The A7 isn't the piece of engineering you think it to be...

[varunsanthanam]

If you read the OP you can easily see that the graphs are to compare the iphone 5 and 5c (which use the same A6) across ios 6 and 7 (in the case of the iphone 5) and the subsequent massive performance gain.



http://www.anandtech.com/show/7399/...review-iris-pro-driving-an-accurate-display/3

Lets look at the imac playing bioshock on windows 8 vs 10.8.5.

41.9 fps on windows 8, 29.5 fps on OSX. Clearly software plays a major difference when CPU limited.

Terrible hardware can show massive optimizations, it just depends on the time and money you through at it (PS3 -cell-uncharted series).

My point is NOT about user experience or anything like that (which I think nobody is getting). Its about the physical hardware. How much of Apple's A7 prowess is due to software (If I put an A7 in an android phone how would it run?) rather than actual hardware?

My point is to those who say the A7 shows that Apple is solidly out-engineering Qualcomm (etc.) and my answer is that the benchmarks make this unclear due to apple's massive software optimization.

Like I said, you clearly understand *NOTHING* about software optimization. Or CPU design. Or register / memory access that comes from optimization. Or cache design.

The Windows version of that game is DirectX based, the OS X version is OpenGL based. Comparing the two is moronic. You need to use the SAME implementation of the software (one optimized, one not optimized) to get a fair comparison. Different implementations of the same software will always perform differently, and its no secret that DirectX has a large leg up over OpenGL.

 

[cirus]

Like I said, you clearly understand *NOTHING* about software optimization. Or CPU design. Or register / memory access that comes from optimization. Or cache design.

The Windows version of that game is DirectX based, the OS X version is OpenGL based. Comparing the two is moronic. You need to use the SAME implementation of the software (one optimized, one not optimized) to get a fair comparison. Different implementations of the same software will always perform differently, and its no secret that DirectX has a large leg up over OpenGL.

Which is the entire point of what I am saying. People are comparing the A7 vs S600/S800/baytrail/etc on software based benchmarks (that are especially prone to software optimization being mostly JS) and saying that the A7 smashes the others out of the park due to hardware design. I'm saying that's not really true; apple is leveraging a massive software advantage to help the A7 perform.

There is no 'same implementation' of benchmarks for the A7 and other ARM chips so making a hardware comparison is flawed.

Those benchmarks are fine for looking at the observed performance (which is fine but beside the point) but inaccurate for actual hardware performance.

 

[Radiating]

The 5S was reported to have 2x the processing power. It just happens to be four times as fast with iOS 7, nobody was misled, they could have said it's 4 times faster.

 

[varunsanthanam]

Which is the entire point of what I am saying. People are comparing the A7 vs S600/S800/baytrail/etc on software based benchmarks (that are especially prone to software optimization being mostly JS) and saying that the A7 smashes the others out of the park due to hardware design. I'm saying that's not really true; apple is leveraging a massive software advantage to help the A7 perform.

There is no 'same implementation' of benchmarks for the A7 and other ARM chips so making a hardware comparison is flawed.

Those benchmarks are fine for looking at the observed performance (which is fine but beside the point) but inaccurate for actual hardware performance.

Okay seriously, you have no idea what you're saying. Nothing is 'prone' to software optimization, In fact the higher level the language, the less you can optimize it (javascript is incredibly high level and that is why it is used for so many benchmarks)

