Tweaked Apple TV Contains Die-Shrunk A5 Chip, Not A5X

[osaga]

I think I've "cracked" a small part of the apple TV problem...

The Apple "iTV" will run ios, and we'll have all the functionality of the ipad on our TV (obvious).

It will include an extensive Apple created content subscription service. (obvious, but apparently difficult.)

...drumroll... The wireless remote will be a tracking device with a button on it. You waive it around in the air, and follow a cursor on the screen, you click on whatever you want. It's an Air Mouse! Forget about air gesturing with your hands, a touchscreen remote, wireless keyboard, or a trackpad. It's just a piece of aluminum with a button.

The remote could include a fingerprint sensor, to identify each user.

P.S. You probably wish there was a down vote button huh? Anyone have any better ideas? Or even care to speculate?

 

[huffcw]

I love my Apple TV but it is lacking serious content when compared to other devices like the roku 3.

Completely not true if you have an iPad which can AirPlay content to the Apple TV -- from a good number of apps, such as HBO, Showtime, etc.

And, if you have a newer MacBook that can AirPlay, anything on your computer can now be put up on your TV screen very easily (this includes Hulu, without having to go with a Hulu Plus subscription, many cable channels that offer online content like Comedy Central, and any other video you can find on the web).

Plus it works pretty seamlessly with the huge amount of iTunes content I have on my computer - making browsing, selecting it and playing it very user-friendly.

AirPlay makes the content available through Apple TV way more robust than Roku 3 (assuming you have an iPad or MacBook with the capability). I had a Roku for a while, but found that while there are a good number of apps, most had nothing really that great to offer in terms of content (other than a handful of major ones, like Netflix and HBO). Apple TV for me offers a much greater value - given the endless content I can get to it through AirPlay and from my iTunes library.

However, I will agree that it would be more convenient to have these apps and content directly on the Apple TV rather than having to AirPlay from another device.

 

[APlotdevice]

Not true if you have an iPad which can AirPlay content to it - from apps for HBO, Showtime, etc.

And, if you have a MacBook that can AirPlay, anything on your computer can now be put up on your TV screen very easily (this includes Hulu, without having to go with a Hulu Plus subscription, many cable channels that offer online content like Comedy Central, etc.)

AirPlay makes the content available through Apple TV way way more robust than Roku 3 (assuming you have an iPad or MacBook with the capability).

However, I will agree that it would be more convenient to have these apps and content directly on the Apple TV rather than having to AirPlay from another device.

Last I checked Showtime does not support AirPlay. Nor do a lot of other media apps. Also there are a lot of other types of apps for the Roku beyond basic media. Sure you can AirPlay mirror apps from your device, but it doesn't take full advantage of the HDTV screen and you have to constantly look down to operate it.

 

[Menneisyys2]

Dumb question. Does Apple TV bought from the States will work in Australia? I know HULU, NETFlix will not work. I am just wondering about the movies that I bought from iTunes.

It will (this is how I'm using it with my US account), albeit, officially, isn't really allowed by Apple (see the dedicated threads here at MR - I can link to some of them if you want). The latest ATV3 firmware even supports quick(er) switching between accounts now (which also means local and US accounts can bbe switched much quicker).

----------

Last I checked Showtime does not support AirPlay. Nor do a lot of other media apps. Also there are a lot of other types of apps for the Roku beyond basic media. Sure you can AirPlay mirror apps from your device, but it doesn't take full advantage of the HDTV screen and you have to constantly look down to operate it.

Yup, simple mirroring sucks, as opposed to transferring iOS-native files to the ATV so that it decodes them locally.

In this regard, third-party player set-top boxes are indeed much more advanced: much better streaming support (SMB, UPnP, FTP, WebDAV, HTTP etc.), much better video format / container support (MKV etc.), much better everything. This means they can stream directly from almost any NAS source (no need to rely on iTunes running locally) and can decode any files right on them, delivering orders of magnitude better image quality and less complicated setup / power usage (taking the power drain made by the desktop client AirPlay is mirroring from) needed than the ATV, which would have to use mirroring with otherwise non-native formats.

Sure, there are proxy e.g. SMB or UPnP solutions (see for example THIS tutorial), but they all require a middleman, even if it's "only" an iDevice.

 

[wizard]

It is a jump to believe it is a 28 nm test drive but not without reason. The bigger jump is the idea that this is a TSMC chip. Being an Apple TV and the low associated volumes it could very well be a Samsung chip running off an 28nm pilot line!

Apple may very well be looking at TSMC and ther foundries for that matter, for its 28nm chips, but nothing about this release says that the chip is coming from anywhere other than Samsung. The physical size of the chip is in otherwords not sufficient evidence that the unit is manufactured by TSMC.

