David Schor: Intel 10nm in big problems

witeken

Diamond Member
Dec 25, 2013
3,899
193
106
David Schor from WikiChip:
David Schor said:
I hope I am wrong but I have good reasons to believe there won't be ANY 10nm volume this year at all. From what I heard, I really hope they have a "Champagne Lake" (made up name) lined up for 2019. They are going to need it.

Everything is true here:
Charlie Demerjian‏ said:
There are three distinct problems with #Intel 10nm process that I know of. I am willing to bet there are many more. Of the three only one has a solution that I am aware of, and that was the easy one.

Charlie also says:
Let me explain this. Intel's 10nm process is years late, underperforming, and PR has curious explanations. I have heard about 3 distinct issues leading to the borking, but I have no doubt those are the only issued facing bringing 10nm to market.

Of those three, I have directly heard one has been root caused, a fix was made, and the fix did what it should have. So in short there were 3 problems each of which crippled 10nm chips. Now there are two remaining, and again with the caveat of, "that I know about"

https://twitter.com/david_schor/status/988107149036478464

A couple days ago, Ashraf also posted this story: https://twitter.com/TMFChipFool/status/987286420132958208

Ashraf Eassa said:
A story about Intel process tech, sit down, my followers, and enjoy.

1/ Apparently, a while back, Intel hired outside consultants to benchmark their progress against the foundries. Those consultants informed Intel that they were behind the foundries in some key areas.

2/ Intel, of course, believing in its own superiority basically didn't take the outside consultants seriously, thinking they were full of shit.

3/ Later on, Intel would start to get clues from key semiconductor equipment makers. Intel would, I am told, refuse help from these equipment makers, again believing in its own superiority.

4/ Eventually, when the equipment makers started to tell Intel things like, "we're surprised you're still struggling with Problem XYZ because Competitor X/Y figured that out a while go..." did Intel finally start to become more receptive to input.

5/ This story points to a fundamental problem within Intel: They actually believe their own marketing bullshit about being way ahead of the foundries -- it's a corporate culture thing. This arrogance hurts them -- badly.

Some additional info on Intel 10nm: Intel is trying its absolute best to get things working, they are, from what I hear, sparing no expense on 10nm and they are literally trying everything, even going to far as to pull unproven experimental stuff from the research labs.

One thing is clear: Intel 10nm is the biggest process related disaster in the company's history, it really makes the 14nm struggles look like a walk in the park by comparison. It is truly a horror show for this process.

Lastly, former Principal Engineer Francois Piednoel has also weighed in (he says the problem will be fixed "shortly"): https://twitter.com/FPiednoel/with_replies

Francois Piednoel said:
You are confusing marketing “naming” of the process and the transistor performance at peak performance level. There are still no match for intel 14nm++ transistor,I can bet a lot that you will see a huge increase in idle power of this new 7nm. What Apple ship in 2018 will tell

You guys will did not learn what happen when intel is in trouble? They set up dungeons, super secretive ones, then, they go dark, for Core2Duo, we were 10 with the real story, the rest of the world was predicting AMD taking over.

Today, I agree that there was management mistake, ICL should have been pull down to 14nm++, process tech should not hold back architecture, but that should tell you that chipzilla is now planning a double whammy, and when this will come out, that is going to break teeth.

And historically, it is a very bad idea to bid again intel recovering, they always do, because that their scale, there are excellent people, the management of nice guys just need to empower them, instead of protecting their sits. IDC here I come again ...

So, when you “hear” stuffs , you most likely hear from an non inform person, so, value is useless, I remember being in meeting with intel senior fellows and not being able to tell them what was going on with Core2Duo, and being destroyed because Cedarmill was sucking ass at SSE3

Few days before we showed Core2Duo to the press for the 1st time, most of intel Top management had no clue. Slowly, those VPs got moved outside intel or retired, I predict this is going to happen again. (Please don’t ask names, they are pretty obvious, but by respect , don’t)

I was in the meeting when we canned Tejas. My friends worked hard on that, it was a cathedral of architecture, sadly, the physic limits killed it, many ideas got recycled in processors you use today. Most people miss-understand what was Tejas, because the fanboys only know the name.