However, even lower level benchmarks show a HARDWARE advantage. You don't understand how benchmarks work. Geekbench, for example, The basic unit of of computing is a floating point operation. Bench marks run THE SAME floating point operations, written in both RISC architectures and x86. They are written directly as CPU instructions, not as 'C code' or any other language. Your video game comparison runs DIFFERENT floating operations on the same hardware, while These (geekbench and others) benchmarks run the SAME instructions on different hardware. It is not easy to optimize a CPU as well as optimized the A7. In order for that, the CPU has to be built / designed around the specific implementation of the optimization, and this can be very difficult. The CPU designer has to factor in the number of registers (data and instruction), location of said registers, size of said registers, and a number of other variables. Try designing a processor and you'll know, I say all this as a software developer who has spent many hours optimizing code for different CPUs. Forget about Apple's ability to software optimize, any developer who was worked with the new iPhone can tell you how incredibly easy Apple has made it to parallelize your code as well as upgrade to the new instruction set, thanks to the design of the A7. Believe me, what Apple has done with the A7 is brilliant, purely from a hardware standpoint. They managed to make a CPU which was the same size as the A6, that consumes less power and has twice as many data registers and twice as many instruction registers. While moore's lawmakers smaller transistors easier to achieve, it assume both the same instruction set/architecture as well as the same power consumption. Apple has managed to double the transistors in the same space, AS WELL as bringing the first implementation of ARMv8 (arm64) to the market AND lowering power consumption. Their register placement is ideal for optimization and incredible feat of HARDWARE engineering (I believe they did this by replacing the older 'swift' style cores with newer 'cyclone' style cores). In fact, I'd argue that software optimization (which Apple is admittedly very good at) is less impressive of a feat. With so many tools to optimize your code (everything form OpenMP, OpenCL, GrandCD and others to parallelize your code, as well as powerful new development runtime environments that can find unnecessary memory writes, compiler optimization that can automatically unroll and fix embedded loops to improve cache hit rates), optimizing code to run faster on the same hardware is far easier than it used to be. The fact that such optimizations lead to SUCH an improvement is purely related to hardware advances in Apple A7.

 

[apprunner]

Okay seriously, you have no idea what you're saying. Nothing is 'prone' to software optimization, In fact the higher level the language, the less you can optimize it (javascript is incredibly high level and that is why it is used for so many benchmarks)

However, even lower level benchmarks show a HARDWARE advantage. You don't understand how benchmarks work. Geekbench, for example, The basic unit of of computing is a floating point operation. Bench marks run THE SAME floating point operations, written in both RISC architectures and x86. They are written directly as CPU instructions, not as 'C code' or any other language. Your video game comparison runs DIFFERENT floating operations on the same hardware, while These (geekbench and others) benchmarks run the SAME instructions on different hardware. It is not easy to optimize a CPU as well as optimized the A7. In order for that, the CPU has to be built / designed around the specific implementation of the optimization, and this can be very difficult. The CPU designer has to factor in the number of registers (data and instruction), location of said registers, size of said registers, and a number of other variables. Try designing a processor and you'll know, I say all this as a software developer who has spent many hours optimizing code for different CPUs. Forget about Apple's ability to software optimize, any developer who was worked with the new iPhone can tell you how incredibly easy Apple has made it to parallelize your code as well as upgrade to the new instruction set, thanks to the design of the A7. Believe me, what Apple has done with the A7 is brilliant, purely from a hardware standpoint. They managed to make a CPU which was the same size as the A6, that consumes less power and has twice as many data registers and twice as many instruction registers. While moore's lawmakers smaller transistors easier to achieve, it assume both the same instruction set/architecture as well as the same power consumption. Apple has managed to double the transistors in the same space, AS WELL as bringing the first implementation of ARMv8 (arm64) to the market AND lowering power consumption. Their register placement is ideal for optimization and incredible feat of HARDWARE engineering (I believe they did this by replacing the older 'swift' style cores with newer 'cyclone' style cores). In fact, I'd argue that software optimization (which Apple is admittedly very good at) is less impressive of a feat. With so many tools to optimize your code (everything form OpenMP, OpenCL, GrandCD and others to parallelize your code, as well as powerful new development runtime environments that can find unnecessary memory writes, compiler optimization that can automatically unroll and fix embedded loops to improve cache hit rates), optimizing code to run faster on the same hardware is far easier than it used to be. The fact that such optimizations lead to SUCH an improvement is purely related to hardware advances in Apple A7.

From Anandtech:

"Apple’s own 64-bit Cyclone cores are incredibly powerful, even more so than I originally expected when I reviewed the iPhone 5s. Apple seems to have built a bigger, higher performance CPU architecture than any other ARM player, including ARM itself. The design isn’t perfect, but it’s a completely different caliber performer than anything else it competes against."

So yeah, the A7 does "smash" the competition in terms of hardware. Software optimization just lets it pull even further away from the comp

 

[varunsanthanam]

From Anandtech:

"Apple’s own 64-bit Cyclone cores are incredibly powerful, even more so than I originally expected when I reviewed the iPhone 5s. Apple seems to have built a bigger, higher performance CPU architecture than any other ARM player, including ARM itself. The design isn’t perfect, but it’s a completely different caliber performer than anything else it competes against."

