Exactly see this is from S6 5.1" Hope they don't delete this post or my other one.
Uh, why did you not pick scores from Geekbench's Results Browser? And you accuse AI of cherry picking?
Edited:
Here are the actual scores for the S6 and S6 Edge from Geekbench.
S6: 1207/4109
S6 Edge: 1317/4405
A far cry from your posted result. Geekbench averages out results as opposed to many individuals who post "loaded" results (both faster or slower than usual).
And here are the iPhone 6S scores:
6S: 2487/4327
6S Plus: 2460/4315
The real difference is Apple devices are remarkably consistent in their results. You never see inflated scores or unusually low scores.
Finally, if you want to know how much of a failure the Note 5 is, just go to Primate Labs and their Geekbench browser. The Note 5 doesn't even show up in the Android chart. Geekbench needs a minimum number of tests before a device shows up in their charts, which is why a new iPhone or iPad usually takes a couple days to hit the iOS chart. Yet almost 2 months after release the Note 5 hasn't had enough results/tests to even appear. I find that hilarious.
Can we already forget the A8 (64) chip? Have we already exploited the full potentiel of that old technologie yet? Half a second gain in speed is ok, i guess, lol! But there is a price to it, the new phone don't come cheap...
Exactly see this is from S6 5.1" Hope they don't delete this post or my other one.
Hey buddy boy, the S6 almost fracking burns up and then trottles and then eats the battery when pushed in real apps (not one that runs a few minutes), and that score means nothing in real life for even short running apps as many have discovered as almost no apps, and even the OS, can't fully use all the cores efficiently in apps people actually use.
That's the same reason AMD gets creamed in the market by Intel, even if AMD has way more cores for the price than Intel.
They're more power hungry and the IPC is much lower.
Cherry picking scores also is helping you supposed point buddy. Samsung, and all other Android phones, are getting their ASS KICKED.
Also, the supposed leaked S7 benchmark (A phone which coming out in 7 months) comes from some lab chip running at 2.5GHz... 45% higher clock rate than the Iphone just to be close to the Iphone's single core score! If that score is not just a pure invention (as it came out strangely enough the day before the Iphone release).
And Androites are all clapping their hands at that like it is their salvation : it is not. Their chip just plain suck (in plain language...)
You know what that would do in a cell phone... An even worse crapola as is done in the S6.
A few hours of battery life, just to be able to play in the same neighborhood as the Iphone.
is DED ok. There was at least 100 words less than his usual opus. There is even less sign of his usual foaming at the mouth rants! If this is a newer leaner and more focused writing style of DED then keep it up!
This op-ed is full of errors and exaggeration as usual. In his last cpu comparison "op-ed", he compared the performance of the A8 and the Exynos 7420 -- and, of course, DED's Exynos benchmark was based on the power-saving A53 quad.
In this particular article, take, for instance, his claim that :
Quote:
... In July, we noted that GameBench reported--after extensive testing of last year's iPhone 6 against this year's flagships from Samsung, HTC and Motorola--that in games, Apple's iOS devices were not only performance competitive but simply faster and delivered better looking graphics, despite packing less RAM and fewer "cores." ...
now, if you know anything about Apple's GPU strategy, they have long outsourced their GPU design to Imgtech and the last three or fours A SOCs had no fewer than 4+ GPU cores -- the A6 had 3 GPU, A7 4, A8 4. Sure, Imgtech's GPUs are known to perform better than ARM's MALIs, and it apparently helps the iPhone perform better in graph intensive games, but it isn't certainly because of Apple's dual cores. And the impulse behind this multi-core arch, on both mobile and non-mobile devices, has been, at least since the mid 2000's, driven by power / heat, as many pointed out here.
This op-ed is full of errors and exaggeration as usual. In his last cpu comparison "op-ed", he compared the performance of the A8 and the Exynos 7420 -- and, of course, the 7420 was benchmarked in "Power-Saving" mode.
