Well, the real answer is that it's not an either/or scenario. Chip design teams design and layout chips based on off-the-shelf tools, layout expertise, etc. Silicon compilers are also constantly improved, but a completely different set of people are involved. Not sure about today's Apple, but the 80s/90s Apple probably would have done it both ways. In fact, even now I think about it, and the way the Intrinsity guys seem to work, it makes sense.
This is sometimes done in PCB layout. Sure, some types of layout, like RF, are pretty much always done by hand. Today's successful autorouters work from a large set of design constraints that describe the circuit at a high level. You may get a layout sufficient fir a prototype in a week rather than the month the hand layout would take. On a tight schedule, that's two free weeks of bring up.
So it might well be that they ran the silicon compiler, tweaked and simulated the piss out of the design, in parallel with the much longer hand layout. Maybe a bit like hand optimizing compiled code with downcoding/recoding and profilers.
As for hand layout/coding always beating compilers, not in practice. One is certainly a time constraint... the compiler will produce a better end product below a certain time limit. Just where that line is drawn depends on the coder and the complexity of the process. As the CPU/PCB/chip rule sets grow, the compiler does better than the human, assuming it can efficiently factor in the rule set. Naturally, when it can't, you need the personal touch. Until it's expanded to embrace the new rules...like vector or GPGPU coding today. The other factor is of course architecture.. human time is usually better spent on the big picture items, architecture and algorithm. No sense hand coding for a 10% improvement in quality when the same time spent on design might yield 100%.
Not to mention the JIT factor -- code is compiled on the fly for the specific CPU and system in use. Custom DSP pipelines created based on the problem at hand. A complex math problem is targeted,on the fly, to somewhere between 32 and 4096 parallel processing elements. Compilers always win here.
4G models are dual core, faster core. 3G models are quad core, same A9 we all know. This has something to do with Samsung mysteriously leaving an LTE port off their quad A9 Exynos SOC. Whoops.
Based on their dropping of the iPad 1 with iOS 6, it looks like Apple's current policy is to drop support once a product is no longer actively sold. The iPad 1 is slightly faster than the iPhone 4, but last sold new in 2011...it was reduced and sold alongside the iPad 2. Gone for good shortly before the iPad 3 debuted. We'll see how well the iPhone 3GS and iPad 2 do with iOS 7 next year... both sold new in 2012, the 3GS is gone, the iPad 2 probably before the next iPad in March.
So, following this formula, the iPhone 5 goes to $100 in 2013 and gets iOS 7, goes free in 2014 and gets iOS 8, maybe even gets iOS 9 on the way out the door in 2015. Great if bought one just recently -- but you'll upgrade anyway before then. Sucks if you're a "free" buyer, you may see only a few months of support, depending on the iOS schedule.
And that's actually the wrong release sequence. Its clear Apple likes to do iOS major upgrades for iPhone, then maybe a tweak for the next iPad. But they seem to be cutting out the iPad from the upgrade sooner in its model life than the iPhone. Much sooner, as they only keep one older iPad around for new sales, so far anyway. But iPads don't follow telco subsidy cycles, so the user is likely to keep them longer, even hand them down, like PCs often are. Looks like Apple's plan here is to seriously de-value older iPads, dropping support as soon as they retire.
How about "you have great 1080p recording capability, and would like to actually use it"? Camcorder quality 1080p runs about 12GB per hour. A 64GB microSDXC card runs about $50. Adding that SD interface to your phone, maybe a $0.25 connector and another $0.10 in passive components -- the SD interface itself is free with your SOC. So why the hell not have the SD slot? Well, it does make charging $100+ for 16GB flash a bit harder for consumers to swallow...
Actually, Samsung's A15 SOC is only sampling now. In real S3s, you either get a dual-core Krait based SOC (Qualcomm's in-house more-or-less A15 class core) or a quad A9. Given the Asus tablet I'm typing this on soundly beats the iPhone 5 on Geekbench, one would expect Samsung's quad to do likewise. The Krait is more evenly matched, a bit slower than the expected A15 result, but clocked higher than Apple's likely to go, unless they finally trust their power/thermal management enough to run at the core's rated speed for peak loads.
The trade dress argument is about the only one that Apple really seems to have on Samsung's older products. It's a judgement call for sure, but there's little question that the iPhone is iconic. The fact that Samsung built very similar looking devices, modified Android to look more like iOS, even used similar packaging and wall wart.. it all adds up. And this can be a trade dress violation, even without any single patent or copyright violation.
It says nothing about Android having a general problem. And it has very nicely illustrated the problems with such patent cases being decided by a neophyte jury, or worse get, a self-proclaimed expert with other priorities. I think it's also been shown that Apple is writing patents specifically designed to confuse such juries in their favor. Patent authoring itself has been an ongoing evolution, to work the PTO and their thin basis fir software patenting as fast and hard as possible.
If my phone is in my pocket screen-out, it's going to be nearly impossible to establish an NFC connection in a crowded subway. The phone itself (plus the battery... on my phone, the NFC antenna is actually in the battery, so that it can be close to the back surface) is a pretty good shield.
Is that that best you can do? Possible faked or overclocked CPU benchmarks?
So.... dozens of Galaxy SIII benchmarks are presumed to have been faked and/or overclocked, even with those having performed some of those here, telling you "no, it's stock hardware". But a random unsourced one-of benchmark claiming to be an iPhone 5 is somehow gospel handed down from on high?
