Intel team of 1,000 people working on iPhone modem, foundry SoC

[IntelUser2000]

The obvious next step is of course ditching the ARM architecture altogether, maybe at 7nm or 5nm, who knows.

I think this tells me rather that they finally admitted the reality that their mobile architecture is not even remotely competitive with Apple SoCs and they won't get any meaningful traction in the mobile space. Heck, these puny sized ARM SoCs are competitive with Core chips in every aspect!

This seems like a Plan C of which came up because they absolutely had no choice and it wasn't about technical leadership or ego but purely for long term financial stability. How do you expect Apple would go x86 at any time in the near future is beyond me.

Considering they are going 10nm at 2018, I'd say even this rumor is absurd. I don't think they have any process lead anymore.

 

[stingerman]

Indeed, I'd imagine that winning Apple's foundry business would be adequate motivation to give them excellent pricing on the modem side too. The extra wafers help to offset fab costs and, more importantly, represent revenue that's not going to Samsung and TSMC. I've always been of the opinion that Intel's long term play is to maneuver themselves into being the only company that can continue making the necessary investments for future node shrinks. And depriving Samsung and TSMC of leading edge customers is the best way to do that.

Actually, it's looking like TSMC is moving at a faster clip than Intel. Consider this, Intel had the transistor (FinFet) and process lead (14/16NM) for a couple years. Now, TSMC is gearing for mass 10NM production on their 3rd generation FinFet next year. So they caught up to Intel in a short period after their 20NM Planar node.

One more piece of Interesting info I came across is that TSMC high performance mobile FinFet version 1 was able to sustain 2.3GHz with no fan cooling on the turnkey ARM 64 bit design (not Apple's.) FinFet+ (second generation) is 10% faster. With a similar pipeline, that indicates a current Max of 2.53Ghz at FinFet+ (which I believe they are currently using.) (And, that's with no other improvements since they published the initial number in 2014.)

IIRC, 10NM adds a 40% speed improvement, so can I extrapolate the A10 could Max out at 3.542 Ghz? Sure it's a rough estimate, but it's telling in the direction the A10 is going. And that's next year!

 

[Arachnotronic]

Actually, it's looking like TSMC is moving at a faster clip than Intel. Consider this, Intel had the transistor (FinFet) and process lead (14/16NM) for a couple years. Now, TSMC is gearing for mass 10NM production on their 3rd generation FinFet next year. So they caught up to Intel in a short period after their 20NM Planar node.

One more piece of Interesting info I came across is that TSMC high performance mobile FinFet version 1 was able to sustain 2.3GHz with no fan cooling on the turnkey ARM 64 bit design (not Apple's.) FinFet+ (second generation) is 10% faster. With a similar pipeline, that indicates a current Max of 2.53Ghz at FinFet+ (which I believe they are currently using.) (And, that's with no other improvements since they published the initial number in 2014.)

IIRC, 10NM adds a 40% speed improvement, so can I extrapolate the A10 could Max out at 3.542 Ghz? Sure it's a rough estimate, but it's telling in the direction the A10 is going. And that's next year!

10nm adds a 19% speed improvement, I believe, not 40%.

 

[stingerman]

10nm adds a 19% speed improvement, I believe, not 40%.

Okay, I found the reference:

TSMC’s 10nm process technology will have 110 per cent higher logic density compared to the company’s 16nm FinFET+ (CLN16FF+) process tech, 20 per cent higher clock-rate potential at the same power and 40 per cent lower power consumption at the same frequency.

http://www.kitguru.net/components/g...igher-density-and-performance-at-lower-power/

So nothing else changing since their early 2014 test, we are looking at max of 3.036 GHz at the same Voltage, however, using 40% less power.

From the timetables I'm reading, that looks more like an A11 Chip for the iPhone 7s, not the A10 for the iPhone 7.

 

[Mondozei]

Or it could be a beachhead for Apple to have its ARM-based chips manufactured by Intel. This is more likely than Apple getting rid of ARM.

Bingo, we have a winner!

The notion that Apple would abandon it's hypersuccessul, multi-year SoC effort in favor of Intel is something only witeken, who has an Intel logotype as his avatar, could conjure.

