Intel interested in building Apple's mobile A4, A5 chips: report - Page 2

We've been hearing a lot about Global Foundries for many years now. Would love to see some objective data on how much and what exactly they produce compared to Intel, TSMC, etc.

i see that we differ on our definitions of the word, "duh".

please, let it be true!

I think it would be a good move on both parts. The A5 could be ported to 32nm, the A6 have access to 22nm. Intel is still king when it comes to process nodes and the mobile business could help drive the advanced processes. Apple's customized designs would benefit from the chip size reduction, power consumption reduction, etc...

The A5 and IOS are completely linked architectures which is one of the big reasons they perform so well together so Apple has built in IP protection. I for one hope Apple and Intel find a way to win together.

Global foundry was spun off from AMD. Everything with AMD label on it came from Global foundry.

Samsung is a major partner in this. Lol Samsung is everywhere

Seems to me that the more diversified Apple's supplier base is, the more insulated they are from unexpected interruptions. Let's see Foxconn in Brazil, and Intel in Cali, etc. Spread the wealth and spread the RISC. Sorry.

And you know this because...?

Oh, I hope not. Global Foundries isn't the highest quality company out there. One of the biggest reason why AMD's chips over the years have been late, or perform poorly has been because of the poor state of their foundry. I've no reason to believe it's any better now that they've sold off so much of it.

By the way, as you can tell from my post, that Global Foundries doesn't have anything to do with ARM. It's an AMD unit, or partly so today.

Yes, they bought Intrinisity right after the A4 came out.

Intel has a factory in China, and other parts of the world. Most of their fabs are here in the USA.

Global Foundries is still mostly a fab for AMD, producing CPU's for them.

TSMC produces the GPU's for both ATI and Nvidia.

I'm not thrilled with them as a manufacturer, as they always have problem with their new processes. Both ATI and Nvidia have been forced to tone down their highest end designs, and to put back the introduction of a number of new products, sometimes form months, because TSMC was having problems. The latest ATI chips were supposed to be on a new process from TSMC, except that they couldn't get yields up high enough. ATI was forced to rework their chips to function on the older process. Nvidia has seen the same problems.

I would also get nervous at the prospect of them producing 150 to 200 million chips a year for Apple, as they will need by the end of this year. I have no problems in expecting that Intel would be able to do it, and on a process that's at least one generation ahead. We've seen how that's worked for them as compared to AMD. Even in a few cases with poorer designs, performance was noticeably better, and as Intel's designs moved ahead, AMD fell into the wayside again.

I'm not denying that at all. Intel has a fab in China and other countries in the world, but most of its fabs are in the USA. I don't doubt that.

All I'm saying is that don't think that Intel keeps its fabs in the USA out of some altruistic concern for American jobs. Intel wants to protect its IP, and it's easier to keep an eye on fabs if they're in the US.

I agree with melgross and I have done tape-in's at Global. TSMC's track record at the smaller feature size isn't all that great either.

Intel is a node ahead for sure and the time between process nodes is stretching out. I say it would be a good move for Apple and Intel if the deal is right.

But if they have a fab in China, and I believe they're building another one there, how does having fabs here protect their IP? We would have to have some idea of what it is they produce in these different fabs. And a number are in countries where foreign IP has little protection.

Intel consistently has problems with their fabs. You're not reading much about it due to the overall total fab capacity they have to use to compensate until they fix their issues. No fab provider is devoid of issues.

The issue isn't fab capacity, as Apple can build several fabs if it wants to see it through. The issue is actually ARM and Apple isn't abandoning that advantage for all the tea in China.

You have a reliable source for that? Everything I've read indicates the A4 in the iPhone is very similar to the Hummingbird core in Samsungs products. The cortex-A9 cores in the A5 may change this trend, but i don't believe the nextgen dualcore cortex-A9-like Samsung chips are out yet...

Processors using older technologies are produced in fabs outside the country, such as China, since that IP is not very important. Processors based on newer technologies are fabricated in US fabs. For example, apparently, the Core 2 Duo processors are manufactured in China or other countries where it is cheaper to produce them.

Intel's fabs are recognized as the best in the industry. This isn't a secret. Every company has problems. But they have fewer. They're also instituting a new technology using "3D" transistors in their 22nm process bode. Not only are they a year ahead in moving to 22nm, at least, but they're a year to three years ahead in this technology as well, which has great benefits.I read industry publications. EE Times is one of the vest for this as well as other industry related subjects. I've quoted them often here, and recommended that people read it.

I don't know why you mentioned that Apple won't abandon ARM. I never said they would. But, if the Atom gets the boost it might from this new tech, Apple might see a use for it as well.