How about comparing default spec versus default spec? Those are pretty irrelevant if they were either falsified or on a device with an overclock because a "phone" is supposed to be usable in your pocket and have battery life measured in hours around at least a work day long rather than minutes.
It's similarly irrelevant to judge one unsupported posting.
On the other hand, if I were working at Apple and knew the iPhone 5 had crazy killer performance, I'd post a number higher than the current competition, and totally faked. To get them all going crazy, posting higher numbers, etc. Then, next week, when real iPhones 5 hit the market, we'd see all those much higher actual numbers from the iPhone, and victory would be mine. Mine, I tell you, mine!
Good thing the iPhone 5 isn't really that fast and I'm not working for Apple.
Any Android running Jellybean can do that just as well. Anything you can actually perceive as being faster, rather than measuring via some kind of benchmark, has very little to do with processor speed, and everything to do with user interface latency. In other words, it's a matter of tuning. Apple's been pretty good about that, though they still fail in places where they should be using multitasking/multiprocessing but aren't.
Android, on the other hand, didn't seem to spend much of an effort on some of the "finesse" issues in Android until fairly recently. Android Ice Cream Sandwich made a great deal of progress on deciding just what Android's supposed to look like (eg, the "magazine" UI that showed up first on Google's web site, rather than just a mish-mosh of Palm, WinCE, Blackberry, and iOS ideas), and improved latency issues nicely. The most recent release, Jellybean, has tweaked latency, eliminated deadlocks, etc. so that it's every bit as good as iOS.
The key here is that humans are just slow. If the OS isn't responding as fast as you need it, that's an algorithm problem, not a CPU speed problem. If you have any doubts, find an old Amiga and try AmigaOS on a 1980's CPU that's not fast enough to be considered as an I/O processor on a modern smartphone SOC. It'll make desktop Windows seem slow, within limits (eg, you're not going to get any H.264 videos playing). This fully illustrates my point... AmigaOS was a realtime OS, and the UI manager ran at a much higher priority than nearly anything else. So there was never any waiting on an app or anything else -- the user was the most important thing in the system. So it seems really, really fast, even on a slow processor. There's no hard reason any modern device, PC or portable, should have a "slow" UI.
The SIII is not using the A15, but Qualcomm's own Krait design. Krait is supposed to be about 32% faster than the A9, while the A15 is expected to run about 40% faster than the A9. There's some suggestion that the A15 was being designed for higher power use, but then again, both Samsung and TI have released (to developers) A15 core processors intended for smartphones, not even just tablets or high processor count servers.
According to Arnand (http://www.anandtech.com/show/6292/iphone-5-a6-not-a15-custom-core), the iPhone 5 doesn't use the A15 either. He suggests this high power idea as a reason, but it doesn't make the slightest bit of sense that ARM would walk even a little bit away from their primary business (mobile devices) by overpowering the A15, or that the chip makers would use a non-phone-suitable core in a phone-oriented chip. And of course, ARM themselves list the A15 as a processor for advanced smartphones. Ok, sure, Intel has a chip for smartphones too....
It's also the case that Apple's second chip company, Intrinsity, were known for making chips faster -- not low power. They have a bunch of tricks to get to higher speeds in older process technologies, at least some of which (like dynamic latches) aren't the right formula for a mobile device that might be clocked way down for power savings (dynamic latches establish minimum clock speeds -- static logic can clock all the way down to DC). Apple's using Samsung's 32nm process, and may move some things to TSMC's 28nm process (along with AMD,TI, nVidia, Qualcomm, Broadcomm, and the rest of the fabless world -- rumor has it Apple actually tried to buy all of TSMC's 28nm capacity for some unspecified period, and were told to go pound sand).
One interesting observation about this: ARM Holdings (the ARM company that does the basic design of ARM processors, MALI GPUs, etc) made $462M in gross income in 2011, and probably something more in 2012. But still -- could Apple or Samsung or Qualcomm or even TI match ARM's level of investment in CPU development? Sure they could... and ARM still gets paid when they do. Apple's got so much money and, in particular, if they just concentrate on the CPU, they might deliver substantially better CPU cores for their specific purposes than the other ARM licensees. It'll be interesting to see if any of the others, aside from Qualcomm of course, get into their own development.
Qualcomm's done their own designs for awhile. Their previous core, the Scorpion, was about 5% faster than the ARM Cortex A8, not as fast as the A9 (about 25% faster than the A8), thus most Qualcomm powered phones being clocked a bit faster than the A9 phones.
Anyway, if true about Apple's core, that make even more interesting, with ARM, Qualcomm, and Apple all making their own ARM cores. And Samsung still in a very strong position, being the only company in the ARM race with their own IC fab (well, TI does, but not for this class of parts). Oh, well, sure, Intel planning to fight it out as well on mobile... they also make their own chips.
Some are overclocked... easy to tell, since the benchmark supposedly calculates the actual processor speed. Not a big surprise -- most smartphone processors run below their rated speeds, in order to use a bit less power.
The big boost seems to be the SIII going from ICS to Jellybean. So far, every release of Android has improved the performance of the Dalvik VM. Not specifically an issue for native code (not sure if the benchmarks are done in the VM or the NDK; you'd want the NDK to best judge hardware performance), but it's going to affect parts of the system, even when using the NDK. You might also expect better performance if run with minimal active background services.
Not in a practical way, though. The iPad Don't-Call-Me-3 has twice the GPU performance (SGX 543MP4 vs. SGX 543MP2), but four times the pixels to paint.
