[semiaccurate] Coffee Lake points to issues with Intel’s 10nm process

[CakeMonster]

Aha, I see. Tick-Tock-Tock was inaccurate (to be generous) then.

So given uncertainties about the new 10nm process, it seems that Coffee Lake should be a pretty damn solid buy then, since I might have to wait for more than 1 year after it. Lets just hope they have tweaked 14nm well enough so that they don't have to reduce the clock speed with 6 cores, since I want my 4.5Ghz for gaming...

 

[IntelUser2000]

And what sources does this magazine base this on? Or is it just speculation?

CanardPC is quite reliable.

Also, at some point you have to admit it rather than just brushing it aside. Intel is having problems. Massive. Even as an outsider we can tell. Also, when Apple can make a 5W SoC that's fully integrated, costs 1/5th and performs better per clock than Intel's Core chips you know Apple is much better at execution. Execution, as in putting ideas and thoughts into reality accurately. And its not surprising because all chip talent went there. And its entirely naive to think its just because Apple pays more money. No, people go to where they believe future exist. What's the point of staying at a company that's at its darkest point in years? Where they would throw away billions of dollars at something then to abandon it in a flash?

 

[AtenRa]

Even if Intel will not sell a single Desktop chip they will still make billions per year from Laptops and Server. They will just get lower revenue the coming ears and increase the desktop release time tables from 12 months to every 18 months (Skylake to Kaby to Coffee). They already stopped the tick-tock cadence and they will use each process for more than 2 years, 10nm may be used for up to 4 generations of Desktop CPUs.

 

[IntelUser2000]

Even if Intel will not sell a single Desktop chip they will still make billions per year from Laptops and Server.

The market is slow, but it eventually catches up with reality. Back when AMD was doing well, they had over 50% of the retail market. No Intel magic marketing trick needed. But trust is developed over a long period of time. If AMD kept leadership, they might have been the one having 70%+ marketshare.

And billions are not enough. They need $30 billion for desktop market just to stay even. The worst case can be quite catastrophic for them. ARM on Windows starts taking low end laptops and desktops. Qualcomm's server is competitive with low-mid range Intel server chips. Apple goes all Ax rather than using Icelake. They lose process advantage to Samsung and TSMC. AMD destroys their HEDT line(due to extremely competitive price/perf), and threatens their LGA115x K line.

 

[WhoBeDaPlaya]

...Apple is much better at execution. Execution, as in putting ideas and thoughts into reality accurately. And its not surprising because all chip talent went there.

Having worked with Apple's (in)famous CSM team, I'd vouch for this.
They are nitpicky down to the very last detail - every tool switch/setting, logfile, warning - and that's just from the physical design implementation / verification side.
While they have some cocky bastards on those team(s), most of them do know what they are talking about in great depth.
Everyone gets grilled thoroughly at the MDR and FDR - analog guys, digital guys, verification guys, DFT guys, top-level integrators, etc.

 

[AtenRa]

The market is slow, but it eventually catches up with reality. Back when AMD was doing well, they had over 50% of the retail market. No Intel magic marketing trick needed. But trust is developed over a long period of time. If AMD kept leadership, they might have been the one having 70%+ marketshare.

And billions are not enough. They need $30 billion for desktop market just to stay even. The worst case can be quite catastrophic for them..

They can literally stop selling any desktop CPU tomorrow and still generate billions per quarter from Laptops and Servers.
Server income alone in Q3 2016 was 2.11Billions and they make another 1.5B income or more from Laptop CPU sales.

 

[fleshconsumed]

Yeah, Intel is struggling and it's been a long time coming. That said they're still the Silicon Juggernaut, they have a lot of cushion behind them to keep them running. But hopefully these struggles can give AMD the breathing room it so desperately needs and give Intel a much needed kick in the nuts. I want to see fast and affordable 6-8 code desktops, and I want them soon

 

[mikk]

Tick broadwell 14nn
Tock skywell new arch
Tock kabylake same but slight process tweak
Tuck coffee lake same but 6c - think i7 replacement q1 2018
Tick canon lake 10nm minor arch improvement as in bw or ib 2019
Tock ice lake new arch...2020 at the earliest imo

Looking at the article and the 10nm woes they might skip cannon lake and go straight to ice lake.
Lets see. If zen is a hit history shows Intel can go into overdrive.

