[glw]

Originally Posted by Exophase

A6 being really close to their projected 32nm A5X size seems to be a good argument that it's a 32nm A5X, with a lot more CPU clock speed. I don't think they realize that Cortex-A15s are probably substantially bigger than Cortex-A9s.

IIRC ARM's own ballpark figure is twice the size in the same process.

[tangey]

Originally Posted by Pressure

Would you feel better if I called it S5L8942 then?

I wasn't feeling either good or bad about what you called it. Rather you stated that apple called it A5R2, whereas I have seen no evidence that apple gave it any different consumer name than the 45nm A5, which if true confirms that their consumer naming system means diddly squat in terms of the technology therein.

Hence my original point that many are just putting too much faith in what is an arbitrary naming system.

[ltcommander.data]

Originally Posted by Jubei

Interesting. They exclude quad-cpu but they never entertain the thought of dual A9s clocked high.

A6 = 22% smaller than 45nm A5

32nm A5 = 41% smaller than 45nm A5

32nm A5X = theoretically 21% smaller than 45nm A5

Based on this we can exclude a simple overclocked 32nm A5 no? Because either GPU or the CPU has to be bigger in A6 to be only 22% smaller

I agree simply an overclocked 32nm A5 doesn't seem to work. The die sizes work out to be:

122.6 mm2 for the 45nm A5
69.6 mm2 for the 32nm A5
165 mm2 for the 45nm A5X
93.7 mm2 for a 32nm A5X (assuming the same scaling factor that the A5 achieved)
95.6 mm2 if the A6 is 78% the size of the A5.

There would seem to be too much extra area for the A6 to simply be largely a 32nm A5 with doubled clock speeds for the CPU and GPU. The 32nm A5 is only 73% the size of the A6. Apple confirmed a new ISP and a new memory controller to support LPDDR3 is probably a safe bet, but would those really add that much extra space compared to the 32nm A5?

The A6 is very close in size to a 32nm A5X and it's quite possible Apple swapped the 4x32-bit LPDDR2 memory controller for a 2x32-bit LPDDR3 memory controller which should give some space savings for the improved ISP and other tweaks. Previous Apple SoC have usually been thought to use a 4x clock speed differential between the CPU and GPU so this would be the first time they move to a 8x divider assuming the CPU is clocked at 1.6GHz while the SGX543MP4 needs to remain at 200MHz to match Apple's claim that GPU performance only increased 2x.

My personal theory preference would be that A5/A5X to A6 does mean something more generational than just clock speed differences and since the 4x CPU/GPU clock speed difference has always been in place, my default thought is that it'll remain so although I have no justification or advantage for that other than momentum and that it limits the possibilities of what the A6 can be so I can settle on an estimate. I also don't think Apple is roping in flash I/O improvements when they claim a "2x faster CPU" rather that really does refer to CPU itself. Assuming those limitations, my guess is that the A6 is a 1.2GHz Cortex A15 with either a 300MHz SGX543MP3 or a 300MHz SGX554MP2 and a 2x32-bit LPDDR3 memory controller. That could satisfy the 2x CPU speed, 2x graphics speed, die size differences, A6 meaning a generational change, and a 4x CPU/GPU clock speed difference.

[silent_guy]

Originally Posted by zed

Im surprised that apple shares went up 3%. In all honesty this phone unveil is prolly less exciting unveil than the 4s (which wasnt that great, though it did have siri which made the masses go ohhh).

If you look at the pure HW specs, the 5 is a bigger jump than the 4S in term of features that are noticeable to the user. (LTE, screen, wifi)

I wonder how it will sell? samsung s3 was 20million in ~3 months, if it doesnt ~double that we may be in the process of the crown being passed.

Everybody around me plans upgrade his 4 to a 5. Just like most 3GS users upgraded to the 4S. But there are way more 4 users than there were 3GS. It should do just fine.