In this particular article, take, for instance, his claim that :
now, if you know anything about Apple's GPU strategy, they have long outsourced their GPU design to Imgtech and the last three or fours A SOCs had no fewer than 4+ GPU cores -- the A6 had 3 GPU, A7 4, A8 4. Sure, Imgtech's GPUs are known to perform better than ARM's MALIs, and it apparently helps the iPhone perform better in graph intensive games, but it isn't certainly because of Apple's dual cores. And the impulse behind this multi-core arch, on both mobile and non-mobile devices, has been, at least since the mid 2000's, driven by power / heat, as many pointed out here.
Sounds like someone is upset that Apple is far ahead of Samsung in processor design.
No, wait. Samsung doesn't design processors, so we really can't say Apple is ahead. Maybe in a year or two Samsung will finally create their own processor and then we can compare. Until then, Apples custom processor cores are far ahead of Samsung's rubber stamping of someone else's designs.
Exactly see this is from S6 5.1" Hope they don't delete this post or my other one.
Hey buddy boy, the S6 almost fracking burns up and then trottles and then eats the battery when pushed in real apps (not one that runs a few minutes), and that score means nothing in real life for even short running apps as many have discovered as almost no apps, and even the OS, can't fully use all the cores efficiently in apps people actually use.
That's the same reason AMD gets creamed in the market by Intel, even if AMD has way more cores for the price than Intel.
They're more power hungry and the IPC is much lower.
Cherry picking scores also is helping you supposed point buddy. Samsung, and all other Android phones, are getting their ASS KICKED.
Also, the supposed leaked S7 benchmark (A phone which coming out in 7 months) comes from some lab chip running at 2.5GHz... 45% higher clock rate than the Iphone just to be close to the Iphone's single core score! If that score is not just a pure invention (as it came out strangely enough the day before the Iphone release).
And Androites are all clapping their hands at that like it is their salvation : it is not. Their chip just plain suck (in plain language...)
You know what that would do in a cell phone... An even worse crapola as is done in the S6.
A few hours of battery life, just to be able to play in the same neighborhood as the Iphone.
I agree, and I think Apple is doing what they do best, and that is "hitting the puck where the competition can't even skate to." Apple's Ax chip is efficient as hell. Which even got better during iterations. Then Apple set out to make the entire hardware more energy efficient and wring another additional hour out of battery life... which keeps the Ax from warming up even as much as it did... further pushing the puck where no cute little android can go.
3D touch is a great addition to the UI - but while making the UI better Apple was/is making it harder for the competition to copy and match operatibility... zoom, the puck is even harder to reach... Everything Apple does, if you look at through the lens of competition, is to further distance Apple's products from the commodity market, leaving everyone else to fight it out down there where the profit is tight, if not completely non-existent.
In trying to catch up to Apple, Samsung is forced to do what you describe above. Like a Volkswagon diesel, Samsung's products only look good during non-world (i.e. test) conditions. As we move toward winter and another wave of Sammy shit heading toward the chute, the best they can hope for is to advertise the forthcoming offerings as "pocket warmers."
This op-ed is full of errors and exaggeration as usual. In his last cpu comparison "op-ed", he compared the performance of the A8 and the Exynos 7420 -- and, of course, DED's Exynos benchmark was based on the power-saving A53 quad.
DED wrote an Op-ed piece, not a research paper. The "OP" part of that means "OPINION." So, no footnotes and clarifying statements needed.
These are all "Power-Saving CPUs, if you will. Not one is designed for a desktop computer, so suck an egg over the Exynos power-saving A53 quad.
Apple R & D is doing many things to make the competition's products suck air, some even the competition could do if they were so inspired. In doing so, Apple is leveraging their ability to better fit hardware to iOS software. That's an advantage Apple has, and they are wisely exploiting it. Anyone could go and write software like Metal, to optimize the gap between (lets call it machine language) and a compiled high-level language. But no one has, and if they start today they will still be years behind Apple,,,meanwhile Apple will continue to move the mileposts, so there's little hope of any competitors closing the gap between Apple and the also-rans.
That is a bit irrelevant, if you talk about real-world. Apart from power consumption consideration, there is also a performance consideration. And the problem with that is that it is FAR easier to write an app that will run on a single core, comparing to a multi-core setup. In fact, ANY task can be executed on a single core-CPU at 100% efficiency (CPU bound task, that is), but it is not always the case that you can execute the same task just as efficiently on 4- or 8- core setup. Some tasks aren't easily parallellizable and sometimes it is simply impossible (even theoretically) to make them parallel. That is why it is always good to have LESS but faster cores, and not more (but weaker) cores. The more cores a CPU has, the less possibly you will have that you get 100% efficiency out of it.