So yeah, the A7 does "smash" the competition in terms of hardware. Software optimization just lets it pull even further away from the comp

Agreed whole heartedly. The OP does not know how benchmarks work, nor does he know how to read them correctly. The A7 is a huge advance in mobile SoCs. Software optimization is just icing on the cake.

 

[blackberrycubed]

Agreed whole heartedly. The OP does not know how benchmarks work, nor does he know how to read them correctly. The A7 is a huge advance in mobile SoCs. Software optimization is just icing on the cake.

Guess OP won't be getting paid by Samsung\MS\Nokia as his "spread misinformation" agenda has been squashed.

http://abcnews.go.com/Technology/sa...ple-criticize-htcs-products/story?id=20671547

Anandtech does one of the best and most thorough reviews I know of and if he says A7 kills the competition then I will believe him not some random troll who does not even know how to read benchmarks.

Thread should be closed.

 

[cirus]

Okay seriously, you have no idea what you're saying. Nothing is 'prone' to software optimization, In fact the higher level the language, the less you can optimize it (javascript is incredibly high level and that is why it is used for so many benchmarks)

However, even lower level benchmarks show a HARDWARE advantage. You don't understand how benchmarks work. Geekbench, for example, The basic unit of of computing is a floating point operation. Bench marks run THE SAME floating point operations, written in both RISC architectures and x86. They are written directly as CPU instructions, not as 'C code' or any other language. Your video game comparison runs DIFFERENT floating operations on the same hardware, while These (geekbench and others) benchmarks run the SAME instructions on different hardware. It is not easy to optimize a CPU as well as optimized the A7. In order for that, the CPU has to be built / designed around the specific implementation of the optimization, and this can be very difficult. The CPU designer has to factor in the number of registers (data and instruction), location of said registers, size of said registers, and a number of other variables. Try designing a processor and you'll know, I say all this as a software developer who has spent many hours optimizing code for different CPUs. Forget about Apple's ability to software optimize, any developer who was worked with the new iPhone can tell you how incredibly easy Apple has made it to parallelize your code as well as upgrade to the new instruction set, thanks to the design of the A7. Believe me, what Apple has done with the A7 is brilliant, purely from a hardware standpoint. They managed to make a CPU which was the same size as the A6, that consumes less power and has twice as many data registers and twice as many instruction registers. While moore's lawmakers smaller transistors easier to achieve, it assume both the same instruction set/architecture as well as the same power consumption. Apple has managed to double the transistors in the same space, AS WELL as bringing the first implementation of ARMv8 (arm64) to the market AND lowering power consumption. Their register placement is ideal for optimization and incredible feat of HARDWARE engineering (I believe they did this by replacing the older 'swift' style cores with newer 'cyclone' style cores). In fact, I'd argue that software optimization (which Apple is admittedly very good at) is less impressive of a feat. With so many tools to optimize your code (everything form OpenMP, OpenCL, GrandCD and others to parallelize your code, as well as powerful new development runtime environments that can find unnecessary memory writes, compiler optimization that can automatically unroll and fix embedded loops to improve cache hit rates), optimizing code to run faster on the same hardware is far easier than it used to be. The fact that such optimizations lead to SUCH an improvement is purely related to hardware advances in Apple A7.

Javascript benchmarks are very poor indications of hardware performance, given that they can differ greatly depending on what browser is used (run sunspider on IE and chrome and see how the scores compare). This alone should point to how much software optimization is key to JS.

http://support.primatelabs.com/disc...3-compile-options-for-ios-and-for-android-arm

You are absolute deluded if you think benchmarks are written as CPU instructions. The programmer never touches low level code or assembly. Geekbench is written (ironically "Mr. Software" developer in C/C++) and then complied using Clang for IOS and GCC for android. The two verisons, while written in the same base code are changed completely when compiled. They are not anything like the 'same' instruction, rather the compiled program orders the CPU to 'run this procedure list to get this value' on the CPU, with the list being a higher level list of instructions that generally works on all similar CPUs (and in the same operating- OS conditions). Using different compilers should immediately raise red flags as the compilers may return differently performing pieces of code depending on how they are tweaked (remember ICC and Intel?).

http://www.anandtech.com/show/7460/apple-ipad-air-review/3

And no the A7 does not, strictly speaking consume less power. AT ipad air review. Ipad Air has significantly lower platform power consumption but uses the same power under load ~7W platform. There appears to be a 3W delta between idle and Kraken on the Air and a 1.5 ish W difference on the ipad 4.