Apple's still going to replace Intel in their laptops when they can make the jump software-wise. And getting Intel to fab for them also makes a lot of sense, since Intel has typically had a superior fab process.

Both make sense and neither cancel each other out.

 

[Arachnotronic]

Apple's still going to replace Intel in their laptops when they can make the jump software-wise. And getting Intel to fab for them also makes a lot of sense, since Intel has typically had a superior fab process.

Cool that you're so confident about this. May I ask what source you have for this claim?

 

[Mondozei]

You can't spit on SemiWiki claiming they have a bias and at the same time use Intel marketing slide to prove your point :biggrin:

You should provide source for that data (which looks correct).

Yes, Shintai, the first graph is total BS. Not only are you blindly relying on Intel marketing slides, even they concede it's a "forecast". A "forecast" from a competitor literally means nothing.

This doesn't mean that TSMC has a superior tech, but you have to have more to stand on than this.

 

[Khato]

Actually, it's looking like TSMC is moving at a faster clip than Intel. Consider this, Intel had the transistor (FinFet) and process lead (14/16NM) for a couple years. Now, TSMC is gearing for mass 10NM production on their 3rd generation FinFet next year. So they caught up to Intel in a short period after their 20NM Planar node.

I always do enjoy the definitions we see for 'caught up' with respect to process node technology. Especially when it comes to talking about TSMC's 10nm '3rd generation FinFet' supposedly making an appearance next year when thus far we've only seen what, a single product using their '1st generation' 16nm FinFet? (Not even enough for them to provide 16nm FinFet revenue percentage by itself no less, instead combining it with 20nm.)

There's no question that both Samsung and TSMC have been making some great progress, but until they're actually selling wafers of a process equivalent to or superior to Intel then they haven't 'caught up'. (Especially considering that volume production for TSMC's 10nm is supposed to start when again? 2017 no? Or at least their most recent earnings transcript implies that they aren't expecting 10nm revenue in Q4 2016.)

So nothing else changing since their early 2014 test, we are looking at max of 3.036 GHz at the same Voltage, however, using 40% less power.

If only it worked that way. I mean, how fast would our desktop processors be if it did?

 

[dealcorn]

If Intel's modems work well and are priced attractively, Intel will get a piece of this business. Speculation on a potential foundry deal is more interesting.

Apple does not like competition and it is not stupid. Moving it's smart phone foundry business to Intel eventually shuts down competitive advanced process nodes. This shutdown happens faster if 3D X Point use in advanced graphics applications reduces demand for non Intel advanced process nodes. If in several years the efficiency and cost benefits of 3D X Point make it tout-able for high end smart phones, Apple must be there.

Typically, a large customer gets sweeteners to close a deal. Apple understands that it must compete with whoever using X86 on high end mobile devices. However, as a sweeter for the Intel deal, I expect exclusivity provisions that effectively preclude ARM Android phones from accessing the good stuff for several years. A world where Apple owns the ARM mobile high end and whoever owns the X86 mobile high end is not perfect, but it is helpful for Apple if the Android riff raff is precluded from the high end because they can not access the good stuff on advanced process nodes and technologies.

What ever happened to the digital radio breakthrough Intel showcased several years ago? Might this be a factor in Apple's deliberations? That Intel is kinda slow in integrating modems into it's SoC's makes more sense if they have 1,000 folks working on the Apple deal. As Aicha Evans noted, Intel had limited resources and must pick it's shots. 2016 may be an interesting year.

 

[Guest1]

Actually, it's looking like TSMC is moving at a faster clip than Intel. Consider this, Intel had the transistor (FinFet) and process lead (14/16NM) for a couple years. Now, TSMC is gearing for mass 10NM production on their 3rd generation FinFet next year. So they caught up to Intel in a short period after their 20NM Planar node.

One more piece of Interesting info I came across is that TSMC high performance mobile FinFet version 1 was able to sustain 2.3GHz with no fan cooling on the turnkey ARM 64 bit design (not Apple's.) FinFet+ (second generation) is 10% faster. With a similar pipeline, that indicates a current Max of 2.53Ghz at FinFet+ (which I believe they are currently using.) (And, that's with no other improvements since they published the initial number in 2014.)