The Samsung Galaxy SIII is faster than four Cray Y-MPs (or a couple of billion HP45 calculators)... at least if you're not too particularly about how your GFLOPS are served up. It's also got memory, unless you upgraded to the Y-MP M90. And uses quite a bit less power. That's 24 years for ya!
Nothing particularly useful or interesting about such observations, I suppose, unless you put the name "Apple" in the title. Or maybe just an Apple fan's way of dealing with the iPad 2 not being a terribly fast device, as mobile devices go. But watch out, all you 80s supercomputers.
My kids Nintendos have also been really abusive to our Cray 1, chasing it around the house, picking on it, calling it "pokey" and "turtle-boy", and just being bratty. Even the Nintendo 64, who's old enough to know better.
I'm cynical enough to believe they make these numbers up just to generate hype over a new product.
The inevitable "shortage" of 'phones over the month after launch? That could be deliberate, too, just to make people who got one feel special (and the ones who didn't to walk around muttering at the floor).
It's all deliberate. And well planned. Apple sold 1 million 4S last year via pre-orders. So naturally, they're going to sell more this year. But not so many that they'll have trouble selling 2.5M or 3M iPhones 5S or 6 next year.
Keep in mind, too, that selling a million smartphones in a 24 hour period is no kind of amazing accomplishment. Last year, around Christmastime, there were several days that topped one million Android devices, and I think Apple even hit two million once or twice. There are currently about 1.3 million Android activations every day.
In short, very credible numbers, given the month long hype circus that leads up to any iPhone pre-order day. And to the iPhone in-store day as well. Apple has generated this fear that you won't be able to get your new iPhone. They did that on purpose because it gets all sorts of free press, and just makes the iPhone seem a more attractive thing -- anything that's hard to get must be worthwhile, eh?
It's the sure sign a new model is coming when supplies of the old one start drying up.
Not really. It's a sign that a new model is coming when orders for the old model start shrinking. Apple keeps selling the old models -- they just rotate them down the price structure. And EVERY knows this. The iPhone 5 launch was the least well kept secret in technology, and that's exactly how Apple wants it. Rather than have the press talking about the new iPhone for a week, they're doing that for a month before the official availability. So people just stopping the 4S, because everyone knew the 5 was coming this month. If you didn't like the 5, you'd be guaranteed to get a 4S for $100 less right after the iPhone 5 announcement. Easy.
Though, an annoyance seems to be Apple never makes "enough" - they always sell out on the weekend and supplies are extremely tight. It's a marketing ploy, maybe (though each Apple store seems to get a few daily), but I guess when you're spending billions on making the things, you get cautious and don't blithely order tens of millions for release day and tie up money on inventory.
No, it's much more than that. Apple does their marketing homework -- they knows with certainty they'll see ten or twenty million iPhones this fall, and they have no concerns about tying up money in inventory -- they have $100+ in cash (though most of it outside the USA). In fact, they probably have a very precise estimate of sales, and they're working to deliver that much product.
This is all part of the marketing process that is an iPhone launch. They could have 5 or more million available for pre-order online. They don't. They have enough to beat last year's record enough to get the press reporting it. But not so much breaking that record next year should be a big problem. That's optimizing the press they get from the pre-release, just as they optimize the impact of the announcement. I mean, if Apple felt like it, they could have the announcement, then start the pre-sale that very day, maybe the store sale that day or the next. But they don't -- they give each of those events time to get reported, all that free publicity. And more, all that reinforcement in the iPhone users' minds that this is something special. No other new device release makes this kind of impact.
They also have to be careful about pre-selling too many online, because the next big event on the calendar is the in-store release. That's also going to break last year's record, and they're also going to sell out. iPhone fans know this, so they line up outside Apple stores across the country, and sure enough, it's another big media event. It gets reported, the lines, the eventual sellout, etc. Again, this isn't like any other "phone".
Magically, once the noise all dies down, though, ample supply of the iPhone 5 will appear. This is also part of the process.. having generated all this excitement, they really don't want to lose that momentum.
This is all part of the Apple iPhone Launch song and dance.
They know pretty accurately what they're going to pre-sell. They pre-sold 1M iPhones 4S last year because that's how many they made available for pre-order. This year, it was 2M iPhones 5. Apple's spent the last month building the anticipation: leaks, event announcements, etc. Even the wait from the announcement to the online ordering being enabled. And the fact they sell out in store every year -- all of this is well planned.
Given that only about 25% of Apple's profits ever enter this country, it seems fairly unlikely the iPhone will have an important effect on the real economy, whatever their effect on the GDP.
If you're a hardcore smartphone user, and run your battery down regularly, you'll use the replaceable battery. The simple fact is that no one's Li-ion battery will last two years of full discharges every day. The ability to replace the cell yourself is a win, even if you don't ever plan on carrying spare cells.
If, on the other hand, you keep your device topped off pretty well, maybe with a charger at work or whatever, the battery will probably last the life of the device.
I'd been holding out with my iPhone 4 for a while, waiting (like many others, I suspect) to see what Apple would wow us with for the iPhone 5. Needless to say, I wasn't that impressed, though to be honest, part of me really didn't expect to be, given that there are only so many innovations they could have come up with. What could they have done? An even bigger screen? NFC? A phone you could roll up?
That's kind of the point of the disappointment, isn't it -- no innovations. And that's precisely because an innovation is something unexpected, not just a safe, incremental improvement.
