That's not the number that shocks the .... out of me. The Image Signal Processor is claimed to perform 100 billion operations in 25ms - that's 4 trillion operations a second.
Gotta be a major deep neural network to get that kind of parallelism...
Sadly the Soc upgrade is not important as the screen density is outdated. Pretty sure they are saving that for next year. Nice of apple for giving people 720p displays on the 4.7 model. Some $100 phone has a better display.
The screen density of the 7+ is within a few % from the screen density of the S7 once you know its a pentile screen (418*1.25=S7 Resolution). But, maybe you're Batman.
Many people complain that the high resolution of many of phones like the LG G5 lead to horrible battery life (that comes from actual users); and the G5 is a larger phone that has a big enough battery.
The 7 is much smaller and that's the reason the resolution is lower, a higher resolution would kill the battery. People have been already complaining about the 6s battery life and you want them to kill it faster!
No, I think the 7 still has 2 GB. They say the 7 plus has 3 because of the extra processing required for both cameras.
It takes another 1 GB of RAM to process an extra 3 MB photo? Nah.
A 12 megapixel camera produces a raw dataset of at least 36 MB. Apple has stated that the new camera captures "wide color", so we could assume that they are using at 16 bits per channel instead of 8 bits. Of course that is just the final raw data. If the camera is taking multiple photos in rapid succession for HDR or capture additional information, I could see this easily requiring more RAM to process. At least 512 MB.
Video Processing with two lenses would take even more RAM.
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
Yep, the iPhone 7 is faster than the entry level MacBook, on both single (3233 vs. 2668) and multi-core (5363 vs. 4886). It's faster than the top line MacBook on single core (3233 vs. 2831) and fairly close on multi-core (5363 vc. 6008). Two more years of this kind of improvement and the iPhone might be benchmarking better than half the Mac line. Something's gotta give.
Imagine they didn't throttle it in a device that had more cooling and wattage available (did I get that right?) with which to contend, such as the AppleTV, or perhaps one day a laptop with A-series. Wonder how this would scream in one of those housings. Very impressive to get all that in a phone, how far before these mobile devices render the need for faster machines for the masses (not those who need power for work and such, but the email/social networking/internet browsing masses) moot?
Theres no secret to the A series, it is the transistor count. At 3.3 billion it is by far the largest cpu ever. It is quite a bit more than the A9 and 1 billion more than the A8. From Anandtech:
Apple SoC Evolution
Die Size
Transistors
Process
A5
122m2
<1B
45nm
A6
97mm2
<1B
32nm
A7
102mm2
>1B
28nm
A8
89mm2
~2B
20nm
A9
96mm2/104.5mm2
>2B
14nm/16nm
For some other reference. The quad core i7 in my 2012 macbook pro retina has a paltry 1.4 billion transistors with a 1.2 billion transistor GPU. The thing is positively massive. Remember when the A7 came out and Apple said they were building desktop class CPUs, it is quite true.
Apples vs Oranges, you're comparing a CPU vs a SOC. Of course a SOC will have more transistors.
Intel CPUs do not have radios, BT, nor disk controllers.
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
Yep, the iPhone 7 is faster than the entry level MacBook, on both single (3233 vs. 2668) and multi-core (5363 vs. 4886). It's faster than the top line MacBook on single core (3233 vs. 2831) and fairly close on multi-core (5363 vc. 6008). Two more years of this kind of improvement and the iPhone might be benchmarking better than half the Mac line. Something's gotta give.
Imagine they didn't throttle it in a device that had more cooling and wattage available (did I get that right?) with which to contend, such as the AppleTV, or perhaps one day a laptop with A-series. Wonder how this would scream in one of those housings. Very impressive to get all that in a phone, how far before these mobile devices render the need for faster machines for the masses (not those who need power for work and such, but the email/social networking/internet browsing masses) moot?
This amount of power leads to very good battery life and a cool stuff (eventually) that even the social networking masses can appreciate.