This is a flawed overview. Cannonlake is (or was) scheduled for Q3-Q4 2017 in older Roadmaps. You could argue that it will be delayed which might turn out true, but claiming 2019 is bold. Furthermore you are expecting that Cannonlake will effect the launch of Icelake which doesn't make sense either. Cannonlake for the Mainstream client is coming only for 15W and below. Similar to Broadwell where Skylake wasn't affected from Broadwell, only from yield issues of 14nm. Even with Broadwell GT3e that came out for desktop Q2 2015 desktop Skylake came just a quarter later. Icelake is a 2019 product as we know. There is lots of time for Intel to improve on 10nm yields until 2019.

 

[raghu78]

Thanks to folks who know French and summarized the Canard PC article. 2017-2019 is going to be a difficult period for Intel if the Canard PC article is true. Intel seems to have given AMD a golden opportunity to catch up and even overtake them. AMD should not make the mistakes of the past and keep up a rapid pace of innovation and excellent execution.

 

[witeken]

Having worked with Apple's (in)famous CSM team, I'd vouch for this.
They are nitpicky down to the very last detail - every tool switch/setting, logfile, warning - and that's just from the physical design implementation / verification side.
While they have some cocky bastards on those team(s), most of them do know what they are talking about in great depth.
Everyone gets grilled thoroughly at the MDR and FDR - analog guys, digital guys, verification guys, DFT guys, top-level integrators, etc.

Why do you assume that everyone knows all the acronyms you use? It sounds condescending.

 

[Arachnotronic]

They can literally stop selling any desktop CPU tomorrow and still generate billions per quarter from Laptops and Servers.
Server income alone in Q3 2016 was 2.11Billions and they make another 1.5B income or more from Laptop CPU sales.

Haha, no. If they stopped selling desktop CPUs tomorrow their utilization rates would plummet and their margins would plunge because the depreciation costs associated with the fabs would be split across half the number of dies.

That would increase the costs of all remaining products including laptop and server.

Intel's profitability is directly tied to its factory utilization rates.

 

[oak8292]

Just listing the high-profile employees that were removed thanks to Kraznich nevermind the thousands of regular workers.
Renee James(software)
Joel Emer(described as an CPU architect now working for Nvidia)
Rani Borkar(director of development)
Kirk Skaugen(PC division)
Keshavan Tiruvallur(Validation chief)
Doug Davis(IoT)

You left off Kevin Zhang (Vice President of Technology and Manufacturing Group and Director of Circuit Technology at Intel) he is now at TSMC with a similar title.

 

[dark zero]

I actually read the full Intel article from Canard since I speak French. Here's the most important points:

• They say Intel's in the most precarious time ever in its existence.
• Employees blame Krzanich as the main cause, he's impatient and eager and keeps changing/canceling projects before they get anywhere.
• Many employees/engineers got fired in recent times and they say Krzanich treats engineers (replacing them) like supermarket cashiers, many new hires are inexperienced and not getting trained well enough. The mood in the company is at an all-time low and many employees are in fear for their jobs.
• Murphy Renduchintala is said to actually be a glimmer of hope in the whole mess as since he's been brought on he's been able to focus R&D properly and actually has technical background and makes decisions that engineers are able to get behind on. Some say they're hoping he'll replace Krzanich as CEO once they hope he'll get the boot from the board.
• 10nm is suffering all kinds of problems, both technical as well as management problems. The first samples that got in apparently were extremely disappointing in performance, not to talk about yields.
• Kaby-Lake as well as Cannon-Lake were supposedly late designs just to save face and management was just hoping AMD wouldn't compete. Cannon-Lake is said to bring almost no architectural improvements.
• 10nm will sample in late 2017 but production won't happen till 2018.
• Intel has nothing concrete till 2019.
• Long-term, Intel is said to be very worried about x86 as ARM is gaining ever more. If they can't keep Apple happy in the Macbooks that would be the first sign of the decline of x86.
• Because of the above, they say that they have one CPU project which can do hybrid x86/ARM execution and they say they'll have the first prototype wafers with this within the next few months.
• There's an Intel CPU AMD GPU MCM in the works.

They state the sources on all of this were off-the record interviews with both current and ex-employees.

