Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Good summary, however there are a few errors

- A5X has 1gb ram, not 512mb
- A6 gpu clock is 325mhz (1/4 CPU clock)
- A6X gpu clock is 280mhz (1/5 CPU clock)

As for predictions (IMO):
- A7 and A7X CPU clock should be higher than A6/A6X
- Not sure but the A7X GPU clock seems too high

Thanks for A5X correction (I actually had 512GB, woops).

Can you provide sources for GPU frequencies? I pulled them from anandtech reviews but I know they tend to bounce around because they're extrapolated.

For the A7 and A7X, a modest clock increase is probably likely, but they've also gone without increasing clocks before, so I left it the same to be conservative, assuming IPC improvements would pull the weight.

For the A7X, yeah the GPU clock likely won't be that high. I expect somewhere in the range of 400 and up though, with at least a G6430.

Thanks for catching the errors!
 
I have a feeling that most users will switch to Android devices if Apple releases the iPhone 5S with an A7 Dual-core chip

MOST users? That's quite an assumption.

And of course it doesn't matter how smooth or fast the iPhone 5S is, because unless it says "quad-core" on the spec sheet thats it I'm switching to Android! :rolleyes:
 
For the sections I have skimmed through, it was a damn good read. I also liked the part of Apple not falling into the spec wars and the reasons behind it. Thoroughly enjoyed it.
 
Well alrighty then. Helluva read bro.

Very well articulated and, well, technical.

Well done!
 
Not addressed by the op is the probable upgrade from the Broadcom BCM4334 to the BCM4335 to add 802.11ac.
 
Thanks for A5X correction (I actually had 512GB, woops).

Can you provide sources for GPU frequencies? I pulled them from anandtech reviews but I know they tend to bounce around because they're extrapolated.
Read them in the beyond3d forums. Basically the frequencies I listed matched up exactly with the GLBench primitive benchmarks. In any case, it can't be 333Mhz and 300Mhz - Apple GPUs are always an exact integer division of the CPU frequency - and we know what the CPU frequencies of the A6 and A6x are

For the A7 and A7X, a modest clock increase is probably likely, but they've also gone without increasing clocks before, so I left it the same to be conservative, assuming IPC improvements would pull the weight.

I think 1.5-1.7 Ghz is likely up from 1.3-1.4Ghz.

For the A7X, yeah the GPU clock likely won't be that high. I expect somewhere in the range of 400 and up though, with at least a G6430.

Thanks for catching the errors!

I can believe approx 450Mhz. 600Mhz is way too high and doubt it would make it into the iPad5 especially since we know the ipad5 is significantly thinner/lighter than the iPad4.
 
What to expect from the iPhone 5S, Apple's A7 and beyond

I get the feeling that Apple not only aren't pushing the envelope as far as they once did, but have run out of envelopes too! Given the exponential increase in competition and capability each year from Samsung and HTC, I think Apple need to cease and desist with this S moniker, pull their fingers out from their collective rear ends and produce something as stunning as the original iPhone was compared to the rest of the mobile market.

I fear tomorrow's announcement will be a case of relative stasis to keep the dollars and cents rolling in until the iPhone 6 is ready to once again produce something worthy of the keynote marketing spin.
 
Read them in the beyond3d forums. Basically the frequencies I listed matched up exactly with the GLBench primitive benchmarks. In any case, it can't be 333Mhz and 300Mhz - Apple GPUs are always an exact integer division of the CPU frequency - and we know what the CPU frequencies of the A6 and A6x are

Ok, I'll double check against that thread. What's your username? I'm anexanhume over there (same as twitter handle).

I think 1.5-1.7 Ghz is likely up from 1.3-1.4Ghz.

Perhaps. TSMC's 28nm HPM process is looking very good on power use for higher frequencies, so we'll see if Samsung's process can match it on that (though it's likely still way behind on transistor density).

I can believe approx 450Mhz. 600Mhz is way too high and doubt it would make it into the iPad5 especially since we know the ipad5 is significantly thinner/lighter than the iPad4.

