Depends on what you mean by dedicated. Some -- QCOM and I believe TI -- use DSP blocks instead of purely fixed-function. Obviously these are fairly customized DSP's for video encode/decode.
Next-Gen iPhone & iPhone Nano Speculation
- Thread starter Arun
- Start date
Depends on what you mean by dedicated. Some -- QCOM and I believe TI -- use DSP blocks instead of purely fixed-function. Obviously these are fairly customized DSP's for video encode/decode.
It was poorly worded; if I would have said dedicated hw blocks it might have been clearer.
VXD isn't that small if you need to support a lot of video standards despite being able to share quite a bit between them. That's the disadvantage versus a DSP-centric (but I wouldn't necessarily say DSP-based?) approach like Qualcomm's. So very roughly VXD scales based on the number of standards supported (selected by the customer based on market requirements) and VXE scales by the number of pipelines (selected by the customer to achieve their target performance at a given clock speed - iirc the clock rates for given standards are stated for 2 pipeline configurations). VXD cannot scale the number of pipelines and VXE always support the same standards. The required clocks are very low for VXD but these are typical clocks and very complex streams require more so combined with 1080p 3D requirements it's actually fairly reasonable.
I could go on rambling about a whole bunch of other video IPs (I just realised I didn't really answer anything!) but I need to go to sleep so I'll stop now...
I could go on rambling about a whole bunch of other video IPs (I just realised I didn't really answer anything!) but I need to go to sleep so I'll stop now...
Quad-core I deem very likely as well. Beyond that - not so likely.
There is no doubt that smart-phones/tablets will migrate to the 64-bit ISA.
I don't think I disagree with much of anything in your post. My response was to Ailuros' claim that GPU performance would outpace CPU performance and that this somehow constituted him going out on a limb, when I felt that if so, it was a remarkably short, stubby and solid branch.
Not much of a risk taking.I do feel we're in the "gold rush" days of mobile computing. Devices with new and comparatively sophisticated capabilities are ballooning in market presence and value, there is a number of underlying hardware players jockeying for position, and there is still room for huge shifts in market share and even completely new segments to develop. We should enjoy it while it lasts, because it won't last forever. My approximation was that until 20nm we'd see some fairly aggressive moves in order to differentiate you products, but that after this we'd be at a point where additional computing resources would do very little to impress the average buyer, or even a buyer a couple of sigmas out from the average. (Joining Ailuros on his limb here) Not to mention that such increases will be progressively harder to come by just as in higher power segments of computing.
ltcommander.data
Regular
http://www.reuters.com/article/2011/12/16/us-apple-samsung-idUSTRE7BF0D420111216
Apparently, Samsung's new Austin Texas fab has now reached full production of 40,000 wafers a month. Reportedly, nearly all of it is going to Apple. There's no word on 32nm status, but you'd think with so much investment just coming online now Apple wouldn't be on the verge of switching to TSMC?
Apparently, Samsung's new Austin Texas fab has now reached full production of 40,000 wafers a month. Reportedly, nearly all of it is going to Apple. There's no word on 32nm status, but you'd think with so much investment just coming online now Apple wouldn't be on the verge of switching to TSMC?
Was it ever confirmed that Apple will switch to TSMC? Judging from the rumors I could believe that Apple sent some sort of testrun chip to TSMC for 28nm, but judging from the general yield/capacity problems for TSMC's 28nm it sounds like bad timing right now. I'm not even sure that Apple wants to switch entirely to another foundry and isn't just considering some sort of dual sourcing scheme in case they'd want even higher SoC volumes than so far.
anexanhume
Veteran
No, it was never confirmed. It's been rumored that they were switching and it's also been rumored that they are sticking with Samsung (as in actively sourced rather than just guessed at they weren't switching). As long as Samsung keeps their process nodes up with TSMC and/or GF, I don't see a huge reason for them to switch if they like their current partnership. There's a lot of demand for TSMC's 28nm processes, but Samsung's demand is thinner and steadier as a result, it would seem.
anexanhume
Veteran
I agree. Samsung's 32nm process is much more proven than TSMC's 28nm process. The last thing Apple needs is poor yield or congested fabs driving up price and/or availability.
Now Intel just needs to open their fab to everyone to make it interesting
Now Intel just needs to open their fab to everyone to make it interesting
To everyone or Apple
Apple's strategy with the GPU is to give it the priority of putting in the most that's reasonable at the time/node, though still factoring in the desire to not disrupt the software ecosystem with new processors every single year.