Ok...thanks for the information, it was very useful and now I give my opinion about the points..
- Intel worst time was the Pentium 4/Itanium era.. that time were the WORST ever for them. The Cherry Trail/Broadwell time was the 2nd biggest nightmare.
- Krzanich is showing up that is not a good CEO. Killing a promising mobile uARCH (Broxton) along a very badly executed one (SOFIA) is a very bad idea for them and a sign of bad management. Broxton could made a big advance to big phone companies like ASUS and including some chinese ones like Leeco or even Xiaomi.
- As expected, 10nm is the sign that FinFet is not the tech we wanted and also the performance is not as better as expecting... Now is EUV for it.
- If Intel didn't expected AMD renaisance, they did wrong BIG time. Many had faith on AMD comeback which was unavoidable since they learned from their errors. And AMD did it right after near 10 years.
- The x86-ARM hybrid is already released thanks to VIA and Zhao Xin (Chinese government partly owned company) and they are working to improve it. Come on Intel, you expected something less from them? Chinese became even faster learners than Japanese in tech! Tralalak knows about that project. Also there were rumors that AMD is planning that hybrid mode after Zen and K12.
- It seems that at last AMD and VIA are saving x86 in a different ways. And the first one are ready to compete Qualcomm with K12.

 

[dark zero]

I wonder if the AMD Intel MCM part was AMD's plan B in case Zen was a dud?

AMD's Plan B is ARM K12.

 

[raghu78]

AMD's Plan B is ARM K12.

The way I see it AMD focussed on x86 as they know its going to be their bread and butter. In fact Lisa Su pulled the plug on ARM K12 to focus their meagre resources on x86 Zen. Right now AMD has stated that ARM will be restricted to semi-custom based on customer requirement.

 

[positivedoppler]

Let's not forget some of the other Kraznich initiatives such as the MICA because his wife likes it,their effort to "diversify" their workforce, and my favorite conflict free silicon. Not saying they're bad things, but just smells like desperation from a company slinging random mud on the wall hoping something sticks.

 

[positivedoppler]

The market is slow, but it eventually catches up with reality. Back when AMD was doing well, they had over 50% of the retail market. No Intel magic marketing trick needed. But trust is developed over a long period of time. If AMD kept leadership, they might have been the one having 70%+ marketshare.

And billions are not enough. They need $30 billion for desktop market just to stay even. The worst case can be quite catastrophic for them. ARM on Windows starts taking low end laptops and desktops. Qualcomm's server is competitive with low-mid range Intel server chips. Apple goes all Ax rather than using Icelake. They lose process advantage to Samsung and TSMC. AMD destroys their HEDT line(due to extremely competitive price/perf), and threatens their LGA115x K line.

Agree on all points. This company is under siege on all front and have no answer for years. Their 170 billion market cap is destined to shrink at least by 1/2. I don't like chat up stocks on tech forums but seems like easy money to short Intel at this point.

 

[Arachnotronic]

Agree on all points. This company is under siege on all front and have no answer for years. Their 170 billion market cap is destined to shrink at least by 1/2. I don't like chat up stocks on tech forums but seems like easy money to short Intel at this point.

Are you currently short or planning to short?

We definitely need a tech business/finance sub forum

 

[moonbogg]

So suddenly Intel is crap and AMD is king again, got it.

 

[positivedoppler]

Are you currently short or planning to short?

We definitely need a tech business/finance sub forum

I'm planning on shorting but tbh I don't know enough about selling/buying stocks. I like the idea of a business subforum since it's extremely agitating to see technical threads get dragged into earnings, stock prices, market shares, etc etc...by people who secretly have financial interests in these companies.

 

[WhoBeDaPlaya]

Why do you assume that everyone knows all the acronyms you use? It sounds condescending.

Apologies, did not mean to come across as such.
Just thought I would offer sort of an insider's $0.02 having actually designed chips for Apple

 

[podspi]

6c skl instead of 4c is super. With oc we had more or less same perf for 6 years so this actually makes a difference for once.
At 14nm they can print them cheap so all is nice. Whats not to like.
It means i3 2c 4t is dead and moved to low end pentium life. About time.
Zen plus is well underway and when Intel 10nm hits at desktop we are probably at 7nm zen plus plus at desktop too. About same process perf. So it looks like we are saved after 10 years of starvation and funding Intel shareholders with our high margin 4c. Now comes 10 golden years

Honestly, that would be great. Something like:

Celeron relegated to small core.
Pentium 2C/2T
i3 4C/4T
i5 4C/8T
i7 8C/8T or 6C/12T
i7-HEDT 8C/16T
[gotta get dem margi]

I typed parts of the article into google translate. Its what the reddit guy said, but its still shocking. Kraznich sounds like someone with a mental instability. Almost sounds like someone trying to ruin the company in purpose. Rash decisions, squeezing employees for cost reductions, firing of its veteran(and most valued) workforce, control freak wanting to rule over everything with an iron fist.

I used to read articles about Blackberry(formerly known as RIM). Companies in decline do so not only because of market force, but because they make decisions that accelerate their demise. Intel is doing it now. Of course the market situation is not easy. Regardless, the current management is crazy.

