at all. Specially not after the node number inflation.

Also for anyone but Intel cost per transistor increases.

Cell-SizeComparison.png

Intel-vs-TSMC-10nm.png

10nm-Hyper-Scaling.png
Click to expand...


Cool chart but so far that is the only thing Intel has released on 10nm.
 
While I am no Intel fanboy, they are the best in the WORLD when it comes to bleeding edge IC manufacturing. That isn't going to change anytime soon unless Samsung decides to take $50+ billion and move their entire 14nm production to exotic materials and we finally see EUV patterning become used outside of small, high risk test runs. I can't believe that anyone here would actually believe that TSMC has a "12nm" and GF has a "7nm" node. Its PURE marketing folks. The gate pitches haven't changed. Sure they have made minor tweaks to get the yields improved but that is a far cry from having an actual 12/7nm chip.

I would LOVE LOVE LOVE to see a Threadripper 2/Intel 79xx 20C/40t+ cpu on a true 7nm process. EUV patterning and graphene/carbon nanotubes are going to give us the last HUGE performance jump we will in our lifetime. It will be the pinnacle of IC design and engineering in most of our lifetimes until we get a break in true Quantum Computing or true AI.
 
Gigantopithecus said:
How advantageous is the transition from 14nm to 10nm going to be, realistically? 28nm to 14nm for GPUs was pretty major in terms of both IPC gain and energy savings. But 22nm Haswell to 14nm Broadwell seemed less impressive...? How much of a difference can 14nm to 10nm make? Is this more relevant to mobile processing?
Click to expand...

It is very easy to increase "IPC" of GPUs because GPU workload is highly parallel and the tactics of more cores actually work. Go to smaller transistors, you can squeeze more cores on the die. That doesn't work for CPUs.
 
Gigantopithecus said:
How advantageous is the transition from 14nm to 10nm going to be, realistically? 28nm to 14nm for GPUs was pretty major in terms of both IPC gain and energy savings. But 22nm Haswell to 14nm Broadwell seemed less impressive...? How much of a difference can 14nm to 10nm make? Is this more relevant to mobile processing?
Click to expand...
32nm Sandy to 14nm Broadwell is unimpressive, so 14 to 10 doesn't look like much of a gain unless there is a major shift in architecture.
 
There might be more to this than just rumour.

If you watch the Nvidia tech tour from a year or two ago, they stated they were sampling 10nm in the defect lab.
IIRC Volta suddenly went from 16nm, to supposedly 10nm, then to 12nm.... but nm is marketing number these days so who knows what the hell is going on.
 
Uvaman2 said:
If freaking Intel is having issues with 10nm how realistic is 7 for Intel or anyone really? .. supposedly Vega 20 will be 7nm from TSMC .. I .mean i think we are going into the single digit atom numbers now.
Click to expand...
Wait, Global Foundries doesn't have AMD by the balls anymore? Will 7nm really be 7nm?
 
You must log in or register to reply here.
Back
Top