The mods Apple made in the A4 were actually made by Intricisity, before Apple bought them. The changes relate to efficiency, which is what Intricisity was known for.

http://www.eetimes.com/design/signal...s-A4-processor

But that was the first "Apple"design. The A5 is acknowledged to be far more modified from the basic ARM design. For example, the die is much larger than any other ARM die. That's an indicator that the chip has extra units inside. As the design is proprietary, no o e actually knows exactly what's in there, but there are some good guesses. It's acknowledged that the A5 is much more an Apple design.

http://www.eetimes.com/electronics-n...y?pageNumber=0

No doubt, newer chips will be coming out by others. But from all the reviews, the iPad2 is so far ahead in performance, that at best, these newer chips, when they do come out late this year, might just come up to the performance of the A5. a couple of articles have stated, that just going by the die size, which is a standard way of getting an idea of chip costs, the A5 cost Apple about $25, whereas the Tegra 2 (the standard chip for current high end Android and RIM tablets), costs about $15—a significant difference at that price level. Other manufacturers may not be willing to spend that much for their chips, and it's also possible that Apple's cost is mitigated by the sheer number they buy, meaning that at the much lower purchasing levels other manufacturers are making, it would cost even more. That would make the price of the Tegra lower at Apple's purchasing levels. That means that the cost and performance differential would be even greater than they seem. Other chip manufacturers may not be able to duplicate the performance this year. And of course, next year, and another new chip from Apple.

It's hard to imagine the advantage Apple could get by going with intel's 22nm process. Especially with Intel's newly announced new technology appearing on 22nm, Apple could get so far ahead of every other ARM user that it would be impossible to overcome, possible for years, possibly forever. A bold statement, but when you get that far ahead, it becomes very difficult for others to catch up because of the chip design experience.

I doubt if the new plant being built there will be used to produce older tech. intel moves new tech to newer plants, while modding the newest of the older plants as upgrades for the new technology. Intel has already had some of their tech stolen by the Chinese, as have other US technology companies. If they really want it, they will get it, even from US based plants.

The only protection that can be gotten these days is the promise by governments that they will regard foreign tech as owned by the companies who developed them. With the Chinese, there are no such realistic promises.

The Dual-Core next gen A9's are already on the Exynos 4210 chip from Samsung on the Samsung Galaxy S II being sold in Europe and Asia right now.

Not going to happen according to this report :


http://www.rcrwireless.com/article/2...PS/110509966/0

Therefore, Apple has no choice but to stick with Samsung or risk the potential IP leaks with TSMC or risk a potential conflict of interest in the future with Intel and its ambition to make its Atom processor the de facto standard in the mobile arena.

Tegra 2 was weak from the start --- it doesn't have NEON and the GPU is not that great. It's only saving grace was that nvidia beat TI in releasing it early --- so it got all the "design wins" at the CES, which were promptly being delayed because of Honeycomb's non-readyness.

Intel did do fab business in the past. They can do so again. The quote is also incorrect in that it's Apple that's licensing the ARM, and adding its own IP. As a fab, intel need not do any of that. All it needs to do is to have the relevant people sign the required non disclosure agreements.

That's true.

Keep in mind that if Intel were to start making ARM SoCs for Apple, that would give them security for the future incase Apple does switch to ARM-based macs. If Apple went ARM in all their products the way it is today, Intel would lose one of their biggest buyers. This seems to be a win-win situation for everyone. Intel is good at making CPUs, they've just failed at ARM in the past for various reasons. I believe that if they let Apple have their way, they could easily get back into the business.

Oh yah, and Intel should totally ditch Atom. With Windows now running on ARM, even netbooks will no longer have any reason to use Atom. It's dead Intel, time to make the switch.

I can't see Apple going to ARM for its other machines. The best I could see is them adding ARM to a machine for total, and easy, iOS compatibility

This should mean some tools compatibility with Samsung. The way I understand it though is that Samsung optimized for low power where as Global was more concerned about performance. Interestingly AMDs new Llano can out perform Intels Sandy Bridge while running at a lower power level.

I think it is fair to say they are AMDs manufacturing wing first. However it wouldn't be proper for Global Foundries to talk about other customers. So you would only find out about those customers if they wanted to share their partnership with Global Foundries publically.

Given all of that last I knew there was plenty of activity at their upstate NY facility.

It is extremely common in the press to underplay issues Intel has while magnifying those that AMD has. I sense you are doing exactly the same thing. Let's look at Sandy Bridge to start, it initially went out the door with a fault SATA controller. Now mistakes happen but the press cut Intel a lot of slack for this screw up, slack I'm certain AMD would never be allowed.


I don't even buy that they have fewer problems, simply by being on the bleeding edge they likely have more. One way to look at this is the number of mask revisions that take place over the course of a designs life. Obviously some of these revisions are for design errors but they also occur to try to improve yields.

Their process tech is only of value if they put designs on it that leverage those advantages. Intel is not at all reluctant to use it's process advantage at the expense of suitable design efforts for it's processors. It is only recently that Intel has become somewhat aggressive in improving it's core designs and frankly their GPUs still suck.

