TSMC 16nm in production, 10nm still on schedule (2017). Intel IDF Shenzhen

[CHADBOGA]

If they wanted to obscure their mobile losses, why didn't they do it in 2012 or 2013, or even earlier in 2014 since there's been a TON of articles (e.g. Intel’s mobile division has lost an astonishing $2 billion so far this year, June) this year condemning their losses, but now suddenly in 2015? That does not make sense.

They may not have tried to obscure their mobile losses earlier because at the time they didn't expect them to be so big, but now the penny has dropped that things are going to be be harder and slower to sort out in mobile for Intel.


In recent years it strikes me that Intel has failed at various points to fully grasp the mobile situation.

One could call their not seeing the potential in the iPhone as a mistake that was easy enough to make, but then after seeing the iPhone & mobile explode, they produce BayTrail not for mobile, but low end desktop & maybe high end tablets? D:

Intel has undoubtedly had the wrong people in their mobile division for a while, perhaps they now have the right people, but things always take a few years to sort out.

When will we see from Intel the first truly mobile designed product that doesn't have its roots in BayTrail?

Will it be 5 years after BayTrail was first released?

 

[kimmel]

They may not have tried to obscure their mobile losses earlier because at the time they didn't expect them to be so big, but now the penny has dropped that things are going to be be harder and slower to sort out in mobile for Intel.

You really think that Intel didn't know how much money it would cost them to hit 40m tablets? They may have missed the boat when it comes to mobile products but at least they can do simple multiplication.

 

[CHADBOGA]

You really think that Intel didn't know how much money it would cost them to hit 40m tablets? They may have missed the boat when it comes to mobile products but at least they can do simple multiplication.

I was largely referring to the years before contra-revenue.

 

[krumme]

You really think that Intel didn't know how much money it would cost them to hit 40m tablets? They may have missed the boat when it comes to mobile products but at least they can do simple multiplication.

Its not like Intel have not tried to go into mobile before bt. Ct and ct+ are serious atempts and the ct+ is in eg samsung tab 3 10.
Notice the "so much". How should Intel topmanagement that set rules for communication and accounting - something that is planned at least 2 year in advance - know in 2013 that 2015 would give a loss in excess of 3b? For a revolutionary bt core so much better than ct+?

Hey even if they knew. Who cares. Saying tablets is desktops is just crap. The sweettalk is unbearable imo.

Intel is in good rights to hide it. There is eg good reasons to hide it for the new bigger competitors. But this tablets is phones is desktop is talking to us like we were stupid imo and the numbers excactly prove it big time!!! The numbers is the most obvious fact that the merging is nonsense.

On the techincal level there is a bit/fair level of truth to it but on the solution level its products from different planets. Dont tell me Intel doesnt know that. They prove each day on dcg side they know what this business is about and cpu is only a part of it. If you even want to compare those 650mm2 dies with 50mm2 ct dies.

 

[shady28]

...
Obscuring its lack of progress :biggrin:. Brian Krzanich was really proud when he announced the 46M tablets in January, for your information. They were eager to tell you how much they had reduced the BOM delta. They were eager to tell you that they plan to reduce the mobile losses by $800M in 2015 (so now you know it should be $3.2B in 2015, and Stacy will probably review how their plan worked out at the IM'15, so the merger is utterly irrelevant). Contrary to Qualcomm or MediaTek, Eul showed their full roadmap for 2015 and (less complete for) 2016.

And finally, do you think that Intel gives one [something] about this $7B when they know (1) that the unique investment is closer to $3.5B (Cf. Infineon acquisition, share buyback, dividends) and (2) that they will get profitable eventually in the long-term? The only thing that might worry them a slight bit is ignorant people thinking they are wasting their cash. So your comment and complaints about their losses is short-sighted.
..

The problem with this is that Intels operating losses in mobile (phones / tablets) have increased each year. Moreover, their revenue (what was sold) has been decreasing :

Mobile Group Revenue (Q4 2014/Q3 2014/Q3 2013) -$6M -700% -102%

I mean, how do you have -6M in revenue? If you sold nothing your revenue would be 0, right?

Intel's vision failed years ago when they sold off StrongARM and intentionally crippled Atom, but this is excusable as no-one really saw what was coming.