IIRC, 10NM adds a 40% speed improvement, so can I extrapolate the A10 could Max out at 3.542 Ghz? Sure it's a rough estimate, but it's telling in the direction the A10 is going. And that's next year!

Caught up? That's laughable they haven't even launched the node yet and you are claiming they caught up not to mention Intel is not standing still either.

http://i.imgur.com/bDlN7Lm.png

The link above is a slide from ASML you should Google them. When TSM launch 10nm it will finally be more equivalent to Intel 14nm. Which by then Intel will be two years in and on the verge of true 10nm.

 

[Guest1]

Bingo, we have a winner!

The notion that Apple would abandon it's hypersuccessul, multi-year SoC effort in favor of Intel is something only witeken, who has an Intel logotype as his avatar, could conjure.

Apple's still going to replace Intel in their laptops when they can make the jump software-wise. And getting Intel to fab for them also makes a lot of sense, since Intel has typically had a superior fab process.

Both make sense and neither cancel each other out.

If apple wants to be competitive with Windows 10 is doing x86 is the way forward not ARM. Windows 10 is a much better solution unifying everything from mobile all the way to desktop. Right now the apple ecosystem is fragmented.

 

[tempestglen]

Hope intel made A10 beat core M badly! LOL

 

[StrangerGuy]

If apple wants to be competitive with Windows 10 is doing x86 is the way forward not ARM. Windows 10 is a much better solution unifying everything from mobile all the way to desktop. Right now the apple ecosystem is fragmented.

Ahaha, Apple is so fragmented that they dominate profits in whatever they compete in and x86 Win 10 is so successful that they need to shove down our throats for free.

So, how many more Intel shill accounts this forum needs?

 

[ShintaiDK]

Yes, Shintai, the first graph is total BS. Not only are you blindly relying on Intel marketing slides, even they concede it's a "forecast". A "forecast" from a competitor literally means nothing.

This doesn't mean that TSMC has a superior tech, but you have to have more to stand on than this.

That's why I used the second one as well. I know the first was a marketing slide. But so is the Semiwiki posts. They are equally credible or lack of same.

 

[ShintaiDK]

Caught up? That's laughable they haven't even launched the node yet and you are claiming they caught up not to mention Intel is not standing still either.

http://i.imgur.com/bDlN7Lm.png

The link above is a slide from ASML you should Google them. When TSM launch 10nm it will finally be more equivalent to Intel 14nm. Which by then Intel will be two years in and on the verge of true 10nm.

Yep. They sure dont lack "creativity" in the foundries PR departments. But when the main tool supplier even says its BS you know its bad.

 

[Thala]

Yep. They sure dont lack "creativity" in the foundries PR departments. But when the main tool supplier even says its BS you know its bad.

They just talk about half-pitch not about any electrical properties. The only conclusion we can draw from this, that Intels process is more dense, but not necessarily better.

 

[scannall]

The obvious next step is of course ditching the ARM architecture altogether, maybe at 7nm or 5nm, who knows.

This makes no sense at all. Apple marches to their own tune, and locking themselves in to a sole source for arguably the most important part of their mobile devices isn't that tune. Using ARM, they can customize SOC's for their own specific needs. Using Intel, they can ask for things and hope the answer isn't 'no'. Kinda like when they asked Intel to make SOC's for the iPhone when they were developing it.

They'd set up their own fabs before switching to Intel, and that's extremely unlikely as well.

 

[NTMBK]

Apple is the master of playing their suppliers off against each other. Using an Intel modem gives them leverage against Qualcomm, and matches their MO. Moving to x86 locks them into Intel indefinitely, and is contrary to how they tend to work. Using Intel fabs as a second or third source at 10nm makes sense, but only with an Apple design.

 

[roob]

It's a dubious report coming a day after Intel tanked that is now helping Intel shares. I would take it with a heavy helping of salt. I'm sure Apple meets with many vendors but actually getting selected as a key supplier will take a lot more. Not least of which is producing a viable alternative at the right price, volume and quality points.

might want to check your numbers. intel finished the day after earnings up 2% compared to the broader market dropping 1%.

