Next-Gen iPhone & iPhone Nano Speculation

[Ailuros]

As far as I know only one manufacturer announced products based on this GPU (ST-Ericcson), obviously Apple never announced any plans whether they will adapt it or not.

When did Apple ever since the beginning pre-announce what kind of graphics processor (or CPU for that matter) their next generation SoC will contain?

While it has nothing to do with the subject there are according to 6 Series6/Rogue licensees. ST Ericsson, Texas Instruments (OMAP6) and MediaTek have been officially announced. Take a guess who the remaining 3 unnamed lead partners could be. However I wouldn't expect from anyone anything from Series6 before 28nm and that for a good reason.

If I'd were Apple I'd take the A5 through a die shrink to the next best smaller manufacturing process, save quite some die area (and resources) and simply bump up the clocks.

I don't see why there would be a need for a MP4 as others suggest; sure four cores are better than two, but so does N*2 frequency vs. N. It doesn't even have to be twice the iPad2 MP2 frequency. Even at 400MHz it's already twice as fast on paper than the MP4 in the PS Vita.

Besides everyone seems to expect that because display resolution could be 4x times as high an equal bump of graphics power would have to happen (or I'm reading a few things the wrong way here).

1. Just because the native resolution could end up being in "iPad3" at 2048 or even higher, it doesn't necessarily mean that 3D applications would run in that resolution. How many mobile games are there out there that support such resolutions? For 1024*768 the MP2 in the iPad2 sounds to me like it's got fill-rates to spare and not the other way around. I assume (but do not know) that the MP2 frequency in iPad2 is =/>250MHz. Given 4 TMUs that equals to 1 GTexel/s raw fill-rate.

2. What exactly has the entire competition in the works for hypothetically early 2012 (and I'm obviously talking about final devices on shelves) that they'd need to move from a MP2 in iPad2 immediately to Series6 in the followup tablet? Yes ST Ericsson has claimed that A9600 will sample in late 2011 (if all goes according to plans I assume) but the SoCs will end up in devices when exactly?

3. Which manufacturing process will be used for Apple's next SoC? I severely doubt they'll move straight from 45 to 28nm, but that's probably just me. But when it comes to squeezing a Rogue into anything <28nm a wild exaggeration could be how you'd fit an elephant in a mainstream household refrigerator: open door, push elephant in, close door. That simple

It'd make psp2 blush for sure.

If you'd take the PS Vita and let it run iOS for sure. In any other case I suppose the latter being a handheld console, it retains for a bit of time it's "closed box" advantage without any sw overheads.

 

[Arun]

You forget that Apple has designed (not just specified) many more chips than they publicly let on. In addition to the A4 and A5, you've got three chips with Apple markings: the 65nm ARM11 in the iPod Touch 2G, the 65nm Cortex-A8 in the 3GS, and the 45nm Cortex-A8 in the iPod Touch 3G. The first and the last were clearly made to gain design experience on the process. Based on that cadence (which btw is faster than TI or NVIDIA) I would expect at least two more chips before a 28nm Rogue-based SoC from Apple for the iPad 4 in late 2012 or early 2013.

 

[ltcommander.data]

If I'd were Apple I'd take the A5 through a die shrink to the next best smaller manufacturing process, save quite some die area (and resources) and simply bump up the clocks.

I don't see why there would be a need for a MP4 as others suggest; sure four cores are better than two, but so does N*2 frequency vs. N. It doesn't even have to be twice the iPad2 MP2 frequency. Even at 400MHz it's already twice as fast on paper than the MP4 in the PS Vita.

Perhaps it's hard to make a general statement, but which is the more power efficient way to double performance, doubling clock speed or doubling transistors (ie. MP4)? Apple would probably choose whichever option is more power efficient even if it costs more to have a bigger die. Presumably with a shrink from 45nm to 28nm, Apple now has die area to spare, and if SGX cores are pretty area efficient, doubling core count should be very viable.

Besides everyone seems to expect that because display resolution could be 4x times as high an equal bump of graphics power would have to happen (or I'm reading a few things the wrong way here).