In any event yes smaller chips are wonderful for the consumer.

This silent update could have major implications if it is a 28-nanometer die test. I'm looking forward to seeing less power and space taken up by the electronics, giving more battery space that is put under less demand. Or allows for new features.

 

[djdj]

I don't see why it would have contained an A5X chip. The A5X was the same chip as the A5, just with more GPU cores. The AppleTV does not need them.

...unless they were planning on turning the Apple TV into a gaming console, in which case it would need the extra power to do even half-way decent 3D graphics at 1080p.

So it kind of rules that out. At least for the foreseeable future.

 

[syriana]

this is not 28 nm...

At the size as measured, this can't be a 28 nm process...
Very good scaling would be 57% (source: anandtech, http://www.anandtech.com/show/5789/the-ipad-24-review-32nm-a5-tested)
So going from 8.19 mm by 8.68 mm at 32 nm, that would be an aera of 71 mm2 to 6*6 or 36 mm2 at the actual node... This is almost half!
So it's better than a full node in the same foundry... So it can't be half a node (from 32 nm to 28 nm is half a node).
That means either 20 nm (global foundries?) or 22 nm with a a more dense process (intel?)

 

[yeah]

Ditto. Anyone who doesn't appreciate how big the introduction of a 28nm die is just doesn't understand the physics that will bring revolution to mobile computing. Less heat. Less power drain. More space for batteries. All good things for mobile devices.

Bad things to come too. In the next few years we will have to find a new platform to work with, instead of silicon. Yes processors will get smaller and smaller, but there is a certain limit to the number of transistors that go on the processor.

 

[wizard]

Come on now this is very significant, if Apple is using 28nm hardware then they most likely have wheel in motion to introduce 28nm hardware in the next iPads. I'm not at all excited about the iPhone so I look at this in terms of iPads!

From a manufacturing ramp up standpoint it only makes sense to start out on a low volume machine like AppleTV. AppleTV might not be that interesting but the reality that 28nm may be here is. For us users it means any of the following possibilities might happen:

1. Faster clock rates on the CPU or GPU or both.
2. More cores on the GPU or the CPU or both.
3. Possibly a more advanced GPU core.
4. Possibly 64 bit ARM cores.
5. Much larger on chip caches for improved performance.
6. VRAM for the GPU on die, increasing performance and lowering power.
7. Higher integration which again lowers power and increases performance.
8. More room externally for RAM. In other words more board space
9. Less heat given the same architecture.

These are just some of the possibilities. 28 nm would most likely allow Apple to deliver many of the above features without a thermal increase in the iPad.

So while the Apple TV may not be all that interesting on its own this is very good news to hear.

While I most certainly appreciate _All_ the news MacRumors brings us, it's things like this that are so trivial, that the average buyer doesn't care about.

It's a reflection of anything bigger to report on, like something truly substantial to report about regarding the next iPhone.

 

[huffcw]

Last I checked Showtime does not support AirPlay. Nor do a lot of other media apps. Also there are a lot of other types of apps for the Roku beyond basic media. Sure you can AirPlay mirror apps from your device, but it doesn't take full advantage of the HDTV screen and you have to constantly look down to operate it.

True - some media apps do not, but many do. And all work when AirPlaying from a MacBook (which, to be honest, is what I use most of the time).

I still agree it would be nice to have the apps directly on Apple TV, but I still find the current Airplay method provides better options for me than what I can get from Roku. (I am talking streaming media here, not other types of apps.)

And, while it would be nice not to have iTunes open to stream from my computer, the interface for browsing though my huge library of media is better than what I have come across on other devices. I do wish Apple would allow to stream directly from network drive - would be so much better.

 

[wizard]

Apple TV seems to now be a testbed for Apple's Tick (die shrink) cycle of A-series chip manufacturing.

Exactly! Apple TV could be more valuable to them right now in this respect than as an actual product. It allows debug and line issue iron out to be done on a machine with limited volumes. If anything this highlights the coming 28 nm chips in the other iOS devices.

 

[alent1234]

Bad things to come too. In the next few years we will have to find a new platform to work with, instead of silicon. Yes processors will get smaller and smaller, but there is a certain limit to the number of transistors that go on the processor.

In a year or two mobile is going to be where PC's are now. The cheapo model being good enough for most people

 

[sammich]

This chip is a hair over 1/4 the area. Fairly amazing in 2 years.

36/(8.19*8.68) = 0.506

Actually, it's roughly 50% of the area of the 'old' A5.