If you put a pathelogical workload on the SOC the A7 clearly consumes more power than the ipad 4. Look at the delta in power. ~11.5-3.7 = 7.8W for the A7 while the A6 has a delta ~7.6-4.8W = 2.8W. This is a non consumer workload and you will likely never run something this demanding on an ipad (unless someone comes out with hardcore rendering software in which case a milder version of this behavior may be seen). Battery life is basically the same. Yes the battery capacity has dropped but it also looks like the platform power consumption has also dropped by just over a watt

And please don't forget that the A6 is fabbed on Samsung HKMG 32nm and the A7 is build on HKMG 28nm. All things being equal a lower power draw should be seen anyways.

Apple's software innovations are due to the fact that they only have to support a couple different chips. That they can design the chips for their software and can code for specific chips (like apps are written for certain resolutions). That they have the profit margins and consequently the time and money to optimize this code (you can see this with intel's compilers where they can be massively faster than GCC for certain tasks.


As for software optimization we can go back and look at the iphone 4s.

iphone 4s 800mhz dual core A9.
Samsung galaxy S2 1.2 ghz core A9

Wow look at the iphone go... on exactly the same CPU (type) but only 66% the frequency. Even the iphone 4 with an 800 mhz single core A8 is doing as well as the dual core A9 phone with a 50% frequency advantage.

Is it unreasonable to expect Apple to still be maintaining this type of software advantage today?

 

[akdj]

"Javascript benchmarks are very poor indications of hardware performance, given that they can differ greatly depending on what browser is used (run sunspider on IE and chrome and see how the scores compare). This alone should point to how much software optimization is key to JS."

Interesting...as it seems the 'better and best' performing devices to the end user are the 'best' performing SunSpider units....some sort of correlation? Same goes with GeekBench. Anand in his review mentions...not in just this review but almost every mobile review he undertakes....that more/better and specific benchmarking tests are necessary in this day and age. That said, we've got what we've got. I'm not sure anyone messing around with low level programming via G/B or Sunspider...but he delves extremely deep into not only the computing power, the graphics and Calman with display performance...on each and every test...the A7/M7 and ISP---the 5s and Air implementations of the A7 are truly, 100% potent SoCs. It's exhausting to see you go on while being educated by someone with the intelligence of SoC design, software implementation and it's inevitable impact on performance....as not only is Apple writing the code for their OS---but THEY are the ONLY OEM currently taking the off the shelf ARM SoC and actually doing the low level programming necessary to A) work well with their OS/UI and the development community on a new 64bit platform and B) optimize energy/efficiency of the battery is such a small package with almost a 25% reduction in battery size while maintaining...in some cases exceeding it's predecessor's performance...

Which brings me to my next point...

"And please don't forget that the A6 is fabbed on Samsung HKMG 32nm and the A7 is build on HKMG 28nm. All things being equal a lower power draw should be seen anyways."

Doesn't matter who it's 'fabbed' by....The A5 (previous to the A6) is when Apple and company began the low level programming and intricacies of their own chip designs....remember when they hired several 'big' chip designers from AMD and Intel a couple years back? They've been mixing and baking and working directly with their SoC 'fabbers' to build the chip to their specs....not just their mobile designs but their computer iGPU work hand in hand with Intel has proven to be significant as well....as Intel recognizing the portability of the 'Air' and the idea now of the 'Ultra Book' and Intel's $300million backing, I think that's pretty telling.....not to mention Intel's announcement yesterday that they, too, will be providing and 'fabbing' 64 bit mobile chips 1st quarter 2014----kinda ironic, wouldn't you say?

Sorry---didn't want to quote your entire response, as that's between you and Shadow%20Mac. I'm simply a side line reader that finds it interesting your dismissal of such an engineering marvel...as well as you pointing to Anand's site while dismissing his findings.

As well....it's up to each OEM to implement (or not, IE Google) their own 'Java' Script layer of frosting to Android---Touch Wiz, Sense, or even with the active community of folks flashing their ROMS, using different launchers....ALL in an effort to achieve the fluency one receives immediately when they open an Apple box....it's a lot like a Windows computer vs. opening an Apple computer. I'm 42 and I've grown up with both---there is such a HUGE difference when a company controls everything---from the design of how the hardware (made by other companies) is put together to the OS'es optimization for that specific computer or tablet or phone...NOW that Apple has been doing their own chip design not just with low level programming but actual, real hardware/transistor design, it's even more evident...again, not only in artificial benchmarking but the 'End User Experience'.