Just listing the high-profile employees that were removed thanks to Kraznich nevermind the thousands of regular workers.
Renee James(software)
Joel Emer(described as an CPU architect now working for Nvidia)
Rani Borkar(director of development)
Kirk Skaugen(PC division)
Keshavan Tiruvallur(Validation chief)
Doug Davis(IoT)

Duties that Justin Rattner was reponsible for was taken over by Kraznich. I've wondered for a while what that guy was doing. I realized since couple of years ago that future-related research drastically dropped at IDF. Thanks Kraznich. Rip out the heart of the company. Ever wondered where their $10 billion annual R&D went to and why certain people were saying R&D to product ratio was crap at Intel? There ya go. Even Mr. Barrett was a star compared to this guy. Actually, Barrett left a great legacy. It was his manufacturing expertise that gave Intel its process leadership. Of course, destruction comes far easier.

Death is coming for this company. I'd give it 10 years. Rather rapid considering how big they are. But if you start to go at its foundation, the whole thing will fall. I think the biggest mistake companies make is they start to focus on the company, rather than its people. So they start slashing R&D and micro-managing people(annoying those involved) to save a few bucks and keep it alive for few more years. Without the people, its worth nothing. You start to treat them like crap, and you get crap in return.

Will a new CEO help? Probably. If he's sane. But based on what CPC is saying, the old-timers with the experience gone, it'll be very difficult to reverse the trend. He'll need to prove the WHOLE thing will change for the better.

I always try to be skeptical when talking about the death of a company. I learned my lesson after Apple, personally. All it takes is the right (or wrong) group of people to come in and succeed (or fail). Intel seems to be in a lull, but it honestly isn't like they are where AMD has been in the past few years, and even they look like they might come back. Let's hope competition kicks everyone in to high gear, I'd love to see competition in the x86 market like we do in the ARM market.

 

[mikk]

You left off Kevin Zhang (Vice President of Technology and Manufacturing Group and Director of Circuit Technology at Intel) he is now at TSMC with a similar title.

Mooly Eden as well.

 

[AtenRa]

The market is slow, but it eventually catches up with reality. Back when AMD was doing well, they had over 50% of the retail market. No Intel magic marketing trick needed. But trust is developed over a long period of time. If AMD kept leadership, they might have been the one having 70%+ marketshare.

And billions are not enough. They need $30 billion for desktop market just to stay even. The worst case can be quite catastrophic for them. ARM on Windows starts taking low end laptops and desktops. Qualcomm's server is competitive with low-mid range Intel server chips. Apple goes all Ax rather than using Icelake. They lose process advantage to Samsung and TSMC. AMD destroys their HEDT line(due to extremely competitive price/perf), and threatens their LGA115x K line.

Haha, no. If they stopped selling desktop CPUs tomorrow their utilization rates would plummet and their margins would plunge because the depreciation costs associated with the fabs would be split across half the number of dies.

That would increase the costs of all remaining products including laptop and server.

Intel's profitability is directly tied to its factory utilization rates.

Nope, they have already started the layoffs and closed fabs, they would simple close one-two more fabs and lay off more workers until they would have the fab capacity they want.

They already gave up on the two year cadence and will use 14nm for more than two Chip generations, each fabrication process will already be used for more than 3 years from now on, much like 28nm.

Desktop doesnt make the fat profit it used to back 10-15 years ago, Mobile and Server is where the fat margins and profits are now, not to mention that they already selling mobile socs to the desktop segment already.

Producing less volume will increase the cost for the consumer but not that much, Intel will make less than what they make today but they will still make billions even without the desktop.

 

[itsmydamnation]

So suddenly Intel is crap and AMD is king again, got it.

No thats not it at all.
But its power house has been its FAB's , 90,65,45,32nm intel was absolutely dominant. But since then the playing field has leveled far more then intel would like to admit.

Then you add on the lack of any real progress after Haswell while ARM and Apple have been on an endless march.
Then you add in the actual market dynamics, 1st world economy is in the toilet because governments cant figure out that if you kill your middle class there is no one here to buy all your buddies in big business's stuff.
Then you add that intel has a lot of mouths to feed and big capital expenditure.
Then you add Zen desktop, Zen server and what a Zen laptop APU looks like.
And maybe people will be interested in the qualcomm server part ( i think this is the least of intels problems at this point)

Its not that intel still wont make lots of money over the short or medium term, its that their revenue, gross profit and thus dividend will suffer and that will put a lot of pressure on the stock prices and onto the exco and upper management. It sounds like they dont have short term answers, so what will really matter is what choices they make under the pressure that is yet to come.