Ultimately, an innovation is a judgement. You do something risky. If it sells, that's an innovation. If not, it's probably on someone's worst products list for that year, and then forgotten. The original iPhone was an innovation, taken as a whole. Not really tech innovation, a business innovation -- Apple was doing things everyone in the phone business absolutely knew were impossible. They were going to sell a smartphone (well, originally, a feature phone) to consumers. As the most expensive phone on the planet, none-the-less. And they were dictating their terms to the carriers, rather than kowtow to them. A relatively big risk (tough even if the phone failed, they might have still launched a decent iPhone Touch).
But Apple's riding high now, with over 30% of the US smartphone market, and most of the profits. Every new iPhone has been just an incremental improvement, some hailed as the Jesus Phone, some written off too little of an improvement, all very, very successful. Apple has the formula down, and they're sticking to it. There's no room here for a gamble, and thus, no room for a significant innovation.
What are the big changes here? Faster CPU/GPU... yeah, that happens pretty much every year, on every phone. Longer screen... pretty much a requirement. Every competitor has a 16:9-ish screen. They didn't try to the match the typical 1280x720 resolution. They did use a new LCD technology, leaving behind IPS for something new from Sharp that integrates the touchscreen in the display, transmits more light -- so, thinner (not as thin as OLED) and lower power. Good place for an improvement, but hardly noticeable to the average user. And then the connector -- maybe that's actually innovative, if it does anything useful that a plain old microUSB/MHL connector can't do. And also the main controversy.
Apple really never going to be the place to look for innovation in smartphones or tablets again -- they're too successful. They'd need a couple of model releases per year to even try out something very different, without risking the whole franchise. With the single model, no chance of risk talking. It's easier to watch all the Android devices life or die in the market, and adopt the innovations from those that rise to the top.
I have the Galaxy Nexus as well... I do believe there's a chip containing unicorn poop in there.. that's where the on-screen rainbows come from.
As well, the TI OMAP 4460 improved the overall system. It wasn't just dual ARM Cortex A9s, but dual memory bus controllers -- much the same deal you find on the average desktop PC. That's one key to keeping the processors (and GPU of course, don't forget all of these SOCs are "shared memory" devices, in the sense of PC graphics being shared) well fed. Some of the quad cores, like the nVidia Tegra 3, still use a single (though somewhat faster) memory bus. So when you have good cache hits, the Tegra 3 pulls ahead, when running memory bound, the 4460 will be faster.
In short, performance is all about the total system, not any given spec. The Tegra 3 ought to be significantly faster on 3D games. That's also what the iPhone has typically been optimized for -- Apple sells a crazy boatload of games, and they've pretty much made that a thing in the iTunes store now.
I think if you chose well, as we did with the GN, you're not going to see a huge upgrade opportunity in just a year. Sure, if someone gave me a free SIII I'd use it, but if I have to spend real money (and maybe a bit of it, if I want to keep my unlimited plan on Verizon), I'll wait for a substantial upgrade. Hopefully one that tunes battery life in as well.
It's not just the cores, but what they do. The four Cortex A9s in your Nexus 7 may or may not be faster than the dual A15 in the iPhone... looks pretty close. The A15 was expected to be about 40% faster than A9, but this the first time one's actually been benchmarked "in the wild"... assuming, as most think, it is an A15.
They also have either a four core GPU based on the PowerVR SGX543MP4, or something new. This is already faster than the Tegra 3's GPU... much as you find, on desktop processors, you need several AMD GPU cores to match a single nVidia core.
Well, they used to be "Apple" in the A4 days -- the A4 was really just a customized version of one of the Exynos chips. Apple could retroactively claim to have some design work on it, for those who want to be pedantic about, though. Samsung worked with Intrinsity, Inc. on the design of the Hummingbird processor. Like every A8 over 600MHz or so, this was a customized version of the Cortex A8. Apple later bought Intrinsity.
These days, they are doing the SOC design, and so far, Samsung's doing all the fab work, though there have been rumors that Apple's been shopping for a second source, and may be going to TSMC. Qualcomm uses TSMC, and in fact, has been frustrated with the output from TSMC's 28nm process. Enough so they tried to buy exclusive access to that from about a billion. Apple reportedly tried the same thing; TSMC said "no" in both cases. Which is in their long-term interest -- an independent fab can't play favorites and stay independent. Qualcomm's moved a bunch of work to Global Foundries (the AMD spinoff), and reportedly also have talked to UMC and Samsung about making the Snapdragon S4. nVidia, also a long time TSMC customer, seems to be doing similarly, looking at UMC and Global Foundries for increased 28nm support.
In short, business as usual. Many of the big names -- even AMD these days -- design the chips, use the big chip foundaries to actually make them.
The dual core versions run the Qualcomm Krait processor, which is about 30% faster per clock cycle than the ARM Cortex A9. So head to head tests will vary. And of course, if your benchmark isn't multi-threaded, you're just testing single core performance. That's at least one way the actual effect of additional cores can be made confusing.
Steve Jobs loved a thin phone -- so iPhones are thin. That's kind of their thing.
They have also been fairly fragile, and yeah, you can see some crazy cases, like the Otter Box, that pretty much erase any purpose to the thin design. I've seen the same thing with laptops... a guy I used to work with had a super think MacBook Pro. Which was always kept in some kind of Pelican case, making it the most bulky laptop ever seen.
Point being, a slightly thicker phone that doesn't need a case to survive works out much thinner than a phone that requires a case. It's not just Apple... the Galaxy Nexus requires a case. The plastic used wouldn't be rugged enough for a kids' toy, and the glass scratches fairly easily, which is close to crime these days, given that's one of the problems that seems to be pretty completely solved by Gorilla Glass.