I don't know how semiconductors work, but with all other things being equal, wouldn't a 50% reduction in die area mean a bigger jump than just 32 to 28nm?

 

[Henriok]

I hate to break it to you but this is probably not a die shrink nor any TSMC stuff.

The package only contains the ARM CPU, not the stacked RAM. That's known since the markings on the package only describes the CPU and not any RAM since there's only writings on two sides, not four (the left and the right is the data for the RAM when available). And the RAM module from Hynix is clearly seen on the mobo.
That would account for the reduced package size. I find it hard to believe that MacRumors actually measured the die size, but just measured the package size.

Secondly.. the markings and the package itself is very similar to the markings and packaging that Samsung is using.
The original A5 was called APL0498 339S0130 (fabbed and packaged by Samsung)
The A5R2 was called APL2498 339S0173 (fabbed and packaged by Samsung)
This new A5R3 is called APL7498 343S0624
It's the second part of this designation that describes what plant and fabrication process is used. It's very similar on the package on the ones we know Samsung has made.
TSMC would probably not mimic Samsung's designations.

 

[wizard]

There is much in the way of fudging when it comes to process size. Some would say that Intel is only halfway there when they say they are at 22 nm for example.

As for Samsung they have often described their 32 nm processes as 32/28 nm. I would suspect that if this is a Samsung chip the process is closer to 22nm that 28 nm. Either way the obvious thing here is a much smaller chip. We really don't know much more beyond that though they could have trimmed some non essential functionality from the chip. Sadly I don't think there is much incentive for somebody to tear this chip down and see how it is made.

At the size as measured, this can't be a 28 nm process...
Very good scaling would be 57% (source: anandtech, http://www.anandtech.com/show/5789/the-ipad-24-review-32nm-a5-tested)
So going from 8.19 mm by 8.68 mm at 32 nm, that would be an aera of 71 mm2 to 6*6 or 36 mm2 at the actual node... This is almost half!
So it's better than a full node in the same foundry... So it can't be half a node (from 32 nm to 28 nm is half a node).
That means either 20 nm (global foundries?) or 22 nm with a a more dense process (intel?)

 

[nunomoreira10]

At the size as measured, this can't be a 28 nm process...
Very good scaling would be 57% (source: anandtech, http://www.anandtech.com/show/5789/the-ipad-24-review-32nm-a5-tested)
So going from 8.19 mm by 8.68 mm at 32 nm, that would be an aera of 71 mm2 to 6*6 or 36 mm2 at the actual node... This is almost half!
So it's better than a full node in the same foundry... So it can't be half a node (from 32 nm to 28 nm is half a node).
That means either 20 nm (global foundries?) or 22 nm with a a more dense process (intel?)

Maybe they made a single A9 core processor in 28nm instead of disabling one on the A5.
It could be cheaper this way.

Even better, they could have put a single core A9 with a single SGX 543MP clocked twice for similar performance and a single 32-bit channel with ddr3 memory for similar memory bandwith.
so basically cutting an A5 in half.

 

[mono1980]

I wonder if they will put this new, smaller chip in the iPhone 4S as well. It would improve its battery life, which would be nice. Or maybe the rumored "cheap" iPhone will use it.

 

[iChrist]

pfffft

Wow, so AppleTV is useful for Apple only as test for new die shrink.

What is AppleTv good for anyway? Netflix and music on remote speakers? It has screen full of "apps", but they are all useless.

Really what has Apple been doing the last 4 years?? iOS... no major improvement (Maps and Siri, half baked), OSX... yearly rehash with some new iOS features, AppleTV... nothing, retina macbooks... engineering issues with fans and screen lag, iMac... move all fat from edge to bulge in back and add more glue, MacPro... uhhhhhhhh...
... this would have been all done in one year in Steve's day.

 

[2bikes]

I think I've "cracked" a small part of the apple TV problem...

The Apple "iTV" will run ios, and we'll have all the functionality of the ipad on our TV (obvious).

It will include an extensive Apple created content subscription service. (obvious, but apparently difficult.)

...drumroll... The wireless remote will be a tracking device with a button on it. You waive it around in the air, and follow a cursor on the screen, you click on whatever you want. It's an Air Mouse! Forget about air gesturing with your hands, a touchscreen remote, wireless keyboard, or a trackpad. It's just a piece of aluminum with a button.

The remote could include a fingerprint sensor, to identify each user.

P.S. You probably wish there was a down vote button huh? Anyone have any better ideas? Or even care to speculate?

Have you ever seen a Nintendo Wii?

 

[hundleton1]

I hate to break it to you but this is probably not a die shrink nor any TSMC stuff.