Then they missed it again by not noticing the iPhone and Android until it was too late, and having no response outside of intentionally crippled Atom.

5 years later they still don't really have a competitor outside of the tablet space and in case you haven't noticed - tablets have flat lined as people are using large screen phones instead.

And still, they focus on Windows / Intel tablets. Windows tablets as far as hardware for price are a terrific bargain no question. The problem is the software / Windows market. I know several people that have windows tablets that sit unused. People buy them because they are cheap and look good on a spec sheet comparison, and then find the software they want isn't there or is poorly maintained.

If intel and MS want this platform to succeed, they need to take a few billion of that "contra-revenue" stuff and start a massive software kick-start for the Windows market. Without that, this platform isn't going anywhere.

 

[witeken]

I mean, how do you have -6M in revenue? If you sold nothing your revenue would be 0, right?

The contra-revenue is counted as negative revenue.

Intel's vision failed years ago when they sold off StrongARM and intentionally crippled Atom, but this is excusable as no-one really saw what was coming.

Then they missed it again by not noticing the iPhone and Android until it was too late, and having no response outside of intentionally crippled Atom.

5 years later they still don't really have a competitor outside of the tablet space and in case you haven't noticed - tablets have flat lined as people are using large screen phones instead.

I don't think you can blame BK for that.

And still, they focus on Windows / Intel tablets. Windows tablets as far as hardware for price are a terrific bargain no question. The problem is the software / Windows market. I know several people that have windows tablets that sit unused. People buy them because they are cheap and look good on a spec sheet comparison, and then find the software they want isn't there or is poorly maintained.

If intel and MS want this platform to succeed, they need to take a few billion of that "contra-revenue" stuff and start a massive software kick-start for the Windows market. Without that, this platform isn't going anywhere.

80-90% of their 46M tablets run Android.

 

[Arachnotronic]

They may not have tried to obscure their mobile losses earlier because at the time they didn't expect them to be so big, but now the penny has dropped that things are going to be be harder and slower to sort out in mobile for Intel.


In recent years it strikes me that Intel has failed at various points to fully grasp the mobile situation.

One could call their not seeing the potential in the iPhone as a mistake that was easy enough to make, but then after seeing the iPhone & mobile explode, they produce BayTrail not for mobile, but low end desktop & maybe high end tablets? D:

Intel has undoubtedly had the wrong people in their mobile division for a while, perhaps they now have the right people, but things always take a few years to sort out.

When will we see from Intel the first truly mobile designed product that doesn't have its roots in BayTrail?

Will it be 5 years after BayTrail was first released?

SoFIA and Broxton. Merrifield/Moorefield were also arguably unrelated to BYT.

 

[ehume]

How about we get back to TSMC's 16nm instead of the interminable argument about contra revenues?

 

[CHADBOGA]

SoFIA and Broxton.

Isn't SoFIA a Baytrail or Cherrytrail core?

Merrifield/Moorefield were also arguably unrelated to BYT.

What makes them unrelated?

 

[witeken]

Isn't SoFIA a Baytrail or Cherrytrail core?

What makes them unrelated?

SoFIA is a separate and cost-optimized platform for the low-end. It doesn't matter that it uses a SVM core since that will be very competitive.

He says they're unrelated because Moorefield is also another platform, the successor of Medfield.

 

[Arachnotronic]

Isn't SoFIA a Baytrail or Cherrytrail core?

SoFIA is the same Silvermont CPU core, but at an SoC/platform level, it's entirely different from Bay Trail.

What makes them unrelated?

Much more competitive ImgTec GPUs, narrower memory interface (aimed at phones), likely built on a process flavor of 22nm more tuned for smartphone-level power consumption.

Moorefield also has a more robust image signal processor, and it integrates a sensor hub.

It also fits nicely into a phone

 

[shady28]

How about we get back to TSMC's 16nm instead of the interminable argument about contra revenues?

Well, in that same release TSMC says they'll have 7nm in 2017 (along with the 10nm in 2016).

None of the foundries have been delivering on these kinds of future promises on time, including Intel who (arguably) has the most capability and resources to make it happen.

Example Reference from May 2012 :

http://www.bit-tech.net/news/hardware/2012/05/14/intel-process-roadmap/1

"Semiconductor giant Intel has revealed its roadmap for process technologies, which will see 10nm, 7nm and 5nm released beginning in 2015."