That is why a two-core A9 with a single core (2500) result is MUCH better then 8 core S6 edge that has a single-core result of 1300.
A single threaded app will use 56% of full A9 performance, and 100% - with two threads.
In comparison, Galaxy S6 Edge CPU running a single threaded app will reach only 30% of the theoretical maximum (1300), while two-threaded app will access 50-60% (2200-2600, depending on what kind if interaction happens between cores) of its CPU potential.
So, if you compare apps that at maximum can be paralleled only to 2 threads, then A9 will always be 2 times faster no matter what and one needs to have an app running tasks on 4 threads in order to be able to stay at the same performance as A9.
It is true that performance wise it is better to have less cores with higher performance, but claiming that a single task can run at a 100% efficiency on a single core, is of course bullshit. You apparently forget that there is always an OS and that there background processes that are running and both need the CPU as well. On a smartphone there are a lot of such processes: network, screen and sound processing, ... These tasks can easily be distributed on different cores.
The efficiency of the OS is also a very important factor. I assume that iOS is more efficient than Android, taking into account that a lot of Android tasks run on the Dalvik virtual machine, which must have some overhead.
But also the efficiency of the apps are important. In terms of performance Safari on iOS has not improved in recent years at the same pace as Safari on OSX or as Google Chrome. And the browser app is one of the apps I use the most. So when surfing, my Android devices feel snappier than my iOS devices, despite the better single core architecture of the Apple devices
" You apparently forget that there is always an OS and that there background processes that are running and both need the CPU as well."
"a single task can run at a 100% efficiency on a single core, is of course bullshit"
1) Nope. You don't understand the concept of time-sharing on a single-core CPU. Inside of its own CPU time slice, an app has full 100% speed of a CPU to itself. There is no OS task running at that time because there is no place for it to do it. I was talking about CPU-time measurements and not wall-time measurements. So, for a task that runs during a certain time-slice - it will be 100% CPU allotted to that task. And by measuring CPU-time and not wall-time you can get a pure single app result on a CPU as if there is no addition load present. In fact, you can test the same app on a free-of-load CPU and on a CPU that runs 3d-rendering. CPU-time for your app will be the same in both cases.
Background tasks are very limited on mobile platforms in what they can do. I have come across a statement(I think it was in Android docs for devs) that Google says that background activity is limited to about 10% of total CPU use. Neither Android nor iOS allows CPU-heavy or GPU-heavy stuff to be computed in the background.
2) "The efficiency of the OS is also a very important factor"
Irrelevant if you measure CPU-time, again..
Yes, system dispatches tasks according to how it divides CPU between processes (managing of queues), giving apps some processor time by using some time-slicing algorithm, but in the context of a single TASK, a single task isn't even aware of ANY other tasks that will be running next time slices on that very CPU. In short, a concept of CPU-time allows to discount that!
3a) "But also the efficiency of the apps are important. I"
And sadly this is even more irrelevant to what I said initially. It runs regardless of how it is efficient. If Java compiler/interpreter/magical parser/whatever is less efficient, so it means it will be the same for ALL programs on that platform because I am talking about a tool that is used when creating ANY app for that particular platform.
If LLVM compiles binaries with a given efficiency, then that is what you get, and if you compare it to a less efficient Dalvik - too bad for Dalvik, but this comparison is fair in the sense that end user has to use one of those indirectly and if he chose a less efficient Dalvik - too bad for him too, but that Google solution will not only affect benchmarks, but also EVERY app that dude is running on his Android - hence fairness is preserved.
3b) However, if you are talking about individual programmer's ability to write an efficient code in the context of the same OS, than that is even less important because benchmarks or any other apps for that matter may be written by different teams of programmers. But all those teams theoretically should be on-par with each other. If not, that is by random chance, and by looking at others benchmarks results we can see that still holds or whether that difference (in quality of developers) is smoothed out because a single bad developer is a single sample in statistics, while 50% of bad developers is a trend and hence should participate in a fair comparison too...