I know the reason of all of those delays, but I can't say it because I am under NDA. It is fixable, and my estimate is that it should be fixed shortly.

Yes, the people working on C2D were in IDC, and the information about the performance of C2D was "roomed", non of the Oregon team had any clue of what was coming (except one apple related), because you did not want to discourage the guys trying to make netbust go faster.

Fellows are very smart, very very smart (There are exceptions ... don't ask ;-) , but if they don't need to know, they don't get to know, especially if you plan to wipe competition as C2D did.

My lawyer agree that in 2019, my knowledge of what is going on at intel will be less significant, and I will be allowed to start speaking.

The ARM camps has for sure won the "perception" battle, at least the "nm" naming battle, those 2 slides are comparing and , on the top of this, non of those graph tell you about the Cdyn other important factors, or leakage at high voltage and low voltage, it is to compare

Keep in mind that increasing IPC is not linear to the increase of transistors,the amount of IPC increase will depend on a lot of simulation, how accurate they are, and the scale of your R&D, this is where IPC increase is going. Now,chipzilla has 3 years of architecture to release

For the record, this is what Mark Bohr said in May last year.
Mark Bohr said:
We’re on track to begin shipping 10nm products in the second half of this year. 7nm technology is already well into its development phase, so I know it will happen. And I’m spending most of my time on our 5nm technology, and we have some interesting and viable ideas to make that happen.

I really have no idea at all how Intel screwed up this badly. I am completely baffled. The phrase "10nm is broken" isn't an overstatement.
 
Last edited:

Mockingbird

Senior member
Feb 12, 2017
733
741
106
So, at this point, who is doing worse?

Intel or GlobalFoundries?

Should we expect mass availability of Intel's 10nm and of GloFo's 7nm next year?
 
  • Like
Reactions: PeterScott

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
Intel is dwarfed by Samsung today. Samsung q4 17 investement equaled the rest of the entire industry.

They can keep saying this is delay. Nm is marketing numbers whatnot. That 10nm is so superior when it comes. Blahh

At the end of the day Intel need to give investors less and use far more on process. Double digit billions extra per year is needed to stay a fab house. Get a business model that supports it or be gone in 20 years. Or use a model like eg amd where basic research is done outside.

This is NOT solved by an engineering fix. Far from. Its not 10nm process problems.
 
  • Like
Reactions: Yuriman

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Francois Piednol referencing the Core 2 Duo days is so out of context and shows how badly he needs a reality check. In those days Intel was anywhere from 18-24 months ahead on process node. The first 65nm Intel CPU (Core Solo and Core Duo Yonah) came out in Jan 2006 . For comparison 8800GT the first 65nm Nvidia GPU came out in late Oct 2007. Thats around 21 months.

https://en.wikipedia.org/wiki/Intel_Core
https://en.wikipedia.org/wiki/GeForce_8_series#8800_GT

Intel has no process time to market lead as TSMC N7 is in volume production and Intel 10nm is nowhere to be seen and will not be till atleast late 2018 or early 2019. Intel's dominance in x86 is linked directly to their process lead. They do develop very good microarchitectures but its the process which keeps them firmly ahead. That is valid even today with CFL on 14++ vs Zen + on 12LP. With Intel's process lead gone I think it will be difficult for Intel to run away from AMD or vice versa. I think Intel will have to vastly improve execution to not fall behind TSMC further. IMO both these x86 vendors will compete aggressively and the winner is the consumer.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Francois Piednol referencing the Core 2 Duo days is so out of context and shows how badly he needs a reality check. In those days Intel was anywhere from 18-24 months ahead on process node.

It can be argued Netburst problems were due to process too. Back then they believed process gains will continue and 10GHz processors would have been no problem. The Pentium 4 era was when voltage scaling stopped, meaning a huge portion of power reductions moving to a new process were suddenly gone.

0.13u was when the voltage scaling stopped, at 1.4V. That's pretty much where we are at today with our 14nm processors. Imagine if it never stopped though. 1.1V at 90nm, 0.8V at 65nm, 0.55V at 45nm.

The "demands" that Intel can their fabs and move to TSMC is at the moment, not based on reality. Coffeelake reaching 5GHz is very likely a result due to years of optimizing their process for the highest performance. What they need is realizing as a company process gains are going to be smaller than ever and designing around that reality.
 