It's just... FUD.

 

[mikk]

Well, in that same release TSMC says they'll have 7nm in 2017 (along with the 10nm in 2016).

None of the foundries have been delivering on these kinds of future promises on time, including Intel who (arguably) has the most capability and resources to make it happen.

Example Reference from May 2012 :

http://www.bit-tech.net/news/hardware/2012/05/14/intel-process-roadmap/1

"Semiconductor giant Intel has revealed its roadmap for process technologies, which will see 10nm, 7nm and 5nm released beginning in 2015."

It's just... FUD.

Check your, source, it's just wrong. Intel didn't claim anything like that.

 

[scannall]

TSMC and ASML have been actually making progress with EUV.

Milestone

I don't know about 2016 and 2017 though. That sounds more like misplaced enthusiasm to me.

 

[shady28]

Check your, source, it's just wrong. Intel didn't claim anything like that.

Do you have a source that refutes it? I cited one link - all you need to find dozens of others is limit a google search to 1/1/2011 to 1/1/2013.


Here are some more sources :



"Intel Technology Roadmap: First 14nm Chips in 2013, 10nm in 2015"

http://news.softpedia.com/news/Inte...-14nm-chips-in-2013-10nm-in-2015-198612.shtml

"Intel's manufacturing cadence suggests that the first 14 nm products will arrive in late 2013, 10 nm in 2015, 7 nm in 2017, and 5 nm in 2019."

http://www.tomshardware.com/news/intel-cpu-processor-5nm,17578.html

""The 14nm technology is in full development mode now and on track for full production readiness at the end of next year," Bohr said."

http://www.crn.com/news/components-...intel-charts-course-to-10nm-chips-by-2015.htm

 

[witeken]

including Intel who (arguably) has the most capability and resources to make it happen.

Predicting the future isn't easy. Intel ran into unforeseen serious issues with 14nm, probably for a large part because of the lack of EUV. You're just saying this because 14nm had those issues. You have to go back many generations to find another such mistake.

"Semiconductor giant Intel has revealed its roadmap for process technologies, which will see 10nm, 7nm and 5nm released beginning in 2015."
It's just... FUD.

No, your article is FUD.

The image this article is most likely referring to:

No company will ever release 3 process nodes at the same time, obviously.

We don't know when Intel will start 10nm production. But Bohr is optimistic, and the last roadmap we got said 2015.

 

[mikk]

Do you have a source that refutes it? I cited one link - all you need to find dozens of others is limit a google search to 1/1/2011 to 1/1/2013.

I told check your source, it's an easy task. You should find your answer.

 

[krumme]

To be frank i didnt even read the source in first post. And havnt read tsmc process stuff for 2 years because it is ....just ...fud. period. They need credibility or we need some more solid data. Unlike Intel that have delivered more consistently imo.

ss seems to be on a roll now and is the interesting part - we had some signs end 2014 that looked good and now we have a 78mm2 core in s6 that sells in millions and millions. Thats solid fact and imo gives ss bragging about 10nm and euv midnode more credibility than what tsmc says.
And that goes for their own core also. I have a nagging feeling ss have hidden some r&d in other areas because i have a hard time beliewing they got it by steeling/lurking key persons from tsmc (within what seems a fairly limited budget for their broad portfolio).

 

[shady28]

Predicting the future isn't easy. Intel ran into unforeseen serious issues with 14nm, probably for a large part because of the lack of EUV. You're just saying this because 14nm had those issues. You have to go back many generations to find another such mistake.


No, your article is FUD.

The image this article is most likely referring to:

No company will ever release 3 process nodes at the same time, obviously.

We don't know when Intel will start 10nm production. But Bohr is optimistic, and the last roadmap we got said 2015.

The multiple articles I linked to were from 2011/2012, describing Intel's roadmap at that time.

According to those roadmaps, we should see 10nm this year (2015). That is not going to happen.

Promising things you can't deliver falls into the category of FUD. TSMC is in the same boat as Intel on this one.

The only company that has delivered ahead of schedule recently was Samsung (and possibly GloFlo by association).

 

[CHADBOGA]

SoFIA is the same Silvermont CPU core, but at an SoC/platform level, it's entirely different from Bay Trail.