We still don't know what Rogue's scaling looks like vs frequency. I've seen talk that it's meant to operate at much higher frequencies than 5XT, but we also have a larger, slower clocked GPU in the A6X versus A6, so nothing would surprise me.

edit: Cleaned up and corrected the GPU frequencies for A6/X. Wasn't able to find a good source on the 320 MHz for A6, but I know it's correct after you pointed out the direct integer multiple. It makes sense using a single oscillator is possible. Thanks for clearing that up.

edit 2: Thanks again everyone for your corrections and questions. Gives me ideas how to expand, elaborate and what to look into more. I'd like to keep this as a template and repeat it next year. I plan to update the post with how well the predictions went too. I partly did this for myself as I wanted a centralized point of info with sources if I ever wanted to remember something about previous Apple SoCs.
 
Last edited:
Any idea if Apple will choose the single radio interface at a time path that they did with the 5 last year? I really hated losing my data connection while i was on the phone with my 5...
 
Any idea if Apple will choose the single radio interface at a time path that they did with the 5 last year? I really hated losing my data connection while i was on the phone with my 5...

Not sure what you're asking. GSM carriers like AT&T have been able to do this for a while. Verizon, and CDMA in general, cannot because they never implemented the SVDO part of the CDMA standard that would have allowed it. They do, however, plan to implement VoLTE, which would provide the feature you're asking for. Qualcomm's radios support it, so it would simply be a matter of Apple implementing it.

------------------

There's a few more things I wanted to cover before tomorrow's event.

WiFi
I neglected to go over WiFi because I thought there was a fair chance it would retain the Broadcom BCM4334, although a strong possibility remained that Apple would upgrade to the Broadcom BCM4335, which is essentially the same chip with ac added (same process node @45nm, same feature set except addition of ac). Although support is low in the industry, apple's newest airport extreme router supports it, and the recent macbook airs also support it. Though scenarios that utilize the peak bandwidth of ac are likely to never occur on the iPhone, ac does have increased efficiency over N which could save battery life. Keep an eye out for ac support mentioned in the keynote and the Broadcom BCM4335 in the ifixit teardown.

Display
I've already covered how advanced the iphone 5 display is and how any improvements would likely be towards power savings (via IGZO transistors or GRAM). There has actually been some more information come to light on the GRAM angle. It turns out that Motorola's G2 is not the only phone to feature this. Motorola's new Moto X also features this, as well as the new Nexus 7 tablet. If you follow this series of tweets, you'll see a discussion between anandtech writers and a qualcomm engineer about the benefits of this method.

The qualcomm engineer points out that the benefit isn't as great as it would seem because android now supports something called "composition bypass." Composition bypass allows openGL apps to draw directly to the frame buffer (the memory that will be written to the display), rather than write to a surface still retained on the software side, which is then written to the framebuffer. This allows the apps to write to only the updated memory locations in the framebuffer rather than having the surface update the entire framebuffer. This just means that you're writing less memory, which saves power. In a full GRAM scenario, the write to the framebuffer would never happen if there was an update, and the display would simply self-refresh from its local GRAM. I don't know if Apple has a similar API that implements a similar framebuffer update paradigm, regardless of their aspirations for display RAM.

I also didn't mention the use of in-cell technology (which laminates touch layer directly to LCD instead of having an intermediary glass layer) because it is expected to be retained for the iphone 5S.

However, there have been developments in cover glass. Since the launch of the iphone 5, which uses gorilla glass 2, gorilla glass 3 has launched. Of course, it boasts improved damage resistance and scratch resistance. This means either can retain glass thickness and get a tougher screen, or it make the screen even thinner and retain the same level of strength.

There have also been rumors of Apple using sapphire crystal instead of glass for its displays. While Macrumors notes that it is considerably more expensive than gorilla glass, it does boast higher scratch resistance, albeit with less strength. There is an argument to be made that strength is eclipsed by scratch resistance once it reaches a certain level. However, the latest reports suggest that sufficient volume doesn't exist for sapphire crystals as a display material.