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,248
136
Francois Piednol referencing the Core 2 Duo days is so out of context and shows how badly he needs a reality check. In those days Intel was anywhere from 18-24 months ahead on process node. The first 65nm Intel CPU Core Solo )and Core Duo Yonah)came out in Jan 2016 . For comparison 8800GT the first 65nm Nvidia GPU came out in late Oct 2007. Thats around 21 months.

I guess you meant 2006.
 

JDG1980

Golden Member
Jul 18, 2013
1,663
570
136
Francois Piednoel is wrong about Conroe. It wasn't a secret project that came out of nowhere; it was an evolution of the Pentium M, which was already a superior design to Netburst, but had been previously relegated to laptops for internal political reasons. Back in May 2005, over a full year before Conroe's release, Tom's Hardware ran an article about how the Pentium M could compete with AMD's best chips, including a concluding paragraph about how Intel's upcoming Conroe design would be based on it. This flatly contradicts Piednoel's claim that the press (and even Intel management!) had no idea about Conroe until just before it came out.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Use NIL and 3D-IC use 10nm process twice, call it 7nm. End of credits. (tip: use 14nm twice to get 10nm)

What I got from Intel^. (Intel is cutting its relationship with ASML for Canon Nanotechnologies, Inc.)
 
Last edited:

Dolan

Junior Member
Dec 25, 2017
14
10
51
Witeken is just troling. Even if they fires him today, he will keep loving them...

Just look at link in his description: https://forums.anandtech.com/threads/apple-a10-geekbench-4-score.2485468/page-4#post-38464277

It was 2016, he already knew how Intel screwed with 14nm (paperlaunch in 2014 with no silicon available, limited availability in 2015 and first bigger silicon in 2016). And he still blindly believed his favorite while trashing others.

How it is now different?

If i can bet: it is kind of "guerrilla marketing". After earnings he will come with something like "based on BK statements 10nm is even better than expected".

We do not allow personal attacks here,
or using terms like "trolling or fanboy".

https://forums.anandtech.com/threads/anandtech-forum-guidelines.60552/

AT Mod Usandthem
 
Last edited by a moderator:

Atari2600

Golden Member
Nov 22, 2016
1,409
1,655
136
Ashraf Eassa said:
2/ Intel, of course, believing in its own superiority basically didn't take the outside consultants seriously, thinking they were full of shit.

Francois Piednoel said:
You are confusing marketing “naming” of the process and the transistor performance at peak performance level. There are still no match for intel 14nm++ transistor,

Arrogant and full of redacted.

[Cast your mind back 12 months - how good were his predictions of how far short Zen would fall?]

There is a ZERO tolerance policy of no profanity in the tech sub-forums.
You can't seem to grasp this concept as you have done it 3 times before.
Perhaps some time off will help you think about it.

Iron Woode
Super Moderator
 
Last edited by a moderator:
  • Like
Reactions: raghu78

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Arrogant and full of redacted.

[Cast your mind back 12 months - how good were his predictions of how far short Zen would fall?]

You are spot on. Well said. Francois exemplifies what is wrong about Intel. Arrogant and disllusioned about Intel's superiority over foundries in process tech (which is now proven false as TSMC N7 is in volume production while Intel 10nm is nowhere close to) and over AMD in architecture (given the fact that AMD have caught up to Broadwell in terms of IPC and only clocks are saving Intel which will get rectified with 7nm Ryzen). I think Zen 2 vs Icelake will prove the point that Intel has no superiority whatsover in process or architecture.
 
Last edited by a moderator:

ksec

Senior member
Mar 5, 2010
420
117
116
Interesting, it was only a few days ago I had this thought. I had wanted to post a thread here right before WWDC, just to give it one month time before confirming.

7nm from TSMC is basically clock work because of iPhone. Following the pattern iPad Pro A11x, will be used as 7nm pipe cleaner. WWDC is in a little more then a month time, and yet Intel still hasn't shipped a single 10nm product.

So in a few months time, TSMC will officially pass Intel in leading node HVM. We are talking about TSMC 7nm being roughly similar to Intel 10nm. With some pros and cons on both side depending on application. But TSMC will have it first in high volume.