1. Just because the native resolution could end up being in "iPad3" at 2048 or even higher, it doesn't necessarily mean that 3D applications would run in that resolution. How many mobile games are there out there that support such resolutions? For 1024*768 the MP2 in the iPad2 sounds to me like it's got fill-rates to spare and not the other way around. I assume (but do not know) that the MP2 frequency in iPad2 is =/>250MHz. Given 4 TMUs that equals to 1 GTexel/s raw fill-rate.

Well, popular casual titles will no doubt quickly make the transition to Retina on the iPad 3. But of course we are more concerned with more graphically aggressive games from major developers. What I'm thinking about here is the state of the asset pipeline for major franchises/developers. Much to the chagrin of PC games, the current consoles target 720p, if not lower, and so game assets target that resolution, which matches well with the current 1024x768 iPad resolution. However, by 2012, the Wii U will presumably be out and Microsoft and Sony could be announcing their next gen consoles which will precipitate a transition to 1080p assets being standard, which will match well with a Retina iPad 3. As the power of mobile devices increases, developers will no doubt want to have as much asset reuse cross-platform as possible. So if the iPad 3 has a Retina Display, it would probably most convenient for major developers to use their 1080p assets if Apple includes a sufficiently powerful GPU to drive the higher resolution, so I don't think it'll be the case where the Retina display resolution will go unused for graphically intensive 3D games.

EDIT: The other thing here is with HDMI out and now AirPlay Apple isn't interested in just driving the native display. In fact, the just leaked release notes for iOS 5.0 Beta 2 say Airplay is now enabled by default for all video content in applications and websites. Secondary display capability will just put further pressure on Apple to keep increasing GPU power. It'll really be something to see an iPad 3 driving a Retina Display and a 1080p TV, even if the native display is only doing 2D information.

You forget that Apple has designed (not just specified) many more chips than they publicly let on. In addition to the A4 and A5, you've got three chips with Apple markings: the 65nm ARM11 in the iPod Touch 2G, the 65nm Cortex-A8 in the 3GS, and the 45nm Cortex-A8 in the iPod Touch 3G. The first and the last were clearly made to gain design experience on the process. Based on that cadence (which btw is faster than TI or NVIDIA) I would expect at least two more chips before a 28nm Rogue-based SoC from Apple for the iPad 4 in late 2012 or early 2013.

Well if Apple continues on their cadence, there will only be 1 more chip before a 28nm Rogue not 2.

90nm 412MHz ARM11/MBX Lite (2007)
65nm 533MHz ARM11/MBX Lite (2008)
65nm 600MHz Cortex A8/SGX535 (2009)
45nm 1GHz Cortex A8/SGX535 (2010)
45nm 1GHz Cortex A9/SGX543MP2 (2011)

Continued cadence:
32nm/28nm High clock Cortex A9/SGX543MP2 (2012)
32nm/28nm Cortex A15/Rogue (2013)

EDIT: I just realized I forgot about the 3rd gen Touch getting a 45nm shrink of the 600MHz Cortex A8/SGX535 SoC. The cadence then doesn't look quite as clear with Samsung's 45nm process seeming to hang around for too long. Unless of course, the rumours of TSMC getting Apple A5 orders are actually correct. The A5 could get a minor shrink to TSMC 40nm just for the iPhone 5 and 5th gen Touch to get experience with TSMC's tools using a known design. The iPad 2 could remain at Samsung until end of life.

 

[Erinyes]

If I'd were Apple I'd take the A5 through a die shrink to the next best smaller manufacturing process, save quite some die area (and resources) and simply bump up the clocks.

I agree and I think thats the general sentiment from past discussions as well, and it fits in with the cadence they've been following so far (with the exception of the ipod 3G 45nm chip). An A6 with dual A9's at 1.5 ghz and 543 MP2 at 400 mhz will still be a powerful chip in early 2012 and i think thats the way they're gonna go. There wont be a match for it until MSM 8960 or OMAP 5 are out in H2 2012. Tegra 3 will be out a few months earlier and ahead in CPU performance but will lose out on graphics performance (and even 8960/OMAP 5 will probably just about match this hypothetical A6 on graphics performance. Of the announced SoC's so far, only A9600 seems like it will beat this hypothetical A6 in graphics performance).