J

 

[cirus]

I think its the reading comprehension.

"Javascript benchmarks are very poor indications of hardware performance, given that they can differ greatly depending on what browser is used (run sunspider on IE and chrome and see how the scores compare). This alone should point to how much software optimization is key to JS."

Interesting...as it seems the 'better and best' performing devices to the end user are the 'best' performing SunSpider units....some sort of correlation? Same goes with GeekBench. Anand in his review mentions...not in just this review but almost every mobile review he undertakes....that more/better and specific benchmarking tests are necessary in this day and age. That said, we've got what we've got. I'm not sure anyone messing around with low level programming via G/B or Sunspider...but he delves extremely deep into not only the computing power, the graphics and Calman with display performance...on each and every test...the A7/M7 and ISP---the 5s and Air implementations of the A7 are truly, 100% potent SoCs. It's exhausting to see you go on while being educated by someone with the intelligence of SoC design, software implementation and it's inevitable impact on performance....as not only is Apple writing the code for their OS---but THEY are the ONLY OEM currently taking the off the shelf ARM SoC and actually doing the low level programming necessary to A) work well with their OS/UI and the development community on a new 64bit platform and B) optimize energy/efficiency of the battery is such a small package with almost a 25% reduction in battery size while maintaining...in some cases exceeding it's predecessor's performance...

Which brings me to my next point...

"And please don't forget that the A6 is fabbed on Samsung HKMG 32nm and the A7 is build on HKMG 28nm. All things being equal a lower power draw should be seen anyways."

Doesn't matter who it's 'fabbed' by....The A5 (previous to the A6) is when Apple and company began the low level programming and intricacies of their own chip designs....remember when they hired several 'big' chip designers from AMD and Intel a couple years back? They've been mixing and baking and working directly with their SoC 'fabbers' to build the chip to their specs....not just their mobile designs but their computer iGPU work hand in hand with Intel has proven to be significant as well....as Intel recognizing the portability of the 'Air' and the idea now of the 'Ultra Book' and Intel's $300million backing, I think that's pretty telling.....not to mention Intel's announcement yesterday that they, too, will be providing and 'fabbing' 64 bit mobile chips 1st quarter 2014----kinda ironic, wouldn't you say?

As well....it's up to each OEM to implement (or not, IE Google) their own 'Java' Script layer of frosting to Android---Touch Wiz, Sense, or even with the active community of folks flashing their ROMS, using different launchers....ALL in an effort to achieve the fluency one receives immediately when they open an Apple box....it's a lot like a Windows computer vs. opening an Apple computer. I'm 42 and I've grown up with both---there is such a HUGE difference when a company controls everything---from the design of how the hardware (made by other companies) is put together to the OS'es optimization for that specific computer or tablet or phone...NOW that Apple has been doing their own chip design not just with low level programming but actual, real hardware/transistor design, it's even more evident...again, not only in artificial benchmarking but the 'End User Experience'.

J

I have already posted numerous times that I am NOT looking at the end user experience, simply the chip design.

And low level programming and optimization is precisely why the A7 is doing so well. You seem to agree.

I said Samsung because there are probably more similarities between their HKMG process and for example TMSC's.

Its not the the A7 is exceeding good, its that apple can extract such a large percentage of its performance through optimization and software that makes it amazing. In my previous post if you look at the iphone 4 or 4s you would expect to see some amazing CPU when you are really getting something much weaker than performs well due to optimization and software.

And like anything you read you must take it with a grain of salt. AT is good but you can see things were you may want to take a second opinion (like in his haswell review where he called Haswell a 'success' when almost every other tech site called it a 'yawn' or something to that effect). Its perfectly acceptable to use someones figures but disagree with their conclusion, especially when they are the only site that publishes certain figures.

 

[GoSh4rks]

And like anything you read you must take it with a grain of salt. AT is good but you can see things were you may want to take a second opinion (like in his haswell review where he called Haswell a 'success' when almost every other tech site called it a 'yawn' or something to that effect). Its perfectly acceptable to use someones figures but disagree with their conclusion, especially when they are the only site that publishes certain figures.