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
Yep, the iPhone 7 is faster than the entry level MacBook, on both single (3233 vs. 2668) and multi-core (5363 vs. 4886). It's faster than the top line MacBook on single core (3233 vs. 2831) and fairly close on multi-core (5363 vc. 6008). Two more years of this kind of improvement and the iPhone might be benchmarking better than half the Mac line. Something's gotta give.
Imagine they didn't throttle it in a device that had more cooling and wattage available (did I get that right?) with which to contend, such as the AppleTV, or perhaps one day a laptop with A-series. Wonder how this would scream in one of those housings. Very impressive to get all that in a phone, how far before these mobile devices render the need for faster machines for the masses (not those who need power for work and such, but the email/social networking/internet browsing masses) moot?
Theres no secret to the A series, it is the transistor count. At 3.3 billion it is by far the largest cpu ever. It is quite a bit more than the A9 and 1 billion more than the A8. From Anandtech:
Apple SoC Evolution
Die Size
Transistors
Process
A5
122m2
<1B
45nm
A6
97mm2
<1B
32nm
A7
102mm2
>1B
28nm
A8
89mm2
~2B
20nm
A9
96mm2/104.5mm2
>2B
14nm/16nm
For some other reference. The quad core i7 in my 2012 macbook pro retina has a paltry 1.4 billion transistors with a 1.2 billion transistor GPU. The thing is positively massive. Remember when the A7 came out and Apple said they were building desktop class CPUs, it is quite true.
Apples vs Oranges, you're comparing a CPU vs a SOC. Of course a SOC will have more transistors.
Intel CPUs do not have radios, BT, nor disk controllers.
The SOC I'm pretty sure doesn't have BT or radios or even disk controller (before the A9, which ). The A9 does include, the NVMe controller, cache, GPU, CPU, M9 and some DSP modules.
Sadly the Soc upgrade is not important as the screen density is outdated. Pretty sure they are saving that for next year. Nice of apple for giving people 720p displays on the 4.7 model. Some $100 phone has a better display.
The human eye can't tell the difference between 720p and 1080p on a screen that little.
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
Yep, the iPhone 7 is faster than the entry level MacBook, on both single (3233 vs. 2668) and multi-core (5363 vs. 4886). It's faster than the top line MacBook on single core (3233 vs. 2831) and fairly close on multi-core (5363 vc. 6008). Two more years of this kind of improvement and the iPhone might be benchmarking better than half the Mac line. Something's gotta give.
Imagine they didn't throttle it in a device that had more cooling and wattage available (did I get that right?) with which to contend, such as the AppleTV, or perhaps one day a laptop with A-series. Wonder how this would scream in one of those housings. Very impressive to get all that in a phone, how far before these mobile devices render the need for faster machines for the masses (not those who need power for work and such, but the email/social networking/internet browsing masses) moot?
Theres no secret to the A series, it is the transistor count. At 3.3 billion it is by far the largest cpu ever. It is quite a bit more than the A9 and 1 billion more than the A8. From Anandtech:
Apple SoC Evolution
Die Size
Transistors
Process
A5
122m2
<1B
45nm
A6
97mm2
<1B
32nm
A7
102mm2
>1B
28nm
A8
89mm2
~2B
20nm
A9
96mm2/104.5mm2
>2B
14nm/16nm
For some other reference. The quad core i7 in my 2012 macbook pro retina has a paltry 1.4 billion transistors with a 1.2 billion transistor GPU. The thing is positively massive. Remember when the A7 came out and Apple said they were building desktop class CPUs, it is quite true.
Apples vs Oranges, you're comparing a CPU vs a SOC. Of course a SOC will have more transistors.
Intel CPUs do not have radios, BT, nor disk controllers.
The older A series do have all of those The majority of transistors typically come from the GPU anyways. A GTX 1080 has 7.2 billion for example
Sadly the Soc upgrade is not important as the screen density is outdated. Pretty sure they are saving that for next year. Nice of apple for giving people 720p displays on the 4.7 model. Some $100 phone has a better display.
The human eye can't tell the difference between 720p and 1080p on a screen that little.
1) I'm not sure that's accurate, since it's missing a distance parameter, not to mention how our brains interpret moving images as opposed to ones that are static.