AMD still has the ability to out do Intel through better design. Their Fusion line up is very impressive. For example at the low end the "Zacate" platform handily beats Intels ATOM processor. In fact one can see far mire integration and innovation in Zacate than you will ever find in Intels half hearted ATOM.

Like wise here, but that doesn't mean we should draw the same conclusions. I honestly believe Intels stupid efforts to protect it's high end product lines is why ATOM is so pathetic. It is also likely a factor in Apple saying screw it and going with ARM. When a company like Intel starts throwing in artificial limitations on products in an effort to rig the market one has to think hard about them as a business partner. Contrast this with ARM and their apparent desire to make the best low power processor possible. A processor that ARM puts few restrictions on.

Or intel could drop ATOM altogether.

I'm just not as convinced as others in this thread that Intel would ever be a good manufacturing partner for Apples A series chips. If they couldn't work as a team to manage ATOM well why would they ever put in the effort for Apple? More so with ARM Apple doesn't need a bleeding edge tech to be competitive.

Someplace in the back of my head I kinda wish that Apple will build it's own fab. There simply isn't enough options for low power processes out there. Rationally though I know that this would cost Apple a good portion of its current cash hord. The costs are far greater than the 2-3 billion a physical plant costs. I also realize that such an adventure might never be cost effective. In the back of my mind though it has a certain appeal.

Not all of these tape outs are for production, for example one foundry (Global Foundries?) has produce working 22nm ARM chips in a lab. Also one of the gate array/programmable logic companies has an ARM core on one of their bleeding edge chips.

There are a lot of questions about what is inside, I would not be surprised to find a lot of ARM IP that anybody can implement.

If nobody knows what is in there then how can they acknowledge that it is Apple IP. If an article ever comes out about A5s internals I will be one of the first to read it. One way to settle this is to look at the SDK.

The performance of the A5 has nothing to do with it. There is a tendency to compare iPad to Android machines but Android has some very serums suckage going on. To start Java is just plain stupid in a handheld device so suckage++. Then we see very poor GPU support so suckage++. Beyond tha Linux is in there, now I love Linux but it isn't a low power OS so suckage++. I could go on but the point is the comparison is pointless as Android is an alpha quality platform.

NVidia is likely highly motivated to succeed here. Other company will look to fill different performance levels. I think the issue us so serious for NVidia that success is a requirement to simply keep the company running.

Nothing lasts forever!

Some roads are rockier than others. Also it is my understanding that this so called 3D tech isn't even Intel original. By the way I hate the use of the term 3D here as these are hardly 3D chips.

You can have your opinions on Intel, but you're a lone wolf on that. The professional counter prevailing opinion is overwhelming. Intel's fabs are considered by pretty much all people in the industry to be the best. Say what you will.

GlobalFoundries is so far behind, it's pitiful. They haven't yet begun 32nm production, and won't, until close to the end of the year. Anything they, or AMD has to say is irrelevant. Three years ago, Ruis stated when criticized at being a full year behind Intel on process size, that they had plans to catch up to no more than 6 months behind in less than two years.

Well, now they are a full TWO years behind. There's no evidence that they will ever catch up. TSMC and one other, which name I can't recall at the moment have said that they are working with 22nm in their labs. But TSMC has historically had problems when attempting to operate a new process size. No confidence there. Samsung has stated that they are looking at 32nm for next year. That's a full two generations behind.

Most other fabs are in the same boat. While memory, because of its simplicity in design and manufacturing, is on smaller die size now, but as it turns out, that has little relation to cpu and gpu production, other than for masking techniques.


Of course, no one is denying that. After all these come from ARM's standard licensed designs. All you have to change is 10- to 20%, and you've made a major change.


If you read the whole report, you would have seen that there are a lot of different areas on the chips that don't appear in the standard design. As they make the point, that shows that there is a lot of differences. You don't have to know what they are to know that.

SDK? You think that looking at Apple's software development kit is going to tell you anything about the chip design that's helpful? Good luck with that!


There are a lot of tests that are OS agnostic. In fact most processor tests are cpu and gpu specific. Android has some problems, but there's no doubt that the A5 is a superior chip.


Of course they are! their business will be destroyed by Intel adding a decent gpu to the cpu die, along with the memory controller. With Ivy Bridge, Intel's on die gpu will be good enough for most tasks for most people, and the fact that Nvidia can no longer mate their chips to Intel's was a major blow. By making ARM SoC's for the mobile market, they hope to keep their business alive.

But as of now, their top of the line SoC doesn't cut it, at least, not for a 10" tablet.


Not literally, of course. But for the life of this new tablet technology era, it will.


It doesn't matter that the new tech isn't original with Intel or not, as all these companies present papers at the various microprocessor conventions during the year. What's important, is who is able to implement these technologies first, and there, Intel has a big lead. Some other companies haven't even implemented Intel's current technology.

It doesn't matter whether you agree with the name or not, this is a big move in process technology, and as the gate does stand vertically, something never done before, 3D seems like as good a name as any.