About the process, well apart from 28nm which other process is there? Rumour is that they're moving away from Samsung and to TSMC, in which case there's only 28nm. And if they are staying with Samsung, does Samsung intend to produce anything on the 32nm node or are they jumping straight to 28nm as well?

 

[Arun]

About the process, well apart from 28nm which other process is there? Rumour is that they're moving away from Samsung and to TSMC, in which case there's only 28nm. And if they are staying with Samsung, does Samsung intend to produce anything on the 32nm node or are they jumping straight to 28nm as well?

Samsung definitely has a 32nm process they intend to go in mass production with. I don't buy the idea that they'd switch completely to TSMC - however, it would make sense to start dual-sourcing some of their chips to alleviate potential capacity constraints and gain even more leverage in wafer pricing negotiations.

 

[Ailuros]

Perhaps it's hard to make a general statement, but which is the more power efficient way to double performance, doubling clock speed or doubling transistors (ie. MP4)? Apple would probably choose whichever option is more power efficient even if it costs more to have a bigger die. Presumably with a shrink from 45nm to 28nm, Apple now has die area to spare, and if SGX cores are pretty area efficient, doubling core count should be very viable.

Twice the frequency with MP2 in theory vs. twice the cores at the same frequency I'd suggest the first wins in terms of die estate, but would like to stand corrected. Granted increasing frequency is definitely not for free, but I have a hard time to swallow that it would cost as much as twice the cores/same frequency.

Besides since from what I can remember Samsung has a 32nm process as Arun states, why would they wait that long until 28nm is ready?

Well, popular casual titles will no doubt quickly make the transition to Retina on the iPad 3. But of course we are more concerned with more graphically aggressive games from major developers. What I'm thinking about here is the state of the asset pipeline for major franchises/developers. Much to the chagrin of PC games, the current consoles target 720p, if not lower, and so game assets target that resolution, which matches well with the current 1024x768 iPad resolution. However, by 2012, the Wii U will presumably be out and Microsoft and Sony could be announcing their next gen consoles which will precipitate a transition to 1080p assets being standard, which will match well with a Retina iPad 3. As the power of mobile devices increases, developers will no doubt want to have as much asset reuse cross-platform as possible. So if the iPad 3 has a Retina Display, it would probably most convenient for major developers to use their 1080p assets if Apple includes a sufficiently powerful GPU to drive the higher resolution, so I don't think it'll be the case where the Retina display resolution will go unused for graphically intensive 3D games.

I don't see the next iPad to have a resolution below 2048 if not even higher. That's way too much for ISVs to jump on that bandwagon just yet. I'd rather have games stay for some time at lower resolutions and add in details and IQ improving features like MSAA and yes why not finally some anisotropic filtering for a change.

EDIT: The other thing here is with HDMI out and now AirPlay Apple isn't interested in just driving the native display. In fact, the just leaked release notes for iOS 5.0 Beta 2 say Airplay is now enabled by default for all video content in applications and websites. Secondary display capability will just put further pressure on Apple to keep increasing GPU power. It'll really be something to see an iPad 3 driving a Retina Display and a 1080p TV, even if the native display is only doing 2D information.

I don't recall the VXD (video decoding) block to be exactly small on the leaked A5 floorplan. I've no idea which version of the VXD it contains, but it should definitely be capable of dual streaming HD video without breaking a sweat and that at lower frequencies as the GPU. Apple's sw constraints considering any video content are a completely different chapter.

 

[ltcommander.data]

http://arstechnica.com/apple/news/2...-soc-production-away-from-samsung-in-2012.ars

So Ars is also reporting that their own source is saying that Apple is moving to TSMC 28nm. For such a common process in terms of usage by other OEMs, I wonder if Apple is able to negotiate first to market releases as Apple is occasionally able to do with Intel processors.