Slashdot Mirror
Well, the real answer is that it's not an either/or scenario. Chip design teams design and layout chips based on off-the-shelf tools, layout expertise, etc. Silicon compilers are also constantly improved, but a completely different set of people are involved. Not sure about today's Apple, but the 80s/90s Apple probably would have done it both ways. In fact, even now I think about it, and the way the Intrinsity guys seem to work, it makes sense.
This is sometimes done in PCB layout. Sure, some types of layout, like RF, are pretty much always done by hand. Today's successful autorouters work from a large set of design constraints that describe the circuit at a high level. You may get a layout sufficient fir a prototype in a week rather than the month the hand layout would take. On a tight schedule, that's two free weeks of bring up.
So it might well be that they ran the silicon compiler, tweaked and simulated the piss out of the design, in parallel with the much longer hand layout. Maybe a bit like hand optimizing compiled code with downcoding/recoding and profilers.
As for hand layout/coding always beating compilers, not in practice. One is certainly a time constraint... the compiler will produce a better end product below a certain time limit. Just where that line is drawn depends on the coder and the complexity of the process. As the CPU/PCB/chip rule sets grow, the compiler does better than the human, assuming it can efficiently factor in the rule set. Naturally, when it can't, you need the personal touch. Until it's expanded to embrace the new rules...like vector or GPGPU coding today. The other factor is of course architecture .. human time is usually better spent on the big picture items, architecture and algorithm. No sense hand coding for a 10% improvement in quality when the same time spent on design might yield 100%.
Not to mention the JIT factor -- code is compiled on the fly for the specific CPU and system in use. Custom DSP pipelines created based on the problem at hand. A complex math problem is targeted,on the fly, to somewhere between 32 and 4096 parallel processing elements. Compilers always win here.
4G models are dual core, faster core. 3G models are quad core, same A9 we all know. This has something to do with Samsung mysteriously leaving an LTE port off their quad A9 Exynos SOC. Whoops.
Based on their dropping of the iPad 1 with iOS 6, it looks like Apple's current policy is to drop support once a product is no longer actively sold. The iPad 1 is slightly faster than the iPhone 4, but last sold new in 2011...it was reduced and sold alongside the iPad 2. Gone for good shortly before the iPad 3 debuted. We'll see how well the iPhone 3GS and iPad 2 do with iOS 7 next year... both sold new in 2012, the 3GS is gone, the iPad 2 probably before the next iPad in March.
So, following this formula, the iPhone 5 goes to $100 in 2013 and gets iOS 7, goes free in 2014 and gets iOS 8, maybe even gets iOS 9 on the way out the door in 2015. Great if bought one just recently -- but you'll upgrade anyway before then. Sucks if you're a "free" buyer, you may see only a few months of support, depending on the iOS schedule.
And that's actually the wrong release sequence. Its clear Apple likes to do iOS major upgrades for iPhone, then maybe a tweak for the next iPad. But they seem to be cutting out the iPad from the upgrade sooner in its model life than the iPhone. Much sooner, as they only keep one older iPad around for new sales, so far anyway. But iPads don't follow telco subsidy cycles, so the user is likely to keep them longer, even hand them down, like PCs often are. Looks like Apple's plan here is to seriously de-value older iPads, dropping support as soon as they retire.
How about "you have great 1080p recording capability, and would like to actually use it"? Camcorder quality 1080p runs about 12GB per hour. A 64GB microSDXC card runs about $50. Adding that SD interface to your phone, maybe a $0.25 connector and another $0.10 in passive components -- the SD interface itself is free with your SOC. So why the hell not have the SD slot? Well, it does make charging $100+ for 16GB flash a bit harder for consumers to swallow...
Actually, Samsung's A15 SOC is only sampling now. In real S3s, you either get a dual-core Krait based SOC (Qualcomm's in-house more-or-less A15 class core) or a quad A9. Given the Asus tablet I'm typing this on soundly beats the iPhone 5 on Geekbench, one would expect Samsung's quad to do likewise. The Krait is more evenly matched, a bit slower than the expected A15 result, but clocked higher than Apple's likely to go, unless they finally trust their power/thermal management enough to run at the core's rated speed for peak loads.
The trade dress argument is about the only one that Apple really seems to have on Samsung's older products. It's a judgement call for sure, but there's little question that the iPhone is iconic. The fact that Samsung built very similar looking devices, modified Android to look more like iOS, even used similar packaging and wall wart.. it all adds up. And this can be a trade dress violation, even without any single patent or copyright violation.
It says nothing about Android having a general problem. And it has very nicely illustrated the problems with such patent cases being decided by a neophyte jury, or worse get, a self-proclaimed expert with other priorities. I think it's also been shown that Apple is writing patents specifically designed to confuse such juries in their favor. Patent authoring itself has been an ongoing evolution, to work the PTO and their thin basis fir software patenting as fast and hard as possible.
If my phone is in my pocket screen-out, it's going to be nearly impossible to establish an NFC connection in a crowded subway. The phone itself (plus the battery... on my phone, the NFC antenna is actually in the battery, so that it can be close to the back surface) is a pretty good shield.
Is that that best you can do? Possible faked or overclocked CPU benchmarks?
So.... dozens of Galaxy SIII benchmarks are presumed to have been faked and/or overclocked, even with those having performed some of those here, telling you "no, it's stock hardware". But a random unsourced one-of benchmark claiming to be an iPhone 5 is somehow gospel handed down from on high?
How about comparing default spec versus default spec? Those are pretty irrelevant if they were either falsified or on a device with an overclock because a "phone" is supposed to be usable in your pocket and have battery life measured in hours around at least a work day long rather than minutes.