Context matters here. Haswell on mobile is a giant leap forward. Haswell on desktop, not so much.

 

[cirus]

Context matters here. Haswell on mobile is a giant leap forward. Haswell on desktop, not so much.

And the context is desktop.

http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/9

...then I’d consider Haswell a success on the desktop

And haswell on mobile is really only a leap in two areas, igp and ULV. SV 35/45 W parts saw maybe a 5-10% increase in performance at max load. Idle power is a little down, partly because of 800 mhz idle clocks. ULT haswell is a major jump forward, mainly because of improved PCH (on chip allows power power savings) and igp, rather than anything at the core level. ULT haswell is awesome and baytrail isn't bad either but they are not sandy bridge impressive.

Efficiency has improved marginally on mobile and recessed slightly on the desktop. Haswell on the desktop is a 'yawn' and almost every other tech site reported the same.

 

[GoSh4rks]

And the context is desktop.

http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/9

Meh, he put in a bunch of qualifiers which framed what he was referring to as a "success." When you take those into consideration, he isn't wrong.

 

[cirus]

Meh, he put in a bunch of qualifiers which framed what he was referring to as a "success." When you take those into consideration, he isn't wrong.

Um his main qualifier is the lack of overclocking. He basically says that haswell improves performance a tiny bit at the same price. Haswell on the desktop is decidedly unimpressive according to every almost every other tech site on the web. .

The AT haswell review is definitely sub par for a site like AT. His sandy bridge and ivy bridge reviews were much better. AT is a great site but please, like anything on the internet, take their reviews with a grain of salt. In their cpu/motherboard section they keep saying that an a8 is a fine match for a 7970 GE when pretty much everyone knows that that will be a tremendous bottleneck on a lot of games.

 

[DanTSX]

Wayyyy too much nerd-rage here.

Bleating about CPU performance is for android dorks.

Apple has chosen an excellent processor that is part of their long-term plan, and will serve as the basis of development for future processors. And in all cases, "it just works" will be the experience for the user with a phone up to 3 or even 4 years old with whatever software comes down the pike for the user's application.

Stop arguing which Samsung/Motorola/HTC POS can draw Mario's hat faster.

 

[gentlefury]

Apple's A7 is largely seen as a very impressive piece of CPU engineering. However, what you don't see is how much software makes a difference; apple's A7 is impressive (and A6) mainly because they leverage Apple's substantial software advantage rather than great hardware prowess.

It can be seen that a lot of improvements in performance across generations are from software upgrades rather than any sort of hardware improvement. Case in point, the improvement from the iphone 5 to the 5c.

AT release article of the iphone 5
http://www.anandtech.com/show/6330/the-iphone-5-review/10

Image

Image

The difference between these benchmarks is that one was taken with IOS 6 and the other with IOS 7. As you can see, the iphone 5 improved by 40%.

Likewise with Octane

Image

Image

71% improvement from IOS 6 to IOS 7.

Sunspider shows a smaller but still present improvement.

I have no doubt that the user experience from the A7 is every good as the charts indicate but the actual hardware performance isn't nearly as close to the baytrail or S800 as benchmarks indicate. Apple's software optimizations allow them to put a very strong showing with weaker hardware. (Geekbench is not used in this comparison as it focuses too much on encryption and tends to operate on single buffers of data). As shown in other benches, 5s performance is notably higher than 5/5c performance; my point here is that the actual hardware is less powerful in comparison to other hardware than the benchmarks indicate.

This is the problem with using JS benchmarks.

Good thing in the real world, user experience trumps white sheets any day! Having a more optimized OS is much more important than greater power. Higher power CPUs only peas to lower battery life, especially when your os is a mess.

 

[Merkie]

I don't understand the OP at all. Running both iOS 7, the 5S is still a lot faster than 5. I don't understand what the relevance of the iOS 7 software optimization is. We're comparing iOS 7 devices here, what does it matter how it compares to older iOS versions?

The fact is that the A7 is an incredibly fast chip. And it's also very energy efficient. I don't see how any performance increase from jumping to iOS 7 from an earlier iOS version is relevent at all.