2) You would help your argument if you stated your case in more precise terms. For example, "The 4.7" iPhone's 1334 × 750 resolution display has a PPI of 325.61. If this was moved to a 1920 × 1080 the PPI would move to 468.70. While there are certainly upper and lower limits to human vision, using Snellen notation we can figure out an individual's visual acuity in minutes of arc (e.g.: 20/20 vision = 1 arc minute) or at least 1 pixel per arcminute, then we have an equation that is simply 3437 ÷ (inches held from face -or- PPI) for those with 20/20 vision. For a display with a 325 PPI (rounding down), the display would have be closer than 10.57" to your eyes before you would be able to discern individual pixels on a still image. If you want to figure out how this works for those with much better eyesight, you can read the attached link below."
Sadly the Soc upgrade is not important as the screen density is outdated. Pretty sure they are saving that for next year. Nice of apple for giving people 720p displays on the 4.7 model. Some $100 phone has a better display.
Oh look everyone, another Android spec whore, who doesn't even know how to properly interpret a full spec sheet either! Indicative of intelligence...
Sadly the Soc upgrade is not important as the screen density is outdated. Pretty sure they are saving that for next year. Nice of apple for giving people 720p displays on the 4.7 model. Some $100 phone has a better display.
That $100 phone will provide its user with endless UI lag and animation stutter as well as poor battery life because the cheap MediaTek SOC will be running flat out just to push that many pixels. Stupid is as stupid does.
This is no longer true. As owner of an app development company, I have a a bunch of test devices (Android and iPhones). Recently, the performance of Android devices is improved significantly (just like the iPhone), only the impact of that improvement is different: a cheap Android phone moved from unacceptable slow to fast enough, an iPhone moved from fast to very fast. For running day to day tasks like switching apps, browsing, email, agenda, taking notes, .... a 2016 Android phone of $100 has no lag. CPU intensive games are of course another matter.
I just looked at GeekBench 4 stats for Intel Core i7-6700K @ 4.00 GHz which comes within a rounding error of what you'd get if you multiplied the single core CPU stats mentioned in this post. You'd have a single core number of 5799.
Of course, there are unknowns here:
How fast is the main memory used in the A10 Fusion SoC?
Would that main RAM still sustain keeping the A10 Fusion fed at that current latency/frequency?
What would be the power usage of that main RAM?
Can the process used for the A10 Fusion scale up in speed to run at 4.00 Ghz (1.79 times as fast) and be reliable?
What is the power used at that speed, if so? (Note: power used is not linear with frequency, but instead is more exponential)
How much waste heat at that speed, if so?
What if the reason that Apple has prolonged the cycle of various laptops (MacBook Pro, I'm waiting for you) is because they have a full A10 quad core (no big/little) in store for their non-iOS devices? Admittedly, even if they had A10 processors comparable in speed to Intel i7 processors in their laptops, that'd still be a slowdown for Intel CPU software even in the best-case scenario, unless Apple included some special hardware to convert it to ARM ops or something even wilder. While they'd be able to make it easy for App Store apps to be cross-CPU, more work would be required for pre-existing apps not sold that way. Note: I write software for a living, and work on low-level software: never assume software built for one CPU will work exactly as you expect for another CPU without testing
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
People love saying this, but have no idea how Apple would pull it off. I no doubt Apple has A-series versions of OSX somewhere just in case they can't squeeze a deal from Intel.
However there is exactly zero percent chance of Apple switching MacBook Pros and iMac's to A-series. It makes far more sense for Apple to make the iPad Pro outrun the MacBook Pro (which the GPU already does http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html ) and bring more "full fledged" Apple software to the iPad Pro that happens to optionally support the keyboard and stylus, rather than to retrofit macOS with A-series and have only iOS software work on it.
The end-goal is full Apple ecosystem lock-in, and no more pesky laptop/desktop "Pro" users with all their non-Apple junk to support. I'm being sarcastic about this point BTW.