 

[argor]

http://arstechnica.com/apple/news/2...-soc-production-away-from-samsung-in-2012.ars

So Ars is also reporting that their own source is saying that Apple is moving to TSMC 28nm.

personally i don’t think apple will stop using samsung as a fab i think maybe they will start dual-sourcing from TSMC and samsung
there may be few reasons for starting dual-sourcing
leverage in wafer pricing negotiations
potential capacity constraints
and for those that say apple will completely stop using samsung as a fab remember why TI stop using samsung as a fab
hint google " samsung texas instruments fab "

 

[Ailuros]

IMHO Apple wills start any dual sourcing starting with 28nm and if you take a wider perspective it's both better for Apple and Samsung. Apple will no longer depend on just one manufacturing source and Samsung will have far less problems with its own SoCs.

There are even reports at the moment that Apple might be supposedly manufacture its next SoC at 28nm/TSMC exclusively. Has any of those even bothered to ask whether TSMC has the capacity to serve Apple's needs and not disadvantage its other customers?

 

[Rys]

Samsung have 28nm and appear to have the capacity to keep serving Apple going forward, so I'm not sure I see the relationship changing if I'm honest. The only sticking point would be Exynos manufacture, but I'm sure they can work that out.

 

[wco81]

Well of course the suggestion is that the legal disputes between Apple and Samsung's mobile division may be affecting the relationship between Apple and Samsung's fab division.

There have also been rumors about Apple possibly striking a deal with Intel to use their 22 nm process, to make ARM SOCs, as strange as that sounds.


Going back to the original post of this thread, there are more rumors about a lower-priced SKU, to address the prepaid market, which is 50% or greater in countries outside of North America. If they can craft a set of SKUs that increases volume the way they would want -- minimal cannibalization of the high-end SKU, upsell from the low-end SKU -- maybe the anticipated increase in volume will drive relationships with new fabs.

Oh and there was one rumor about Apple trying to push suppliers for a crazy level of volume in the first month or two, something on the order of 15 million in August/September or September/October.

 

[Lazy8s]

I don't think anyone can ramp up that steeply. They'd just have to build inventory for a few months.

 

[Lazy8s]

For iPhone 5, I'm looking for Apple to pay some attention to OpenCL and also to have something new from Epic to showcase.

 

[rpg.314]

What about the possibility of thunderbolt in it?

Macs and cinema display are now thunderbolt enabled. If they have temporary exclusivity over thunderbolt, then it would make sense for them to include it. Besides, thunderbolt seems like an awful waste of time and effort if it's going to be limited to just Macs. They could have gone in with a USB3 chip as well since thunderbolt isn't on southbridge either. It's not like it has any useful advantage over USB3 unless you can build a real ecosystem around it.

iDevice paired to a cinema display would be cool setup. Though you can be sure that Apple won't let any body plugin their USB devices into such a system.

 

[wco81]

TB plug is small enough but can a mobile system generate that kind of bandwidth?

Seems like support for an HDMI dongle (because you know it's going to be too thin for any ports in addition to the dock connector) would make more sense.

Of course TB would support faster syncs as well as the ability to send video. But they cut out Firewire to save costs, even though USB 2.0 is poor for syncing even a 16 GB (a full restore which fills most of it up can take well over an hour).

 

[rpg.314]

Not in storage subsystem alone, no. But they already have HDMI out in iPad 2.

TB would be a nice add on though.

 

[Pressure]

The 30-pin connector already has support for USB 3.0 and Thunderbolt, as a former patent showed.

 

[rpg.314]

Then they'll probably keep a few features for thunderbolt only.

 

[RudeCurve]

Whatever form the iPhone 5 takes you'll be glad to hear that existing iPhones have incredible resale value which is good for people who want to get some money back when upgrading to the next iPhone.

http://9to5mac.com/2011/05/19/you-can-now-trade-in-your-ios-gadget-on-amazon/

A sampling of other iOS device trade-in prices includes up to $114 for iPod classics, $235 for first-gen 16 GB Wi-Fi iPads and $162 for an 8 GB iPhone 3GS. That’s definitely way more than you’ll get if you let your old iPhone rattle around in a desk drawer after upgrading to a new one.

 

[iwod]

Yes, Although you only get 10 - 20 dollars more if you have 32GB 3GS.

I really really hope Apple will have a 4 inch + screen.

Apple's record they have Max 2 - 3 Weeks between announcement and releasing date. Since it is pretty much confirmed that Next iPhone will be released on Late September or more likely early October.

This give Apple Less then 2 weeks to make an announcement.