 

[Guest1]

Ahaha, Apple is so fragmented that they dominate profits in whatever they compete in and x86 Win 10 is so successful that they need to shove down our throats for free.

So, how many more Intel shill accounts this forum needs?

You got me. After studying the foundry model and intel I chose to to put money in intel stock. This forced me to spend countless hours researching the various claims of the arm camp and Intel camp. In the end Intel method is superior. All of the arm shills regurgitate FUD that has been proven to be false in university research papers such as arm is more power efficient than x86 when it is not, or that 14nm foundry is the same as Intel 14nm etc. I got in Intel stock at $23ish so no complaints even with the recent pull back. No doubt Apple makes huge profits but it doesn't mean they are not fragmented. Look at all of the products still receiving updates with various features missing depending on how old the hardware. Not to mention iOS and OS X are completely different. Windows still dominates enterprise and with the amount off money that has been spent for decades enterprise will not be switching their backbone anytime soon.

If Intel manages to get baseband and AP business it will be the end of TSM and maybe Samsung at the bleeding edge. Even at lower margins Intel would be starving the foundries of funds required for capex and r&d. It truly is a matter of time before it is over.

Look at what Intel was able to do with 22nm moorefield on the zenfone2. Great performance good battery life in a $300 phone. I have waited for morganfield on 14nm as that will be the most advanced SoC in terms of process node on the market. When SoFIA x3 launches it will be 14nm integrated SoC which means the baseband will be on bleeding edge process more advanced even than the Qualcomm baseband on 20nm foundry. People give Intel grief for "missing" but they were either smart to see the commoditization or maybe got lucky as the high end SoC market is not growing fast as the and low end where coincidentally their SoFIA x3 is launching not to mention every single mobile device connects to the cloud which is dominated by Intel Xeon powers data centers.

 

[Guest1]

They just talk about half-pitch not about any electrical properties. The only conclusion we can draw from this, that Intels process is more dense, but not necessarily better.

So funny when the new TSM or Sammy nodes launch it is about how dense it is and therefore superior but when an Intel process is dense than it doesn't necessarily mean they are better even when said density is coming from the tool maker that supplies all of the fabs.

 

[Guest1]

This makes no sense at all. Apple marches to their own tune, and locking themselves in to a sole source for arguably the most important part of their mobile devices isn't that tune. Using ARM, they can customize SOC's for their own specific needs. Using Intel, they can ask for things and hope the answer isn't 'no'. Kinda like when they asked Intel to make SOC's for the iPhone when they were developing it.

They'd set up their own fabs before switching to Intel, and that's extremely unlikely as well.

Very true. If they pick Intel for baseband and then foundry at 10nm there is no going back to TSM and Samsung at which point Intel will surely pitch x86 mobile and the benefits of a unified code for the entire Apple ecosystem. I know unified software sounds terrible to Apple fans but it does make writing software and maintenance a lot simpler. Imagine being able to run your apps on the Desktop or vice versa terrible I know.

 

[stingerman]

thus far we've only seen what, a single product using their '1st generation' 16nm FinFet

Have you confirmed that it's FinFet and not FinFet+ somewhere? All indicators are that they are on TSMC's 2nd generation.

 

[stingerman]

Quote:
Originally Posted by stingerman
So nothing else changing since their early 2014 test, we are looking at max of 3.036 GHz at the same Voltage, however, using 40% less power.
If only it worked that way. I mean, how fast would our desktop processors be if it did?

I'm looking at top level performance at a mobile level voltage. This is not speaking to what happens when you up the voltage, if the material can handle it, and put a better cooling solution on it.

 

[stingerman]

If apple wants to be competitive with Windows 10 is doing x86 is the way forward not ARM. Windows 10 is a much better solution unifying everything from mobile all the way to desktop. Right now the apple ecosystem is fragmented.

You're trolling dude. A unified user interface hasn't worked yet (Windows 8) and Microsoft has pivoted to Apple's approach. Windows based PC sales are on the decline and Windows 10 has done nothing to help. It's a good gaming platform, but there is still little to move businesses off of Windows 7, especially after the Windows 8 fiasco.