Realisticly, Apple has no reason to migrate to their own chips for Laptops and Desktops because we're nearing the end of the performance we can get. Intel has not released any CPU parts faster than Haswell (i7-4790K,) and that was an entire 3 years ago. All other CPU parts operate at 1.6Ghz at the low-end which is less than half the speed of the top end part. Apple's A-series in the iPhone 6S is 2518 on geekbench, the iPhone 7 is maybe 3233. The 4790K gets 5794 at 4Ghz. So the A9 is closer to a 1.6Ghz Haswell than it is a 4Ghz Haswell. This also completely ignores the GPU.
If the goal is to put the A-series on the desktop, it has to hit 20,000 on the multicore with 4 cores while beating 7000 on the singlecore. We are less than halfway there. Apple could hit the multicore score right now if they created an 8-core model, but that would still not hit the required performance for a desktop. For a laptop it's only comparable to the bottom-of-the-line 1.6 to 2.0 ghz laptops.
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
People love saying this, but have no idea how Apple would pull it off. I no doubt Apple has A-series versions of OSX somewhere just in case they can't squeeze a deal from Intel.
However there is exactly zero percent chance of Apple switching MacBook Pros and iMac's to A-series. It makes far more sense for Apple to make the iPad Pro outrun the MacBook Pro (which the GPU already does http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html ) and bring more "full fledged" Apple software to the iPad Pro that happens to optionally support the keyboard and stylus, rather than to retrofit macOS with A-series and have only iOS software work on it.
The end-goal is full Apple ecosystem lock-in, and no more pesky laptop/desktop "Pro" users with all their non-Apple junk to support. I'm being sarcastic about this point BTW.
Realisticly, Apple has no reason to migrate to their own chips for Laptops and Desktops because we're nearing the end of the performance we can get. Intel has not released any CPU parts faster than Haswell (i7-4790K,) and that was an entire 3 years ago. All other CPU parts operate at 1.6Ghz at the low-end which is less than half the speed of the top end part. Apple's A-series in the iPhone 6S is 2518 on geekbench, the iPhone 7 is maybe 3233. The 4790K gets 5794 at 4Ghz. So the A9 is closer to a 1.6Ghz Haswell than it is a 4Ghz Haswell. This also completely ignores the GPU.
If the goal is to put the A-series on the desktop, it has to hit 20,000 on the multicore with 4 cores while beating 7000 on the singlecore. We are less than halfway there. Apple could hit the multicore score right now if they created an 8-core model, but that would still not hit the required performance for a desktop. For a laptop it's only comparable to the bottom-of-the-line 1.6 to 2.0 ghz laptops.
Why would you think that ARM based MacOS computers would run only iOS software? I expect that if Apple fielded ARM based Macs, they'd run MacOS apps, just as they always have.
If they continue down this path, it will eventually be difficult for developers to write CPU specific code. At that point, Apple could switch out the underlying hardware architecture at will. Given the depth and breadth of Apple's CPU/GPU design expertise, it's not unthinkable that they could actually field their own CPU and GPU free of anyone else's IP. The licensing fees and product margins currently being paid to others (ARM, Intel) would be retained by Apple. The economics may not favor such a move, but it's not unthinkable.
With Intel's recent taking of an ARM license in order to support a foundry role for large customers, it seems Apple stands to gain from consolidating down to one CPU/GPU architecture and playing multiple silicon vendors off each other.
Sadly the Soc upgrade is not important as the screen density is outdated. Pretty sure they are saving that for next year. Nice of apple for giving people 720p displays on the 4.7 model. Some $100 phone has a better display.
The human eye can't tell the difference between 720p and 1080p on a screen that little.
Well sitting here with a whole range of devices (since I develop for both Android and iOS... Prefer iOS of course) I can say even with my messed up eyes I can tell the different (in resolution, not talking color, etc) on a > 4.5 screen at 12 inches.
What is surprising is that iPhone 7 Plus is already FASTER than iPad Pro! Just think how it keeps getting faster and faster with no signs of leveling off at all. So, I can see how Apple may eventually use ARM chips in future Macs. Their chip is just way MUCH faster in terms of performance per watt.
Comments
Gotta be a major deep neural network to get that kind of parallelism...
Many people complain that the high resolution of many of phones like the LG G5 lead to horrible battery life (that comes from actual users); and the G5 is a larger phone that has a big enough battery.