Funny how Ashraf Eassa left the blind side of Intel.

Just wanted to point out about Core2Duo. That was the era when Intel was still king, its competitor was only AMD, it had the best engineers on the planet, they had the volume, revenue and profits to make those investments, both in IPC and leading node, and successfully turn around, not to mention AMD's misstep after Athon 64.

The guy that knew they were going to hit a power wall with Netburst? He was properly the best guy to lead Intel, somehow now CEO of VMware. Best talents of engineers? Many are now in Qualcomm and Apple. Its competitors? AMD rely on their CPU and GPU to compete against Intel, giving price pressure in all x86 segments, Samsung, Qualcomm and Apple both have their near monopoly profits generator to compete with Intel, and are much better equipped. TSMC has the volume and profits that now exceed Intel.

I am not sure if Intel 10nm will reach HVM this year or not, what I do know is nothing on the Roadmap and leaked info suggest we have any decent 10mn chip coming this year.
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Francois Piednoel is wrong about Conroe. It wasn't a secret project that came out of nowhere; it was an evolution of the Pentium M, which was already a superior design to Netburst, but had been previously relegated to laptops for internal political reasons. Back in May 2005, over a full year before Conroe's release, Tom's Hardware ran an article about how the Pentium M could compete with AMD's best chips, including a concluding paragraph about how Intel's upcoming Conroe design would be based on it. This flatly contradicts Piednoel's claim that the press (and even Intel management!) had no idea about Conroe until just before it came out.

Yes. Core architecture is the successor to the Pentium M architecture but it still derived a lot of its low power DNA from Pentium M.

https://en.wikipedia.org/wiki/Pentium_M

Successor to Pentium M is the Core architecture.
 

Eug

Lifer
Mar 11, 2000
23,583
996
126
As just a simple end user: Like I've mentioned before, I'm just glad I bought Kaby Lake Y (with its media improvements in the iGPU) instead of trying to wait for Cannon Lake Y.

However, my original plan several years ago was to wait until 2016 to purchase a 10 nm laptop.

intelroadmap.jpg


f8d31bac-8796-433b-90c1-909a04be350a.jpg
 
  • Like
Reactions: lightmanek

Hi-Fi Man

Senior member
Oct 19, 2013
601
120
106
This was bound to happen, with all the money being injected into TSMC's and Samsung's fabs from Apple and others.

Intel should have opened their fabs to Apple, Qualcomm, and others ages ago. I honestly think if they had done so they wouldn't be in this situation right now because let's be honest, the writing has been on the wall for about a decade now. Intel's arrogance has cost them dearly but we'll see what happens.

Does anybody have any data on how TSMC's 10nm and 7nm compare to Intel's 14++nm and 10nm? These marketing terms really make things annoying.
 
  • Like
Reactions: CHADBOGA

scannall

Golden Member
Jan 1, 2012
1,944
1,638
136
This was bound to happen, with all the money being injected into TSMC's and Samsung's fabs from Apple and others.

Intel should have opened their fabs to Apple, Qualcomm, and others ages ago. I honestly think if they had done so they wouldn't be in this situation right now because let's be honest, the writing has been on the wall for about a decade now. Intel's arrogance has cost them dearly but we'll see what happens.

Does anybody have any data on how TSMC's 10nm and 7nm compare to Intel's 14++nm and 10nm? These marketing terms really make things annoying.
The first shot to the foot was turning down Apple for the iPhone SOC. Kinda went downhill from there.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Intel should have opened their fabs to Apple, Qualcomm, and others ages ago. I honestly think if they had done so they wouldn't be in this situation right now because let's be honest, the writing has been on the wall for about a decade now. Intel's arrogance has cost them dearly but we'll see what happens.

Intel back then weren't ideal for Apple and Qualcomm either. Their process focused on the aspect which propped up their processors, like high transistor performance necessary for clock speeds.

Even they themselves admitted their 22nm process was only 30% denser than 28nm from TSMC. It wasn't until 14nm they had this density focus, which is misguided considering the most important product for their fabs come from selling their own silicon. They lacked many things necessary for a foundry.
 