Even without OpenCL, Apple use the GPU to assist in various not-strictly-graphics tasks of the UI, browser, and imaging, and they plan for a lot more on the imaging side in the future. Plenty of iOS games feature visual enhancements for the A5 SoC, and the majority of all games play noticeably smoother on iPad 2/iPhone 4S (quite a few various projects are actually making use of UE3 recently.)
Apple know that the standards of software, including games, are escalating quickly. They were the driving force behind OpenCL, so they'll be letting developers in on the additional general purpose resources sometime down the line, as well.
Even without OpenCL, Apple use the GPU to assist in various not-strictly-graphics tasks of the UI, browser, and imaging, and they plan for a lot more on the imaging side in the future. Plenty of iOS games feature visual enhancements for the A5 SoC, and the majority of all games play noticeably smoother on iPad 2/iPhone 4S (quite a few various projects are actually making use of UE3 recently.)
Apple know that the standards of software, including games, are escalating quickly. They were the driving force behind OpenCL, so they'll be letting developers in on the additional general purpose resources sometime down the line, as well.
EPOCH.
ltcommander.data
Regular
Including the Epic Citadel demo, I think we're up to 10 UE3 games on the App Store.
Epic Citadel
Infinity Blade
Dungeon Defenders
Gyro13
Warm Gun
Epoch
Dark Meadow
Infinity Blade II
Desert Zombies: Last Stand
Batman Arkham City Lockdown
Afterlife: Ground Zero has also be announced as well as 4 UE3 games from Gameloft, one of which March of Heroes was cancelled.
Epic Citadel
Infinity Blade
Dungeon Defenders
Gyro13
Warm Gun
Epoch
Dark Meadow
Infinity Blade II
Desert Zombies: Last Stand
Batman Arkham City Lockdown
Afterlife: Ground Zero has also be announced as well as 4 UE3 games from Gameloft, one of which March of Heroes was cancelled.
If we take a strong look at Apple's previous 5 SoCs, we see a pattern of designing a chip for one process, then bumping up clocks for a smaller process. It's a nice, consistent way to "guarantee" performance increases every generation of phone Apple makes. Apple should at least have a very good idea what kind of CPU 1-2 years out.
As nifty as quad-core sounds, I don't think it's really all that great, especially on a mobile device. Heck, even on a desktop, you have to think really hard to find consumer-level apps that take advantage of a quad-core chip. You have to take a look and see what if developers would really take advantage of having 4 cores instead of 2 faster ones (most would not). And a faster dual-core CPU means 0 rewrites to take advantage of performance improvements. Plus, are we really expecting Apple to pull a quad-core A15 out of their hat in late-2012 to 2013? Otherwise, a dual-core A15 would mean giving a quad-core CPU to developers for one generation, only to take it away in the next.
Lastly, I strongly suspect there's only one major SoC development team at Apple which is expected to produce a brand new SoC with a new architecture every two years. In the interim years, Apple can rely on die shrinks and clock speed increases to get them more performance. When you take that into account, Apple knows that building that they are building the A5 not only for the iPad 2 and iPhone 4S, but a potential high-resolution iPad 3 as well.
As nifty as quad-core sounds, I don't think it's really all that great, especially on a mobile device. Heck, even on a desktop, you have to think really hard to find consumer-level apps that take advantage of a quad-core chip. You have to take a look and see what if developers would really take advantage of having 4 cores instead of 2 faster ones (most would not). And a faster dual-core CPU means 0 rewrites to take advantage of performance improvements. Plus, are we really expecting Apple to pull a quad-core A15 out of their hat in late-2012 to 2013? Otherwise, a dual-core A15 would mean giving a quad-core CPU to developers for one generation, only to take it away in the next.
Lastly, I strongly suspect there's only one major SoC development team at Apple which is expected to produce a brand new SoC with a new architecture every two years. In the interim years, Apple can rely on die shrinks and clock speed increases to get them more performance. When you take that into account, Apple knows that building that they are building the A5 not only for the iPad 2 and iPhone 4S, but a potential high-resolution iPad 3 as well.
Going from multi-core to "more multi-core" doesn't fall outside the refresh ballpark since you're not really changing architectures. In such a case it's more a design decision if with a smaller process you want to only go for higher frequencies or only for more cores or a fine balance between the two.