It's similarly irrelevant to judge one unsupported posting.
On the other hand, if I were working at Apple and knew the iPhone 5 had crazy killer performance, I'd post a number higher than the current competition, and totally faked. To get them all going crazy, posting higher numbers, etc. Then, next week, when real iPhones 5 hit the market, we'd see all those much higher actual numbers from the iPhone, and victory would be mine. Mine, I tell you, mine!
Good thing the iPhone 5 isn't really that fast and I'm not working for Apple.
Any Android running Jellybean can do that just as well. Anything you can actually perceive as being faster, rather than measuring via some kind of benchmark, has very little to do with processor speed, and everything to do with user interface latency. In other words, it's a matter of tuning. Apple's been pretty good about that, though they still fail in places where they should be using multitasking/multiprocessing but aren't.
Android, on the other hand, didn't seem to spend much of an effort on some of the "finesse" issues in Android until fairly recently. Android Ice Cream Sandwich made a great deal of progress on deciding just what Android's supposed to look like (eg, the "magazine" UI that showed up first on Google's web site, rather than just a mish-mosh of Palm, WinCE, Blackberry, and iOS ideas), and improved latency issues nicely. The most recent release, Jellybean, has tweaked latency, eliminated deadlocks, etc. so that it's every bit as good as iOS.
The key here is that humans are just slow. If the OS isn't responding as fast as you need it, that's an algorithm problem, not a CPU speed problem. If you have any doubts, find an old Amiga and try AmigaOS on a 1980's CPU that's not fast enough to be considered as an I/O processor on a modern smartphone SOC. It'll make desktop Windows seem slow, within limits (eg, you're not going to get any H.264 videos playing). This fully illustrates my point... AmigaOS was a realtime OS, and the UI manager ran at a much higher priority than nearly anything else. So there was never any waiting on an app or anything else -- the user was the most important thing in the system. So it seems really, really fast, even on a slow processor. There's no hard reason any modern device, PC or portable, should have a "slow" UI.
The SIII is not using the A15, but Qualcomm's own Krait design. Krait is supposed to be about 32% faster than the A9, while the A15 is expected to run about 40% faster than the A9. There's some suggestion that the A15 was being designed for higher power use, but then again, both Samsung and TI have released (to developers) A15 core processors intended for smartphones, not even just tablets or high processor count servers.
According to Arnand (http://www.anandtech.com/show/6292/iphone-5-a6-not-a15-custom-core), the iPhone 5 doesn't use the A15 either. He suggests this high power idea as a reason, but it doesn't make the slightest bit of sense that ARM would walk even a little bit away from their primary business (mobile devices) by overpowering the A15, or that the chip makers would use a non-phone-suitable core in a phone-oriented chip. And of course, ARM themselves list the A15 as a processor for advanced smartphones. Ok, sure, Intel has a chip for smartphones too....
It's also the case that Apple's second chip company, Intrinsity, were known for making chips faster -- not low power. They have a bunch of tricks to get to higher speeds in older process technologies, at least some of which (like dynamic latches) aren't the right formula for a mobile device that might be clocked way down for power savings (dynamic latches establish minimum clock speeds -- static logic can clock all the way down to DC). Apple's using Samsung's 32nm process, and may move some things to TSMC's 28nm process (along with AMD,TI, nVidia, Qualcomm, Broadcomm, and the rest of the fabless world -- rumor has it Apple actually tried to buy all of TSMC's 28nm capacity for some unspecified period, and were told to go pound sand).
One interesting observation about this: ARM Holdings (the ARM company that does the basic design of ARM processors, MALI GPUs, etc) made $462M in gross income in 2011, and probably something more in 2012. But still -- could Apple or Samsung or Qualcomm or even TI match ARM's level of investment in CPU development? Sure they could... and ARM still gets paid when they do. Apple's got so much money and, in particular, if they just concentrate on the CPU, they might deliver substantially better CPU cores for their specific purposes than the other ARM licensees. It'll be interesting to see if any of the others, aside from Qualcomm of course, get into their own development.
Qualcomm's done their own designs for awhile. Their previous core, the Scorpion, was about 5% faster than the ARM Cortex A8, not as fast as the A9 (about 25% faster than the A8), thus most Qualcomm powered phones being clocked a bit faster than the A9 phones.
Anyway, if true about Apple's core, that make even more interesting, with ARM, Qualcomm, and Apple all making their own ARM cores. And Samsung still in a very strong position, being the only company in the ARM race with their own IC fab (well, TI does, but not for this class of parts). Oh, well, sure, Intel planning to fight it out as well on mobile... they also make their own chips.
Some are overclocked... easy to tell, since the benchmark supposedly calculates the actual processor speed. Not a big surprise -- most smartphone processors run below their rated speeds, in order to use a bit less power.
The big boost seems to be the SIII going from ICS to Jellybean. So far, every release of Android has improved the performance of the Dalvik VM. Not specifically an issue for native code (not sure if the benchmarks are done in the VM or the NDK; you'd want the NDK to best judge hardware performance), but it's going to affect parts of the system, even when using the NDK. You might also expect better performance if run with minimal active background services.
Not in a practical way, though. The iPad Don't-Call-Me-3 has twice the GPU performance (SGX 543MP4 vs. SGX 543MP2), but four times the pixels to paint.
The Samsung Galaxy SIII is faster than four Cray Y-MPs (or a couple of billion HP45 calculators)... at least if you're not too particularly about how your GFLOPS are served up. It's also got memory, unless you upgraded to the Y-MP M90. And uses quite a bit less power. That's 24 years for ya!