4) I don't like "snappiness" criterion simply because it is hard to understand what is described. Is it the speed of a user interaction with an app UI (animation speed), or CPU-wise speed of the app that performs certain tasks regardless of the UI speed? As far as I know, Safary animations are much slower on iOS, when compared to generally faster and jerkier animations on Android. Since you didn't test it properly and you simply reported a "feel" I would need a really big grain of salt with your "So when surfing, my Android devices feel snappier than my iOS devices, despite the better single core architecture of the Apple devices ". Without proper blind study it is really hard to deal with a human bias.
Just ran a quick geekbench on my S6 edge on latest version and it scores as below. suffice to say I don't see most mobile tasks coming close to needing these numbers apart from running intensive games. Interested to knoe what's better for battery life though. Dropping down to a smaller and lower powered core when needed or dropping down to a larger more powerful core?
While you techie folks are arguing about benchmarks.... there's others out there actually USING the new iPhone 6s and 6s+ to create some stunning pictures and films.
[URL= the Scenes First Documentary Filmed on New iPhone in 4k[/URL] (be sure to choose HD Quality under the gear settings for 2160p HD)
The film itself... naturally also in 4K - [URL= Painter of Jalouzi[/URL]
In terms of performance Safari on iOS has not improved in recent years at the same pace as Safari on OSX or as Google Chrome. And the browser app is one of the apps I use the most. So when surfing, my Android devices feel snappier than my iOS devices, despite the better single core architecture of the Apple devices
Curious, since in browser benchmarks (Kraken, Octane, Sunspider, Basemark) iOS devices routinely beat Android devices.
Curious, since in browser benchmarks (Kraken, Octane, Sunspider, Basemark) iOS devices routinely beat Android devices.
It doesn't merely beat them, it creams them. The S6 has a notorious rep for gobbling the battery, especially when using processor intensive app (like Games, video processing, etc). You hear about that all the time in Android forums (but nobody will ever admit in an Apple forum...).
The key in a processor is run fast then sleep using as little power as you can, transition from one state to another as efficiently as you can, and use the cores as efficiently as you can. That's better done by using powerful wide little cores.
As for what's "better", since to run an App as fast as Apple's A9, those multi core monsters would have to wake up the cores when doing something intense (by groups of 4!) and schedule crap accross all of them, even if its not suited to it, and then wait for those threads to end to put the CPU sections (big-little cores) to sleep, I'm doubting the power efficiency of all this very very much.
BTW, Intel seems to have gone the Apple way in their low power devices, not the multi-core way of other Android makers.
Multi-core does make sense if you have an energy budget to feed them and the ability to dissipate the heat (like in a Desktop or a large laptop).
When running relatively less power intensive tasks those multicore beast would perform relatively OK, but still slower than the A9, but in intensive tasks it would gobble up battery, heat and throttle. That's what happens in real life right now.
Of course the S7 will launch next February (as rumored) with a Snapdragon 820 that will edge the 6s in benchmarks. And the world continues to turn.
The Galaxy S6 Edge is so stupid fast as is, I can't even imagine why people care so much about epeen numbers for their phone. There's a reason why we are seeing $400 unlocked phones that are reviewing well. The chips are satisfactory enough.
iPhones are grease lightening for more reasons than their chips (they're great, no question). But they are still sitting on a sub 1080p resolution for the standard model. The Galaxy S6 Edge is 1440p. So while both chips are benchmarking the same outside of single core (Anandtech has a great article on squashing multicore myths with Android), the iPhone is always going to be the performance champion.
I don't even think the Android community doubts that though. At least the sites I frequent. Some of you guys need new Android friends if that's your perception. Most Android people I know are pumped about the Moto X Pure, or the Nexus 5x, both with weaker Snapdragon 808's but are sub $400 unlocked with major carrier support out of the box. It's the age of "Corolla phone" in the Android community, not the Mercedes.