 

[Troneas]

macrumors at its best: suggest that apple's hardware is not all that apple claims to be or that there is a caveat to apple's comparisons between new and older devices (ie: the 64-bit chip isn't all that responsible for those lovely benchmarks) and everyone attacks the OP dismissing the point as irrelevant.

funny how when qualcomm VP said the A7 was a gimmick everyone forgot about the "end user experience" and it was all about 64-bit this and 64-bit that.



at the end of the day, its just *interesting* to know... at least for some. not everything has to be *valuable* information.


i had no idea the differences in performance would be so huge changing from an older to a new OS. And now I am kept wondering which OS apple uses when it claims the new idevice is 10000000 faster than the previous one...


very interesting OP, thank you.

 

[akdj]

I think its the reading comprehension.



I have already posted numerous times that I am NOT looking at the end user experience, simply the chip design.

And low level programming and optimization is precisely why the A7 is doing so well. You seem to agree.

I said Samsung because there are probably more similarities between their HKMG process and for example TMSC's.

Its not the the A7 is exceeding good, its that apple can extract such a large percentage of its performance through optimization and software that makes it amazing. In my previous post if you look at the iphone 4 or 4s you would expect to see some amazing CPU when you are really getting something much weaker than performs well due to optimization and software.

And like anything you read you must take it with a grain of salt. AT is good but you can see things were you may want to take a second opinion (like in his haswell review where he called Haswell a 'success' when almost every other tech site called it a 'yawn' or something to that effect). Its perfectly acceptable to use someones figures but disagree with their conclusion, especially when they are the only site that publishes certain figures.

You seem to contradict yourself here---not that it's exceedingly good---but what they do @ the L/L makes it 'amazing'---sooooo, which is it? Not good or Amazing? I'm confused...BUT I think you're starting to 'see the light' and that's a good thing

Apple has created an amazing device(s) (iPhone/iPad) all the while utilizing off the shelf parts and doing something an OEM hasn't done to this point...tweak the chip, work with the silicon designer, low level A8 instruction set---all while (re)building their own OS on top of the optimized chip. Your reference to the 4/4s in comparison with other 'like' chipped phones I think actually cements when Apple started to take advantage of this---it's certainly a year or two after they hired well respected chip designers and engineers away from AMD and Intel...it was the first phone on ANY platform that went 'retina' or high rez, can't see the pixels, yada, yada, yada....they needed to eek out that extra performance from the silicon to support the display, graphics and not lose battery life to terribly in the process

As far as Haswell----sure seems to have made a helluva improvement on the MacBook Air---along with Mavericks some folks now declaring up to 15 hours unplugged??? (13"). That's pretty awesome! Not to mention an iGPU that ought to drive that 13" rMBP pretty well...even with the ability to play some AAA games I'm sure. Now you DID mention this...but to dismiss A/T's review and everyone else being 'correct' or in agreement, I think is sad! They managed to double the performance of their iGPU---gain incredible efficiency from the '4th' generation of their current chip design. Those that yawn because desktop performance (along with mobile) in the CPU posted modest 5-12% gains aren't taking in to account the continued movement to mobile, en masse, and the lack of the need for a traditional 'desktop' each year. Of course there will the builders, hobbyists and gamers....but they're the minority. I'd bet...if polled, that 80-90% of the population sampled if asked what their next 'computer' purchase is going to be the answer would far and away be a laptop or tablet.
Tl/Dr---I agree with A/T on Haswell as well. Not Ars, Not T/C and not Gizmodo.

Sorry---as an owner of a 2011 Air that's still just fine, looking at the differences between Sandy Bridge and Haswell (just a 2 generation jump)...as you mentioned, the two differences discussed, energy and iGPU power has been exponentially increased in comparison. That's pretty amazing.

I'm with ya---take what you read on the internet with a grain of salt...especially that picture on Craigslist

J

 

[WilliamG]

Mine hasn't crashed ever. May want to try a clean install.

Then maybe you're not as heavy a user? My 5s and my wife's 5s crash occasionally. It's usually the Apple logo followed by the springboard a few seconds later. Not the end of the world, but there's no doubt iOS7 crashes more than any other iOS.

 

[cirus]

I don't understand the OP at all. Running both iOS 7, the 5S is still a lot faster than 5. I don't understand what the relevance of the iOS 7 software optimization is. We're comparing iOS 7 devices here, what does it matter how it compares to older iOS versions?

The fact is that the A7 is an incredibly fast chip. And it's also very energy efficient. I don't see how any performance increase from jumping to iOS 7 from an earlier iOS version is relevent at all.