Nothing particularly useful or interesting about such observations, I suppose, unless you put the name "Apple" in the title. Or maybe just an Apple fan's way of dealing with the iPad 2 not being a terribly fast device, as mobile devices go. But watch out, all you 80s supercomputers.
My kids Nintendos have also been really abusive to our Cray 1, chasing it around the house, picking on it, calling it "pokey" and "turtle-boy", and just being bratty. Even the Nintendo 64, who's old enough to know better.
I'm cynical enough to believe they make these numbers up just to generate hype over a new product.
The inevitable "shortage" of 'phones over the month after launch? That could be deliberate, too, just to make people who got one feel special (and the ones who didn't to walk around muttering at the floor).
It's all deliberate. And well planned. Apple sold 1 million 4S last year via pre-orders. So naturally, they're going to sell more this year. But not so many that they'll have trouble selling 2.5M or 3M iPhones 5S or 6 next year.
Keep in mind, too, that selling a million smartphones in a 24 hour period is no kind of amazing accomplishment. Last year, around Christmastime, there were several days that topped one million Android devices, and I think Apple even hit two million once or twice. There are currently about 1.3 million Android activations every day.
In short, very credible numbers, given the month long hype circus that leads up to any iPhone pre-order day. And to the iPhone in-store day as well. Apple has generated this fear that you won't be able to get your new iPhone. They did that on purpose because it gets all sorts of free press, and just makes the iPhone seem a more attractive thing -- anything that's hard to get must be worthwhile, eh?
It's the sure sign a new model is coming when supplies of the old one start drying up.
Not really. It's a sign that a new model is coming when orders for the old model start shrinking. Apple keeps selling the old models -- they just rotate them down the price structure. And EVERY knows this. The iPhone 5 launch was the least well kept secret in technology, and that's exactly how Apple wants it. Rather than have the press talking about the new iPhone for a week, they're doing that for a month before the official availability. So people just stopping the 4S, because everyone knew the 5 was coming this month. If you didn't like the 5, you'd be guaranteed to get a 4S for $100 less right after the iPhone 5 announcement. Easy.
Though, an annoyance seems to be Apple never makes "enough" - they always sell out on the weekend and supplies are extremely tight. It's a marketing ploy, maybe (though each Apple store seems to get a few daily), but I guess when you're spending billions on making the things, you get cautious and don't blithely order tens of millions for release day and tie up money on inventory.
No, it's much more than that. Apple does their marketing homework -- they knows with certainty they'll see ten or twenty million iPhones this fall, and they have no concerns about tying up money in inventory -- they have $100+ in cash (though most of it outside the USA). In fact, they probably have a very precise estimate of sales, and they're working to deliver that much product.
This is all part of the marketing process that is an iPhone launch. They could have 5 or more million available for pre-order online. They don't. They have enough to beat last year's record enough to get the press reporting it. But not so much breaking that record next year should be a big problem. That's optimizing the press they get from the pre-release, just as they optimize the impact of the announcement. I mean, if Apple felt like it, they could have the announcement, then start the pre-sale that very day, maybe the store sale that day or the next. But they don't -- they give each of those events time to get reported, all that free publicity. And more, all that reinforcement in the iPhone users' minds that this is something special. No other new device release makes this kind of impact.
They also have to be careful about pre-selling too many online, because the next big event on the calendar is the in-store release. That's also going to break last year's record, and they're also going to sell out. iPhone fans know this, so they line up outside Apple stores across the country, and sure enough, it's another big media event. It gets reported, the lines, the eventual sellout, etc. Again, this isn't like any other "phone".
Magically, once the noise all dies down, though, ample supply of the iPhone 5 will appear. This is also part of the process.. having generated all this excitement, they really don't want to lose that momentum.
This is all part of the Apple iPhone Launch song and dance.
They know pretty accurately what they're going to pre-sell. They pre-sold 1M iPhones 4S last year because that's how many they made available for pre-order. This year, it was 2M iPhones 5. Apple's spent the last month building the anticipation: leaks, event announcements, etc. Even the wait from the announcement to the online ordering being enabled. And the fact they sell out in store every year -- all of this is well planned.
Given that only about 25% of Apple's profits ever enter this country, it seems fairly unlikely the iPhone will have an important effect on the real economy, whatever their effect on the GDP.
If you're a hardcore smartphone user, and run your battery down regularly, you'll use the replaceable battery. The simple fact is that no one's Li-ion battery will last two years of full discharges every day. The ability to replace the cell yourself is a win, even if you don't ever plan on carrying spare cells.
If, on the other hand, you keep your device topped off pretty well, maybe with a charger at work or whatever, the battery will probably last the life of the device.
I'd been holding out with my iPhone 4 for a while, waiting (like many others, I suspect) to see what Apple would wow us with for the iPhone 5. Needless to say, I wasn't that impressed, though to be honest, part of me really didn't expect to be, given that there are only so many innovations they could have come up with. What could they have done? An even bigger screen? NFC? A phone you could roll up?
That's kind of the point of the disappointment, isn't it -- no innovations. And that's precisely because an innovation is something unexpected, not just a safe, incremental improvement.