My Samsung Exynos 7420 on s6 edge delivers a punch, multicore score of 4958 to apple A9 chip multicore score 4423 when I run geekbench3 on my device. Making a headline like that shows no integrity when you don't include the two top samsung phones the galaxy s6 edge and note 5. These samsung devices use 4 gb of faster ram. Based on the results shown here, apple did have a one, two punch.... unfortunately the top two fighters from Samsung didn't show up to the fight or were intentionally removed from the results. Next time please don't bench..... errr I mean do geekbench3 the Samsung Galaxy s6 edge and Samsung Note 5.
"Multicore performance on higher end Android devices is often far higher than their listed single core performance, in part because Samsung and Qualcomm have relied on adding multiple cores (four or even eight) to support the slower performance of smaller, individual cores. "
I laughed here. "more cores to support slower performance"? yeah ok, im sorry, i have an iphone, but appleinsider is seriously biased
Even if a Samsung whatever matches or bests the latest iPhone on Geek-bench scores it is irrelevant, it is still running a crappy OS in a crappy eco system.
really? Android is a crappy system? sure android might be bloated but thats because Google made the software layer not the heart of it. At the heart of it is Linux, and linux is one the best operating systems out there.
Comments
Exactly see this is from S6 5.1" Hope they don't delete this post or my other one.
Last post was for @Chrispoe
Uh, why did you not pick scores from Geekbench's Results Browser? And you accuse AI of cherry picking?
Edited:
Here are the actual scores for the S6 and S6 Edge from Geekbench.
S6: 1207/4109
S6 Edge: 1317/4405
A far cry from your posted result. Geekbench averages out results as opposed to many individuals who post "loaded" results (both faster or slower than usual).
And here are the iPhone 6S scores:
6S: 2487/4327
6S Plus: 2460/4315
The real difference is Apple devices are remarkably consistent in their results. You never see inflated scores or unusually low scores.
Finally, if you want to know how much of a failure the Note 5 is, just go to Primate Labs and their Geekbench browser. The Note 5 doesn't even show up in the Android chart. Geekbench needs a minimum number of tests before a device shows up in their charts, which is why a new iPhone or iPad usually takes a couple days to hit the iOS chart. Yet almost 2 months after release the Note 5 hasn't had enough results/tests to even appear. I find that hilarious.
Uh, why did you not pick scores from Geekbench's Results Browser? And you accuse AI of cherry picking?
I second that, plus if you go to post #21 on this thread I attached my results with my 6S plus and it destroys the S6.
Exactly see this is from S6 5.1" Hope they don't delete this post or my other one.
Hey buddy boy, the S6 almost fracking burns up and then trottles and then eats the battery when pushed in real apps (not one that runs a few minutes), and that score means nothing in real life for even short running apps as many have discovered as almost no apps, and even the OS, can't fully use all the cores efficiently in apps people actually use.
That's the same reason AMD gets creamed in the market by Intel, even if AMD has way more cores for the price than Intel.
They're more power hungry and the IPC is much lower.
Cherry picking scores also is helping you supposed point buddy. Samsung, and all other Android phones, are getting their ASS KICKED.
Also, the supposed leaked S7 benchmark (A phone which coming out in 7 months) comes from some lab chip running at 2.5GHz... 45% higher clock rate than the Iphone just to be close to the Iphone's single core score! If that score is not just a pure invention (as it came out strangely enough the day before the Iphone release).
And Androites are all clapping their hands at that like it is their salvation : it is not. Their chip just plain suck (in plain language...)
You know what that would do in a cell phone... An even worse crapola as is done in the S6.
A few hours of battery life, just to be able to play in the same neighborhood as the Iphone.
is DED ok. There was at least 100 words less than his usual opus. There is even less sign of his usual foaming at the mouth rants! If this is a newer leaner and more focused writing style of DED then keep it up!
This op-ed is full of errors and exaggeration as usual. In his last cpu comparison "op-ed", he compared the performance of the A8 and the Exynos 7420 -- and, of course, DED's Exynos benchmark was based on the power-saving A53 quad.