  • Like
Reactions: lightmanek
Mar 10, 2006
11,715
2,012
126
Francois Piednoel is wrong about Conroe. It wasn't a secret project that came out of nowhere; it was an evolution of the Pentium M, which was already a superior design to Netburst, but had been previously relegated to laptops for internal political reasons. Back in May 2005, over a full year before Conroe's release, Tom's Hardware ran an article about how the Pentium M could compete with AMD's best chips, including a concluding paragraph about how Intel's upcoming Conroe design would be based on it. This flatly contradicts Piednoel's claim that the press (and even Intel management!) had no idea about Conroe until just before it came out.

A typical IC design requires significant work across a wide range of teams, including the core architecture definition teams, uncore teams, physical implementation, microcode, etc. It's an absolutely gigantic effort and I find it hard to believe that Conroe was such a well kept secret internally.

Also worth noting that Intel works with major OEMs to let them know what's coming, to allow them to prepare system designs around the chip and to let them know how the thing will perform to line up their marketing messages with what Intel will be delivering with the silicon.
 
Last edited:

dark zero

Platinum Member
Jun 2, 2015
2,655
138
106
Seems that Intel is in REAL problems and their fans are on negation.

Even more, even at 10 nm, once the competition (GloFo is targeting it along Samsung) goes EUV, the game is over, since EUV for them is literally going neck to neck to the next Intel tech, and they won't catch up until they are on EUV too.

But that means something else too: Once we hit EUV, the silicon Barrier must be bypassed by other materials.

I mean, silicon based processors are over beyond that. EUV might work for 5 or even 3 nm, but after that we need to see other materials like Gallium in order to continue to picometers.

And even if Intel works with major OEMs, they tends to fail too, remember ITANIUM?

Meanwhile Lenovo took measures and they are going to VIA-ZhaoXin and puting that chip on their laptops. Including the octocore ones.
 
  • Like
Reactions: Henk Poley

ksec

Senior member
Mar 5, 2010
420
117
116
This was bound to happen, with all the money being injected into TSMC's and Samsung's fabs from Apple and others.

Intel should have opened their fabs to Apple, Qualcomm, and others ages ago. I honestly think if they had done so they wouldn't be in this situation right now because let's be honest, the writing has been on the wall for about a decade now. Intel's arrogance has cost them dearly but we'll see what happens.

Does anybody have any data on how TSMC's 10nm and 7nm compare to Intel's 14++nm and 10nm? These marketing terms really make things annoying.


Intel used to get the profits so they can prepaid all the investment in fabs and equipment for next leading node, without borrowing much, a lower cost of money. Nowadays Apple could help fund the development in partnership with TSMC for guarantee Fab Capacity.

The original pieces I wanted to write, if there are anything could sway Apple away from TSMC, to Samsung or Intel. And I could not think of one. The IP Apple has invested into its SoC, are well worth a few billion at least. Even if Samsung or Intel could Fab for Apple at slightly better node then TSMC, while giving a billion dollar discount then TSMC is not worth the risk of their IP.

Just think of TSMC 7nm is roughly the same as Intel 10nm. Of course there will also be pros and cons. Comparing density without leakage or performance is useless. Then there is the cost of design and optimization for those node. It is hard to compare. But just say the are roughly the same. At least TSMC has both 7nm LP and HP available in volume by the end of 2018, still waiting for such certainty from Intel.

Intel's best day were when paranoid survives. Now it is nothing more than blind Silicon Valley optimism.
 

Henk Poley

Junior Member
Apr 28, 2011
3
0
66
I had wanted to post a thread here right before WWDC, just to give it one month time before confirming.

7nm from TSMC is basically clock work because of iPhone. Following the pattern iPad Pro A11x, will be used as 7nm pipe cleaner. WWDC is in a little more then a month time, and yet Intel still hasn't shipped a single 10nm product.
To be fair, WWDC is not really the Apple event where very new hardware is released. But yeah, Apple has a pretty good idea of the hardware they are able to deliver in September by the time of WWDC. We will just not get to know by then.

MacRumors has some experience in following Apple and made a prediction on the sort of things they will release at WWDC: https://www.macrumors.com/roundup/wwdc/

btw, a nice overview of the different actual sizes of the different company's lithography processes ought to be nice in the discussion.