The problem with your analysis is that it makes sense.
A bit more seriously, and this one I mean, is that the scheme is too predictable.
Regrettable as it may be, part of the design of any consumer device is marketability. And if Apple believes that superficial tech websites are in any way a significant part of what shapes market perception of their devices, then that might be reflected in SoC designs. GIven that multi-threaded cache resident benchmarks are the order of the day at such sites, and the lack of significant (for Apple) die cost of adding cores, particularly at 32/28nm, going quad A9 could simply be an issue of playing it safe, marketing wise. Quad has to be better than dual, right? Going to quad A15 would be the next probable step. It would cost a bit, but not too much, and Apple is in a good position to put the squeeze on their competition in that regard.
The big question is how aggressive Apple wants to be. The tablet market is theirs to loose, and they have been aware for some time that Android will be joined by a Wintel attack, come Windows 8. Increasing screen resolution and quality is but one way to stay a step ahead of the competition, they are bound to use any other hardware means at their disposal as well, and SoC cost/complexity is one area where they are capable of being very competitive. Caches, data paths, GPU width, core count et cetera are all examples of where they could strengthen their offering and distance themselves from the also-rans. If the cost in terms of battery life is small enough, I wouldn't be surprised to see Apple going above and beyond the scheme you outlined, for market positioning reasons rather than purely technical ones, basically, ensuring that technological predictability isn't taken advantage of. I'd say they already did this with their A4->A5 move on the same lithographic process.
PS. Unfortunately the Cortex A9 doesn't really support going beyond four cores. Otherwise it would have been wonderfully ironic to see Apple put 16 A9s on their SoC, typically "turned off" obviously, but there when called upon to play the numbers game.
The projection that Apple would use 543MP2 + dual A9s came well before the rumors and driver discoveries started to surface simply because that configuration best matched the die budget trends the semiconductor companies appeared to be targeting.
I don't think Apple went beyond expectations, assuming one already accepted the fact that their concurrent development of end product and SoC and also their strategic partnership with IMG gave them an advantage in time-to-market for the performance crown.
I expect mobile CPUs to settle on quad core configs soon enough, not because of the meager gain in performance but because of the improved opportunity to manage power consumption. Balancing workloads across independently controlled cores, getting to use a lower voltage from a lesser need for clock speed, etc.
I don't think Apple went beyond expectations, assuming one already accepted the fact that their concurrent development of end product and SoC and also their strategic partnership with IMG gave them an advantage in time-to-market for the performance crown.
I expect mobile CPUs to settle on quad core configs soon enough, not because of the meager gain in performance but because of the improved opportunity to manage power consumption. Balancing workloads across independently controlled cores, getting to use a lower voltage from a lesser need for clock speed, etc.
One quick note: Apple controls their OS. Therefore it should be relatively simple for them to implement ARM big.LITTLE in a true heterogeneous multiprocessing (A7 & A15 cores doing useful work simultaneously). That makes the 'problem' of switching from 4xA9 to 2xA15 irrelevant since it would really be 4xA9->2xA15+XxA7 where X >= 2.
Mind you, I don't think Apple will go for 4xA9. But I don't think it would be a bad thing either if it doesn't slow down their move to A15+A7 (which is a big if - that's clearly a much more important upgrade). As I've said time and again, there is *no* power cost to going quad-core. It either results in the exact same level of performance at the same power (power gating for lightly multithreaded workloads) or higher performance at the same power (undervolting all cores for highly multithreaded workloads). It's purely an area cost - and that's something Apple can clearly afford.
I agree that their next-gen SoC will probably be very similar but merely increasing the number of cores is much easier than changing the CPU or GPU was in the old days. So given their area budget I don't think a straight shrink is very likely either.
Mind you, I don't think Apple will go for 4xA9. But I don't think it would be a bad thing either if it doesn't slow down their move to A15+A7 (which is a big if - that's clearly a much more important upgrade). As I've said time and again, there is *no* power cost to going quad-core. It either results in the exact same level of performance at the same power (power gating for lightly multithreaded workloads) or higher performance at the same power (undervolting all cores for highly multithreaded workloads). It's purely an area cost - and that's something Apple can clearly afford.
I agree that their next-gen SoC will probably be very similar but merely increasing the number of cores is much easier than changing the CPU or GPU was in the old days. So given their area budget I don't think a straight shrink is very likely either.