Given all of this evidence, it seems as though Apple will stay with Gorilla Glass this generation, likely opting for Gorilla Glass 3. It is also likely Apple will coat the display in an oleophobic coating, something they've done since the iPhone 3GS.

Camera
While I deferred to Macrumors' roundup on the camera rumors, I would be remiss if I didn't mention the possibility of Apple including Optical Image Stabilization (OIS). This is a vibration damping scheme that takes shakiness (and/or blurriness) out of image and video capture. LG's new G2 implements this, with HTC and Nokia also having implementations in their latest smartphones.

OIS, coupled with a windowed still image capture (such as HTC's "Zoe" feature), 120 FPS recording for slow motion, and dual LEDs would make a compelling camera upgrade package for the 5S along with an increased megapixel camera that retains or improves pixel size and retains or improves aperture size.

Other things to watch out for
Look for an update to the PMIC (power management intregrated circuit). This will be an apple branded chip with Dialog silicon IP in it. An update to this could mean a more aggressive asynchronous clock/voltage management between CPU cores, like the Qualcomm solution, that saves power.

As was discussed on the previous page, also look for an update to the Cirrus audio codec, which could allow for higher bitrate music output.

There are a lot of chips I didn't cover like RF power amplifiers and demultiplexers, gyroscope, accelerometer, touch screen controllers, trackpad controllers (used in conjunction with with the touch screen controller).

I'll try to track all of the things that change and give an explanation of the differences between parts once the teardown is complete.

Happy watching!
 
A7X, GPU Performance - 153.6-230.4 GFlops

Just as a comparison, that is awfully close to the xbox360's GPU, of 240 GFlops.
Just a tiny increase, and they could say it's "graphically more powerful than an xbox"...
 
Not sure what you're asking. GSM carriers like AT&T have been able to do this for a while. Verizon, and CDMA in general, cannot because they never implemented the SVDO part of the CDMA standard that would have allowed it. They do, however, plan to implement VoLTE, which would provide the feature you're asking for. Qualcomm's radios support it, so it would simply be a matter of Apple implementing it.

------------------

What I mean is apple's decision to only use one radio interface at a time because of only including one antenna. This makes it impossible to use cdma voice and LTE at the same time, something android phones have been doing since last summer on vzw and Sprint...
 
What I mean is apple's decision to only use one radio interface at a time because of only including one antenna. This makes it impossible to use cdma voice and LTE at the same time, something android phones have been doing since last summer on vzw and Sprint...

Oh, sorry for the confusion. I assumed your question was much more basic than it is :)

Apple did beef up their antenna design between the 4 and 4S, but I doubt they budge on this one. I'm sure they're expecting carriers to get with it and provide VoLTE already. CDMA users have been living without it this long so I'm sure they don't feel any onus to start including a solution that involves CDMA, particularly when most of the rest of the world doesn't use it. Adding more antennas just adds complexity to the design without a clear homerun on winning users.

Just as a comparison, that is awfully close to the xbox360's GPU, of 240 GFlops.
Just a tiny increase, and they could say it's "graphically more powerful than an xbox"...

Yes, and as GPU architectures get more efficient, it's easier to expose the hardware, so a "Rogue" 240 GF is actually better than a Xenos 240 GF.
 
Oh, sorry for the confusion. I assumed your question was much more basic than it is :)

Apple did beef up their antenna design between the 4 and 4S, but I doubt they budge on this one. I'm sure they're expecting carriers to get with it and provide VoLTE already. CDMA users have been living without it this long so I'm sure they don't feel any onus to start including a solution that involves CDMA, particularly when most of the rest of the world doesn't use it. Adding more antennas just adds complexity to the design without a clear homerun on winning users.



Yes, and as GPU architectures get more efficient, it's easier to expose the hardware, so a "Rogue" 240 GF is actually better than a Xenos 240 GF.

I thought so as well. Really sucks though.
 
I thought so as well. Really sucks though.

I agree. Apple is very good downplaying the significance of features they don't have though. Last year Schiller proudly announced iPhone 5 had sv&d while glossing over the fact it still wasn't true for CDMA phones.
 