In this particular article, take, for instance, his claim that :
... In July, we noted that GameBench reported--after extensive testing of last year's iPhone 6 against this year's flagships from Samsung, HTC and Motorola--that in games, Apple's iOS devices were not only performance competitive but simply faster and delivered better looking graphics, despite packing less RAM and fewer "cores." ...
now, if you know anything about Apple's GPU strategy, they have long outsourced their GPU design to Imgtech and the last three or fours A SOCs had no fewer than 4+ GPU cores -- the A6 had 3 GPU, A7 4, A8 4. Sure, Imgtech's GPUs are known to perform better than ARM's MALIs, and it apparently helps the iPhone perform better in graph intensive games, but it isn't certainly because of Apple's dual cores. And the impulse behind this multi-core arch, on both mobile and non-mobile devices, has been, at least since the mid 2000's, driven by power / heat, as many pointed out here.
Sounds like someone is upset that Apple is far ahead of Samsung in processor design.
No, wait. Samsung doesn't design processors, so we really can't say Apple is ahead. Maybe in a year or two Samsung will finally create their own processor and then we can compare. Until then, Apples custom processor cores are far ahead of Samsung's rubber stamping of someone else's designs.
Exactly see this is from S6 5.1" Hope they don't delete this post or my other one.
Hey buddy boy, the S6 almost fracking burns up and then trottles and then eats the battery when pushed in real apps (not one that runs a few minutes), and that score means nothing in real life for even short running apps as many have discovered as almost no apps, and even the OS, can't fully use all the cores efficiently in apps people actually use.
That's the same reason AMD gets creamed in the market by Intel, even if AMD has way more cores for the price than Intel.
They're more power hungry and the IPC is much lower.
Cherry picking scores also is helping you supposed point buddy. Samsung, and all other Android phones, are getting their ASS KICKED.
Also, the supposed leaked S7 benchmark (A phone which coming out in 7 months) comes from some lab chip running at 2.5GHz... 45% higher clock rate than the Iphone just to be close to the Iphone's single core score! If that score is not just a pure invention (as it came out strangely enough the day before the Iphone release).
And Androites are all clapping their hands at that like it is their salvation : it is not. Their chip just plain suck (in plain language...)
You know what that would do in a cell phone... An even worse crapola as is done in the S6.
A few hours of battery life, just to be able to play in the same neighborhood as the Iphone.
I agree, and I think Apple is doing what they do best, and that is "hitting the puck where the competition can't even skate to." Apple's Ax chip is efficient as hell. Which even got better during iterations. Then Apple set out to make the entire hardware more energy efficient and wring another additional hour out of battery life... which keeps the Ax from warming up even as much as it did... further pushing the puck where no cute little android can go.
3D touch is a great addition to the UI - but while making the UI better Apple was/is making it harder for the competition to copy and match operatibility... zoom, the puck is even harder to reach... Everything Apple does, if you look at through the lens of competition, is to further distance Apple's products from the commodity market, leaving everyone else to fight it out down there where the profit is tight, if not completely non-existent.
In trying to catch up to Apple, Samsung is forced to do what you describe above. Like a Volkswagon diesel, Samsung's products only look good during non-world (i.e. test) conditions. As we move toward winter and another wave of Sammy shit heading toward the chute, the best they can hope for is to advertise the forthcoming offerings as "pocket warmers."
This op-ed is full of errors and exaggeration as usual. In his last cpu comparison "op-ed", he compared the performance of the A8 and the Exynos 7420 -- and, of course, DED's Exynos benchmark was based on the power-saving A53 quad.
DED wrote an Op-ed piece, not a research paper. The "OP" part of that means "OPINION." So, no footnotes and clarifying statements needed.
These are all "Power-Saving CPUs, if you will. Not one is designed for a desktop computer, so suck an egg over the Exynos power-saving A53 quad.
Apple R & D is doing many things to make the competition's products suck air, some even the competition could do if they were so inspired. In doing so, Apple is leveraging their ability to better fit hardware to iOS software. That's an advantage Apple has, and they are wisely exploiting it. Anyone could go and write software like Metal, to optimize the gap between (lets call it machine language) and a compiled high-level language. But no one has, and if they start today they will still be years behind Apple,,,meanwhile Apple will continue to move the mileposts, so there's little hope of any competitors closing the gap between Apple and the also-rans.