The package only contains the ARM CPU, not the stacked RAM. That's known since the markings on the package only describes the CPU and not any RAM since there's only writings on two sides, not four (the left and the right is the data for the RAM when available). And the RAM module from Hynix is clearly seen on the mobo.
That would account for the reduced package size. I find it hard to believe that MacRumors actually measured the die size, but just measured the package size.

Secondly.. the markings and the package itself is very similar to the markings and packaging that Samsung is using.
The original A5 was called APL0498 339S0130 (fabbed and packaged by Samsung)
The A5R2 was called APL2498 339S0173 (fabbed and packaged by Samsung)
This new A5R3 is called APL7498 343S0624
It's the second part of this designation that describes what plant and fabrication process is used. It's very similar on the package on the ones we know Samsung has made.
TSMC would probably not mimic Samsung's designations.

the guys got a very good point BUT the older ATV did have the 8gb nand flash near the cpu as well ?

 

[osaga]

Have you ever seen a Nintendo Wii?

Ha, yes. Apple obviously would have to do a much better job. By using a combination of different sensors and software.

Apple has always placed a lot of emphasis on the interface device. Steve talked about it extensively at the first iphone keynote.

Doesn't it seem like an "air mouse" would be crucial? Otherwise a remote app for your iPhone? -You'd have to be looking at your phone instead of the tv. Or a physical keyboard with a track pad on it? -seems too complex for Apple. I guess we won't know what's best without actually trying it, but my money is on air mouse.

 

[APlotdevice]

Have you ever seen a Nintendo Wii?

The remote on the higher end Roku streamers can also behave as an air mouse.

 

[puckhead193]

I've been thinking of getting an apple tv..too many times I rent a movie thinking i'll watch it on my iPad but never do, then i'm forced to watch it on day 28 before it expires on my iPad...
How would one know if i'm getting this new tweaked version?

 

[somethingelsefl]

You are correct. Have you seen the size of the lock on the new chip? Nobody makes a lock pick set that small! Don't even get me started on finding a source of tiny bits to program with.

Comment of the day material. Genius.

 

[repoman27]

While I most certainly appreciate _All_ the news MacRumors brings us, it's things like this that are so trivial, that the average buyer doesn't care about.

It's a reflection of anything bigger to report on, like something truly substantial to report about regarding the next iPhone.

Are you serious? This is real reporting, an actual story broken by MacRumors, and it's great to see that.

At the size as measured, this can't be a 28 nm process...
Very good scaling would be 57% (source: anandtech, http://www.anandtech.com/show/5789/the-ipad-24-review-32nm-a5-tested)
So going from 8.19 mm by 8.68 mm at 32 nm, that would be an aera of 71 mm2 to 6*6 or 36 mm2 at the actual node... This is almost half!
So it's better than a full node in the same foundry... So it can't be half a node (from 32 nm to 28 nm is half a node).
That means either 20 nm (global foundries?) or 22 nm with a a more dense process (intel?)

I was wondering when someone would do the math. It's a near perfect 50% of the 32 nm die, or a full node shrink. This is 22 nm Intel!

Edit: Or, as you mentioned, 20 nm GloFo. Hmmm... That might make even more sense.

I hate to break it to you but this is probably not a die shrink nor any TSMC stuff.

The package only contains the ARM CPU, not the stacked RAM. That's known since the markings on the package only describes the CPU and not any RAM since there's only writings on two sides, not four (the left and the right is the data for the RAM when available). And the RAM module from Hynix is clearly seen on the mobo.
That would account for the reduced package size. I find it hard to believe that MacRumors actually measured the die size, but just measured the package size.

Secondly.. the markings and the package itself is very similar to the markings and packaging that Samsung is using.
The original A5 was called APL0498 339S0130 (fabbed and packaged by Samsung)
The A5R2 was called APL2498 339S0173 (fabbed and packaged by Samsung)
This new A5R3 is called APL7498 343S0624
It's the second part of this designation that describes what plant and fabrication process is used. It's very similar on the package on the ones we know Samsung has made.
TSMC would probably not mimic Samsung's designations.

Pretty sure the DRAM is to the left in that photo and the Hynix chip is actually the NAND package. Also, MacRumors measured the package at 6x6 mm, whereas the other sizes quoted were the actual die sizes as measured by Chipworks. Thus this is at least a 50% reduction, which means a full node down from 32 nm. Apple can have the package silkscreened however they like (and I'm sure they do). I don't see how this can't be 22 nm (or smaller).

Edit: Hynix H2JTCG8T22MBR = 64Gb (8 GB) E2NAND3.0