The 7 is much smaller and that's the reason the resolution is lower, a higher resolution would kill the battery.
People have been already complaining about the 6s battery life and you want them to kill it faster!
Intel CPUs do not have radios, BT, nor disk controllers.
The A9 does include, the NVMe controller, cache, GPU, CPU, M9 and some DSP modules.
2) You would help your argument if you stated your case in more precise terms. For example, "The 4.7" iPhone's 1334 × 750 resolution display has a PPI of 325.61. If this was moved to a 1920 × 1080 the PPI would move to 468.70. While there are certainly upper and lower limits to human vision, using Snellen notation we can figure out an individual's visual acuity in minutes of arc (e.g.: 20/20 vision = 1 arc minute) or at least 1 pixel per arcminute, then we have an equation that is simply 3437 ÷ (inches held from face -or- PPI) for those with 20/20 vision. For a display with a 325 PPI (rounding down), the display would have be closer than 10.57" to your eyes before you would be able to discern individual pixels on a still image. If you want to figure out how this works for those with much better eyesight, you can read the attached link below."
Of course, there are unknowns here:
What if the reason that Apple has prolonged the cycle of various laptops (MacBook Pro, I'm waiting for you) is because they have a full A10 quad core (no big/little) in store for their non-iOS devices? Admittedly, even if they had A10 processors comparable in speed to Intel i7 processors in their laptops, that'd still be a slowdown for Intel CPU software even in the best-case scenario, unless Apple included some special hardware to convert it to ARM ops or something even wilder. While they'd be able to make it easy for App Store apps to be cross-CPU, more work would be required for pre-existing apps not sold that way. Note: I write software for a living, and work on low-level software: never assume software built for one CPU will work exactly as you expect for another CPU without testing
However there is exactly zero percent chance of Apple switching MacBook Pros and iMac's to A-series. It makes far more sense for Apple to make the iPad Pro outrun the MacBook Pro (which the GPU already does http://www.notebookcheck.net/AMD-Radeon-R9-M370X.142763.0.html ) and bring more "full fledged" Apple software to the iPad Pro that happens to optionally support the keyboard and stylus, rather than to retrofit macOS with A-series and have only iOS software work on it.
The end-goal is full Apple ecosystem lock-in, and no more pesky laptop/desktop "Pro" users with all their non-Apple junk to support. I'm being sarcastic about this point BTW.
Realisticly, Apple has no reason to migrate to their own chips for Laptops and Desktops because we're nearing the end of the performance we can get. Intel has not released any CPU parts faster than Haswell (i7-4790K,) and that was an entire 3 years ago. All other CPU parts operate at 1.6Ghz at the low-end which is less than half the speed of the top end part. Apple's A-series in the iPhone 6S is 2518 on geekbench, the iPhone 7 is maybe 3233. The 4790K gets 5794 at 4Ghz. So the A9 is closer to a 1.6Ghz Haswell than it is a 4Ghz Haswell. This also completely ignores the GPU.
If the goal is to put the A-series on the desktop, it has to hit 20,000 on the multicore with 4 cores while beating 7000 on the singlecore. We are less than halfway there. Apple could hit the multicore score right now if they created an 8-core model, but that would still not hit the required performance for a desktop. For a laptop it's only comparable to the bottom-of-the-line 1.6 to 2.0 ghz laptops.
And, Apple has been moving towards CPU independence for some time...
http://www.eejournal.com/archives/articles/20150701-bitcode/ ;
If they continue down this path, it will eventually be difficult for developers to write CPU specific code. At that point, Apple could switch out the underlying hardware architecture at will. Given the depth and breadth of Apple's CPU/GPU design expertise, it's not unthinkable that they could actually field their own CPU and GPU free of anyone else's IP. The licensing fees and product margins currently being paid to others (ARM, Intel) would be retained by Apple. The economics may not favor such a move, but it's not unthinkable.
With Intel's recent taking of an ARM license in order to support a foundry role for large customers, it seems Apple stands to gain from consolidating down to one CPU/GPU architecture and playing multiple silicon vendors off each other.