"multi-threading than initially expected"
That is a bit irrelevant, if you talk about real-world. Apart from power consumption consideration, there is also a performance consideration. And the problem with that is that it is FAR easier to write an app that will run on a single core, comparing to a multi-core setup. In fact, ANY task can be executed on a single core-CPU at 100% efficiency (CPU bound task, that is), but it is not always the case that you can execute the same task just as efficiently on 4- or 8- core setup. Some tasks aren't easily parallellizable and sometimes it is simply impossible (even theoretically) to make them parallel. That is why it is always good to have LESS but faster cores, and not more (but weaker) cores. The more cores a CPU has, the less possibly you will have that you get 100% efficiency out of it.
That is why a two-core A9 with a single core (2500) result is MUCH better then 8 core S6 edge that has a single-core result of 1300.
A single threaded app will use 56% of full A9 performance, and 100% - with two threads.
In comparison, Galaxy S6 Edge CPU running a single threaded app will reach only 30% of the theoretical maximum (1300), while two-threaded app will access 50-60% (2200-2600, depending on what kind if interaction happens between cores) of its CPU potential.
So, if you compare apps that at maximum can be paralleled only to 2 threads, then A9 will always be 2 times faster no matter what and one needs to have an app running tasks on 4 threads in order to be able to stay at the same performance as A9.
It is true that performance wise it is better to have less cores with higher performance, but claiming that a single task can run at a 100% efficiency on a single core, is of course bullshit. You apparently forget that there is always an OS and that there background processes that are running and both need the CPU as well. On a smartphone there are a lot of such processes: network, screen and sound processing, ... These tasks can easily be distributed on different cores.
The efficiency of the OS is also a very important factor. I assume that iOS is more efficient than Android, taking into account that a lot of Android tasks run on the Dalvik virtual machine, which must have some overhead.
But also the efficiency of the apps are important. In terms of performance Safari on iOS has not improved in recent years at the same pace as Safari on OSX or as Google Chrome. And the browser app is one of the apps I use the most. So when surfing, my Android devices feel snappier than my iOS devices, despite the better single core architecture of the Apple devices
" You apparently forget that there is always an OS and that there background processes that are running and both need the CPU as well."
"a single task can run at a 100% efficiency on a single core, is of course bullshit"
1) Nope. You don't understand the concept of time-sharing on a single-core CPU. Inside of its own CPU time slice, an app has full 100% speed of a CPU to itself. There is no OS task running at that time because there is no place for it to do it. I was talking about CPU-time measurements and not wall-time measurements. So, for a task that runs during a certain time-slice - it will be 100% CPU allotted to that task. And by measuring CPU-time and not wall-time you can get a pure single app result on a CPU as if there is no addition load present. In fact, you can test the same app on a free-of-load CPU and on a CPU that runs 3d-rendering. CPU-time for your app will be the same in both cases.
Background tasks are very limited on mobile platforms in what they can do. I have come across a statement(I think it was in Android docs for devs) that Google says that background activity is limited to about 10% of total CPU use. Neither Android nor iOS allows CPU-heavy or GPU-heavy stuff to be computed in the background.
2) "The efficiency of the OS is also a very important factor"
Irrelevant if you measure CPU-time, again..
Yes, system dispatches tasks according to how it divides CPU between processes (managing of queues), giving apps some processor time by using some time-slicing algorithm, but in the context of a single TASK, a single task isn't even aware of ANY other tasks that will be running next time slices on that very CPU. In short, a concept of CPU-time allows to discount that!
3a) "But also the efficiency of the apps are important. I"
And sadly this is even more irrelevant to what I said initially. It runs regardless of how it is efficient. If Java compiler/interpreter/magical parser/whatever is less efficient, so it means it will be the same for ALL programs on that platform because I am talking about a tool that is used when creating ANY app for that particular platform.
If LLVM compiles binaries with a given efficiency, then that is what you get, and if you compare it to a less efficient Dalvik - too bad for Dalvik, but this comparison is fair in the sense that end user has to use one of those indirectly and if he chose a less efficient Dalvik - too bad for him too, but that Google solution will not only affect benchmarks, but also EVERY app that dude is running on his Android - hence fairness is preserved.