I agree. Apple is very good downplaying the significance of features they don't have though. Last year Schiller proudly announced iPhone 5 had sv&d while glossing over the fact it still wasn't true for CDMA phones.

Exactly, as do I. I also remember saying I wouldnt get it but i ended up getting it anyway and keeping for 7 months (long for me)...and i still will this year. Just cant stay away. :apple:
 
Well damn, i guess we got a 64 bit phone ahead of everybody else...:eek:

Yes, I was very surprised as was anandtech, even with the rumors.

@jonst0kes has a lot of good tweets on the topic, pointing out that increasing the register allocation can really increase power consumption, and it's also a question of compiler optimization. At some point, more registers is diminishing returns.
 
edit: nevermind. Audio specs seem the same as the previous iphones. We'll have to see the teardown to know if it's any different.

edit 2: To make this post useful, here's the standard features of ARM's vanilla 64 bit architecture. Sounds like Apple followed it pretty closely on register numbers. We'll have to see how custom it is.

http://en.wikipedia.org/wiki/ARMv8#ARMv8_and_64-bit

AArch64 features:

New instruction set, A64
Has 31 general-purpose 64-bit registers.
Separate dedicated SP and PC.
Instructions are still 32 bits long and mostly the same as A32 (with LDM/STM instructions and most conditional execution dropped).
Most instructions can take 32-bit or 64-bit arguments.
Addresses assumed to be 64-bit.
Advanced SIMD (NEON) enhanced
Has 32 × 128-bit registers (up from 16), also accessible via VFPv4.
Supports double-precision floating point.
Fully IEEE 754 compliant.
AES encrypt/decrypt and SHA-1/SHA-2 hashing instructions also use these registers.

edit 3: Apple's "M7" chip is likely based on ARM's own Cortex M series, used in microcontrollers and for a lot of fitness applications (they are in the FitBit, for example). TI's OMAP 4 and 5 series SoCs feature them too (the new Motorola X has Cortex M processors too).

http://en.wikipedia.org/wiki/ARM_Cortex-M

example: Freescale's "Kinetis" line: http://en.wikipedia.org/wiki/Freescale#Kinetis

Also, no 128GB, LTE advanced or WiFi ac. None of this is necessarily shocking. You can bet the iPhone 6 will have LTE-A and ac WiFi though.

edit 4: With 2X GPU claim, we're talking about a Roge G6430 at about 250 MHz. G6230 would be double the frequency which is too high and G6630 would just be too big. I think 2X rules out any idea that they're staying with Series 5 anymore. Has to be rogue because of openGL ES 3.0 support.

Also, the die size is now 102 mm, up 6 mm over the A6. This is a small change if the process stayed 32nm, and a modest one if the process moved to 28nm. Without knowing Rogue core sizes, it's hard to tell where the die space is being divided up between custom CPU cores and this GPU cores. Also not sure if anything has moved off of the SoC into the M7 coprocessor. I assume all of their gyro and accelerometer chips have moved into the M7, though, saving board space.

edit 5: to get to exactly 2x on GPU, G6430 would be 270 MHz. Given that CPU clock speeds must be integer multiples (thanks thunng8!), we're looking at 1.35 GHz for 5x or 1.62 GHz for 6x. I think the latter is much more likely, given that 2x on CPU would be very hard to reach on architectural improvements alone. The clock jump on A6 from A5 was 62% alone. This would only be a 25% clock jump. 7x would be 1.89 GHz, which is quite high for Apple who has lagged in CPU frequency since the 4S.

edit 6: After talking with some other people, even with a 1.6 GHz clock speed, I think the 2x number is just too big for it to still be dual core. That's still a 60% jump when you adjust for clock speed. We're probably in quad (or maybe just triple) core territory. Asymmetric core count for CPU seems unlikely though. There was also discussion of GPU clocks not being locked to CPU clocks, which is likely given separate PLLs are a possibility. I was hesitant to put that idea forward without knowledge to the contrary, but it seems pretty evident when you consider these parts can usually ramp independently for any semblance of a smart thermal/power profile.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.