Much more competitive ImgTec GPUs, narrower memory interface (aimed at phones), likely built on a process flavor of 22nm more tuned for smartphone-level power consumption.

Moorefield also has a more robust image signal processor, and it integrates a sensor hub.

It also fits nicely into a phone

So they all clearly come from a processor that was far from ideal to begin with, which means at best they will allow Intel to get a foothold, but not beat ARM at its own game.

I'm waiting for the truly optimised for mobile, chip from Intel, to see if they can stop the attack by the ARMed forces.

 

[witeken]

According to those roadmaps, we should see 10nm this year (2015). That is not going to happen.

The image is talking about production (which you can see with the 2011 for 22nm with 2012 products). Production doesn't mean products. Production can still happen this year.

 

[AtenRa]

SoFIA is the same Silvermont CPU core, but at an SoC/platform level, it's entirely different from Bay Trail.



Much more competitive ImgTec GPUs, narrower memory interface (aimed at phones), likely built on a process flavor of 22nm more tuned for smartphone-level power consumption.

Moorefield also has a more robust image signal processor, and it integrates a sensor hub.

It also fits nicely into a phone

SoFIA is using ARM iGPUs (Mali) and will be build at 28nm TSMC.

 

[Arachnotronic]

SoFIA is using ARM iGPUs (Mali) and will be build at 28nm TSMC.

I was talking about Moorefield.

 

[Idontcare]

Well, in that same release TSMC says they'll have 7nm in 2017 (along with the 10nm in 2016).

None of the foundries have been delivering on these kinds of future promises on time, including Intel who (arguably) has the most capability and resources to make it happen.

Example Reference from May 2012 :

http://www.bit-tech.net/news/hardware/2012/05/14/intel-process-roadmap/1

"Semiconductor giant Intel has revealed its roadmap for process technologies, which will see 10nm, 7nm and 5nm released beginning in 2015."

It's just... FUD.

I don't think you understand what the foundries do.

The foundries don't design the chips that are to be produced on their process nodes.

Just because TSMC says they have 10nm in 2016 doesn't mean you, the consumer, are going to be getting your hands on a 10nm based chip in 2016. It means that TSMC's customers (Qualcomm, etc) can get their hands on 10nm based chips in 2016.

In chip development, it is a solid year of debug/validate/ramp time to bring a chip to the consumer from the time you get first silicon back from the fab.

That isn't TSMC's fault. And it really isn't even the fault of their customers. It is just a problem with consumers who are ignorant of what these timelines even mean.

But real delays can and do happen from time to time. GlobalFoundries had a significant issue with their 32nm process flow which delayed Llano by a good 6 months. That had nothing to do with Llano's design and validation cycle, the fab process tech simply failed to show up on schedule.

 

[shady28]

I don't think you understand what the foundries do.

The foundries don't design the chips that are to be produced on their process nodes.

Just because TSMC says they have 10nm in 2016 doesn't mean you, the consumer, are going to be getting your hands on a 10nm based chip in 2016. It means that TSMC's customers (Qualcomm, etc) can get their hands on 10nm based chips in 2016.

In chip development, it is a solid year of debug/validate/ramp time to bring a chip to the consumer from the time you get first silicon back from the fab.

That isn't TSMC's fault. And it really isn't even the fault of their customers. It is just a problem with consumers who are ignorant of what these timelines even mean.

But real delays can and do happen from time to time. GlobalFoundries had a significant issue with their 32nm process flow which delayed Llano by a good 6 months. That had nothing to do with Llano's design and validation cycle, the fab process tech simply failed to show up on schedule.

That's just a twist on a quote below from Witeken, which I was considering putting in as a funny sig line but decided to drop it until your post :

Regarding Intel's roadmaps : Production doesn't mean products. - Witeken

What would you say to Ford if they said they were going to have an all new redesigned Mustang in production in 2015, and showed off a prototype in 2015 but didn't sell one until 2017?

The simple fact is Intel is 12-24 months behind its old road maps. TSMC did the same thing with its 20nm, albeit only falling 6-9 months behind.

It does not matter one bit why, nor does anyone care why. This industry, just like every other industry, is about products. Pre-announcing said products and being unable to deliver said product is, and should be, a black mark on the record of said company.