3b) However, if you are talking about individual programmer's ability to write an efficient code in the context of the same OS, than that is even less important because benchmarks or any other apps for that matter may be written by different teams of programmers. But all those teams theoretically should be on-par with each other. If not, that is by random chance, and by looking at others benchmarks results we can see that still holds or whether that difference (in quality of developers) is smoothed out because a single bad developer is a single sample in statistics, while 50% of bad developers is a trend and hence should participate in a fair comparison too...
4) I don't like "snappiness" criterion simply because it is hard to understand what is described. Is it the speed of a user interaction with an app UI (animation speed), or CPU-wise speed of the app that performs certain tasks regardless of the UI speed? As far as I know, Safary animations are much slower on iOS, when compared to generally faster and jerkier animations on Android. Since you didn't test it properly and you simply reported a "feel" I would need a really big grain of salt with your "So when surfing, my Android devices feel snappier than my iOS devices, despite the better single core architecture of the Apple devices ". Without proper blind study it is really hard to deal with a human bias.
[IMG]http://forums.appleinsider.com/content/type/61/id/63417/width/200/height/400[/IMG]
[URL=
The film itself... naturally also in 4K - [URL=
Enjoy!
[VIDEO]
[VIDEO]
In terms of performance Safari on iOS has not improved in recent years at the same pace as Safari on OSX or as Google Chrome. And the browser app is one of the apps I use the most. So when surfing, my Android devices feel snappier than my iOS devices, despite the better single core architecture of the Apple devices
Curious, since in browser benchmarks (Kraken, Octane, Sunspider, Basemark) iOS devices routinely beat Android devices.
Curious, since in browser benchmarks (Kraken, Octane, Sunspider, Basemark) iOS devices routinely beat Android devices.
It doesn't merely beat them, it creams them. The S6 has a notorious rep for gobbling the battery, especially when using processor intensive app (like Games, video processing, etc). You hear about that all the time in Android forums (but nobody will ever admit in an Apple forum...).
The key in a processor is run fast then sleep using as little power as you can, transition from one state to another as efficiently as you can, and use the cores as efficiently as you can. That's better done by using powerful wide little cores.
As for what's "better", since to run an App as fast as Apple's A9, those multi core monsters would have to wake up the cores when doing something intense (by groups of 4!) and schedule crap accross all of them, even if its not suited to it, and then wait for those threads to end to put the CPU sections (big-little cores) to sleep, I'm doubting the power efficiency of all this very very much.
BTW, Intel seems to have gone the Apple way in their low power devices, not the multi-core way of other Android makers.
Multi-core does make sense if you have an energy budget to feed them and the ability to dissipate the heat (like in a Desktop or a large laptop).
When running relatively less power intensive tasks those multicore beast would perform relatively OK, but still slower than the A9, but in intensive tasks it would gobble up battery, heat and throttle. That's what happens in real life right now.
Of course the S7 will launch next February (as rumored) with a Snapdragon 820 that will edge the 6s in benchmarks. And the world continues to turn.
The Galaxy S6 Edge is so stupid fast as is, I can't even imagine why people care so much about epeen numbers for their phone. There's a reason why we are seeing $400 unlocked phones that are reviewing well. The chips are satisfactory enough.
iPhones are grease lightening for more reasons than their chips (they're great, no question). But they are still sitting on a sub 1080p resolution for the standard model. The Galaxy S6 Edge is 1440p. So while both chips are benchmarking the same outside of single core (Anandtech has a great article on squashing multicore myths with Android), the iPhone is always going to be the performance champion.
I don't even think the Android community doubts that though. At least the sites I frequent. Some of you guys need new Android friends if that's your perception. Most Android people I know are pumped about the Moto X Pure, or the Nexus 5x, both with weaker Snapdragon 808's but are sub $400 unlocked with major carrier support out of the box. It's the age of "Corolla phone" in the Android community, not the Mercedes.
I laughed here. "more cores to support slower performance"? yeah ok, im sorry, i have an iphone, but appleinsider is seriously biased
Even if a Samsung whatever matches or bests the latest iPhone on Geek-bench scores it is irrelevant, it is still running a crappy OS in a crappy eco system.
really? Android is a crappy system? sure android might be bloated but thats because Google made the software layer not the heart of it. At the heart of it is Linux, and linux is one the best operating systems out there.