Ultimately, an innovation is a judgement. You do something risky. If it sells, that's an innovation. If not, it's probably on someone's worst products list for that year, and then forgotten. The original iPhone was an innovation, taken as a whole. Not really tech innovation, a business innovation -- Apple was doing things everyone in the phone business absolutely knew were impossible. They were going to sell a smartphone (well, originally, a feature phone) to consumers. As the most expensive phone on the planet, none-the-less. And they were dictating their terms to the carriers, rather than kowtow to them. A relatively big risk (tough even if the phone failed, they might have still launched a decent iPhone Touch).
But Apple's riding high now, with over 30% of the US smartphone market, and most of the profits. Every new iPhone has been just an incremental improvement, some hailed as the Jesus Phone, some written off too little of an improvement, all very, very successful. Apple has the formula down, and they're sticking to it. There's no room here for a gamble, and thus, no room for a significant innovation.
What are the big changes here? Faster CPU/GPU ... yeah, that happens pretty much every year, on every phone. Longer screen... pretty much a requirement. Every competitor has a 16:9-ish screen. They didn't try to the match the typical 1280x720 resolution. They did use a new LCD technology, leaving behind IPS for something new from Sharp that integrates the touchscreen in the display, transmits more light -- so, thinner (not as thin as OLED) and lower power. Good place for an improvement, but hardly noticeable to the average user. And then the connector -- maybe that's actually innovative, if it does anything useful that a plain old microUSB/MHL connector can't do. And also the main controversy.
Apple really never going to be the place to look for innovation in smartphones or tablets again -- they're too successful. They'd need a couple of model releases per year to even try out something very different, without risking the whole franchise. With the single model, no chance of risk talking. It's easier to watch all the Android devices life or die in the market, and adopt the innovations from those that rise to the top.
I have the Galaxy Nexus as well... I do believe there's a chip containing unicorn poop in there.. that's where the on-screen rainbows come from.
As well, the TI OMAP 4460 improved the overall system. It wasn't just dual ARM Cortex A9s, but dual memory bus controllers -- much the same deal you find on the average desktop PC. That's one key to keeping the processors (and GPU of course, don't forget all of these SOCs are "shared memory" devices, in the sense of PC graphics being shared) well fed. Some of the quad cores, like the nVidia Tegra 3, still use a single (though somewhat faster) memory bus. So when you have good cache hits, the Tegra 3 pulls ahead, when running memory bound, the 4460 will be faster.
In short, performance is all about the total system, not any given spec. The Tegra 3 ought to be significantly faster on 3D games. That's also what the iPhone has typically been optimized for -- Apple sells a crazy boatload of games, and they've pretty much made that a thing in the iTunes store now.
I think if you chose well, as we did with the GN, you're not going to see a huge upgrade opportunity in just a year. Sure, if someone gave me a free SIII I'd use it, but if I have to spend real money (and maybe a bit of it, if I want to keep my unlimited plan on Verizon), I'll wait for a substantial upgrade. Hopefully one that tunes battery life in as well.
And the Java apps will run on x86 Android... which is coming to the mainstream. At least according to Intel.
It's not just the cores, but what they do. The four Cortex A9s in your Nexus 7 may or may not be faster than the dual A15 in the iPhone... looks pretty close. The A15 was expected to be about 40% faster than A9, but this the first time one's actually been benchmarked "in the wild"... assuming, as most think, it is an A15.
They also have either a four core GPU based on the PowerVR SGX543MP4, or something new. This is already faster than the Tegra 3's GPU... much as you find, on desktop processors, you need several AMD GPU cores to match a single nVidia core.
Well, they used to be "Apple" in the A4 days -- the A4 was really just a customized version of one of the Exynos chips. Apple could retroactively claim to have some design work on it, for those who want to be pedantic about, though. Samsung worked with Intrinsity, Inc. on the design of the Hummingbird processor. Like every A8 over 600MHz or so, this was a customized version of the Cortex A8. Apple later bought Intrinsity.
These days, they are doing the SOC design, and so far, Samsung's doing all the fab work, though there have been rumors that Apple's been shopping for a second source, and may be going to TSMC. Qualcomm uses TSMC, and in fact, has been frustrated with the output from TSMC's 28nm process. Enough so they tried to buy exclusive access to that from about a billion. Apple reportedly tried the same thing; TSMC said "no" in both cases. Which is in their long-term interest -- an independent fab can't play favorites and stay independent. Qualcomm's moved a bunch of work to Global Foundries (the AMD spinoff), and reportedly also have talked to UMC and Samsung about making the Snapdragon S4. nVidia, also a long time TSMC customer, seems to be doing similarly, looking at UMC and Global Foundries for increased 28nm support.
In short, business as usual. Many of the big names -- even AMD these days -- design the chips, use the big chip foundaries to actually make them.
The dual core versions run the Qualcomm Krait processor, which is about 30% faster per clock cycle than the ARM Cortex A9. So head to head tests will vary. And of course, if your benchmark isn't multi-threaded, you're just testing single core performance. That's at least one way the actual effect of additional cores can be made confusing.
Steve Jobs loved a thin phone -- so iPhones are thin. That's kind of their thing.
They have also been fairly fragile, and yeah, you can see some crazy cases, like the Otter Box, that pretty much erase any purpose to the thin design. I've seen the same thing with laptops... a guy I used to work with had a super think MacBook Pro. Which was always kept in some kind of Pelican case, making it the most bulky laptop ever seen.
Point being, a slightly thicker phone that doesn't need a case to survive works out much thinner than a phone that requires a case. It's not just Apple... the Galaxy Nexus requires a case. The plastic used wouldn't be rugged enough for a kids' toy, and the glass scratches fairly easily, which is close to crime these days, given that's one of the problems that seems to be pretty completely solved by Gorilla Glass.