Yes the 5s is faster than the 5, its roughly twice as fast. That is certainty true. My point is when drawing a comparison to any non-apple phone and possibly apple software optimization.

Good thing in the real world, user experience trumps white sheets any day! Having a more optimized OS is much more important than greater power. Higher power CPUs only peas to lower battery life, especially when your os is a mess.

Which while certainty true is completely irrelevant to this thread. I have stated numerous time that I AM NOT LOOKING AT USER EXPERIENCE! Only the physical hardware in the Soc.

You seem to contradict yourself here---not that it's exceedingly good---but what they do @ the L/L makes it 'amazing'---sooooo, which is it? Not good or Amazing? I'm confused...BUT I think you're starting to 'see the light' and that's a good thing

Apple has created an amazing device(s) (iPhone/iPad) all the while utilizing off the shelf parts and doing something an OEM hasn't done to this point...tweak the chip, work with the silicon designer, low level A8 instruction set---all while (re)building their own OS on top of the optimized chip. Your reference to the 4/4s in comparison with other 'like' chipped phones I think actually cements when Apple started to take advantage of this---it's certainly a year or two after they hired well respected chip designers and engineers away from AMD and Intel...it was the first phone on ANY platform that went 'retina' or high rez, can't see the pixels, yada, yada, yada....they needed to eek out that extra performance from the silicon to support the display, graphics and not lose battery life to terribly in the process

As far as Haswell----sure seems to have made a helluva improvement on the MacBook Air---along with Mavericks some folks now declaring up to 15 hours unplugged??? (13"). That's pretty awesome! Not to mention an iGPU that ought to drive that 13" rMBP pretty well...even with the ability to play some AAA games I'm sure. Now you DID mention this...but to dismiss A/T's review and everyone else being 'correct' or in agreement, I think is sad! They managed to double the performance of their iGPU---gain incredible efficiency from the '4th' generation of their current chip design. Those that yawn because desktop performance (along with mobile) in the CPU posted modest 5-12% gains aren't taking in to account the continued movement to mobile, en masse, and the lack of the need for a traditional 'desktop' each year. Of course there will the builders, hobbyists and gamers....but they're the minority. I'd bet...if polled, that 80-90% of the population sampled if asked what their next 'computer' purchase is going to be the answer would far and away be a laptop or tablet.
Tl/Dr---I agree with A/T on Haswell as well. Not Ars, Not T/C and not Gizmodo.

Sorry---as an owner of a 2011 Air that's still just fine, looking at the differences between Sandy Bridge and Haswell (just a 2 generation jump)...as you mentioned, the two differences discussed, energy and iGPU power has been exponentially increased in comparison. That's pretty amazing.

I'm with ya---take what you read on the internet with a grain of salt...especially that picture on Craigslist

J

I agree that what apple has done with the A7 is pretty amazing but again I have nothing to really compare it to except android. There is lots of evidence that the iphone 4/4s were much faster than the competition even though the competition had theoretically faster hadware.

As far as the A7 goes at the launch of the 4s apple had a phone that beat the pants off any ARM chip by the same margin on the CPU as the 5s. The gpu was also roughly twice as fast as the S2 mali 400 (seems to be the top or very close to the top at that time) while the S800 is extremely competitive vs the iphone 5s. I'm not seeing anything different today and Apple achieved that with a 800 mhz dual A9 against 1.2 ghz dual A9s. I honestly have no idea how fast the A7 actually is and it it impossible to really know (though I expect it to be significantly less than seen in benchmarks). And unless someone can test the A7 under android with android like optimization it would not be a bad idea to assume that something similar is happening.

 

[Trius]

If you actually look at what I'm trying to say you would see than i'm mainly talking about intel and Qualcomm and their hardware. And really, someone says something you don't like and this is your response?



My point is that it is wrong to say that Apple's engineering of the A7 is what makes it an amazing chip. Its apple's software integration that can take a chip like the A6 and turn it into a solid S600 competitor. People saying that apple is outengineering Qualcomm, intel, or samsung based on the performance of the A7 are wrong. A7 performs well because of Apple's software integration.


Its well known that AT loves apple (and intel) products.

I get what you're trying to do here.. but now you're just arguing semantics.. The "software integration" you speak of, has all been engineered by Apple. They created a thing that is a collaboration of hardware components and software.. Together, they outperform the competition. It's all engineering, bro.