[anexanhume]

Originally Posted by ltcommander.data

I agree simply an overclocked 32nm A5 doesn't seem to work. The die sizes work out to be:

122.6 mm2 for the 45nm A5
69.6 mm2 for the 32nm A5
165 mm2 for the 45nm A5X
93.7 mm2 for a 32nm A5X (assuming the same scaling factor that the A5 achieved)
95.6 mm2 if the A6 is 78% the size of the A5.

There would seem to be too much extra area for the A6 to simply be largely a 32nm A5 with doubled clock speeds for the CPU and GPU. The 32nm A5 is only 73% the size of the A6. Apple confirmed a new ISP and a new memory controller to support LPDDR3 is probably a safe bet, but would those really add that much extra space compared to the 32nm A5?

The A6 is very close in size to a 32nm A5X and it's quite possible Apple swapped the 4x32-bit LPDDR2 memory controller for a 2x32-bit LPDDR3 memory controller which should give some space savings for the improved ISP and other tweaks. Previous Apple SoC have usually been thought to use a 4x clock speed differential between the CPU and GPU so this would be the first time they move to a 8x divider assuming the CPU is clocked at 1.6GHz while the SGX543MP4 needs to remain at 200MHz to match Apple's claim that GPU performance only increased 2x.

My personal theory preference would be that A5/A5X to A6 does mean something more generational than just clock speed differences and since the 4x CPU/GPU clock speed difference has always been in place, my default thought is that it'll remain so although I have no justification or advantage for that other than momentum and that it limits the possibilities of what the A6 can be so I can settle on an estimate. I also don't think Apple is roping in flash I/O improvements when they claim a "2x faster CPU" rather that really does refer to CPU itself. Assuming those limitations, my guess is that the A6 is a 1.2GHz Cortex A15 with either a 300MHz SGX543MP3 or a 300MHz SGX554MP2 and a 2x32-bit LPDDR3 memory controller. That could satisfy the 2x CPU speed, 2x graphics speed, die size differences, A6 meaning a generational change, and a 4x CPU/GPU clock speed difference.

If they were willing to do 554MP2 here, why not on iPad 3?

[Rys]

Originally Posted by anexanhume

Actually, looking at GPU, it has to be frequency doubled 543MP2 or die shrunk MP4. Minimum Rogue performance should be 5X from MP2 going by triangle count alone.

Or an 800 MHz 543MP1, or....

[ltcommander.data]

Originally Posted by anexanhume

If they were willing to do 554MP2 here, why not on iPad 3?

The iPad 3 needs to push a lot more pixels than the iPad 2 so moving from a SGX543MP2 to SGX543MP4 configuration works well because everything scales up together. The SGX554 doubles the ALUs without doubling the TMUs or ROPs which would seem to fit well with the iPhone 5 since the pixel count hasn't increased that much against the iPhone 4S. I was thinking a 300MHz SGX554MP2 in the A6 vs a 200MHz SGX543MP2 in the A5 would give a 3x increase in ALU performance, but only a 50% increase in TMU and ROP performance which may result in an effective overall performance increase around 2x to match Apple's claims. However, marketing-wise even if only the ALUs were 3x faster I would have expected Apple to advertise that 3x figure. Wasn't Apple's 7x graphics performance claims between the iPhone 4 and iPhone 4S heavily based on improved ALU performance? That they didn't and only claimed 2x for the A6 would seem to make a SGX554MP2 less likely.

[anexanhume]

Originally Posted by ltcommander.data

The iPad 3 needs to push a lot more pixels than the iPad 2 so moving from a SGX543MP2 to SGX543MP4 configuration works well because everything scales up together. The SGX554 doubles the ALUs without doubling the TMUs or ROPs which would seem to fit well with the iPhone 5 since the pixel count hasn't increased that much against the iPhone 4S. I was thinking a 300MHz SGX554MP2 in the A6 vs a 200MHz SGX543MP2 in the A5 would give a 3x increase in ALU performance, but only a 50% increase in TMU and ROP performance which may result in an effective overall performance increase around 2x to match Apple's claims. However, marketing-wise even if only the ALUs were 3x faster I would have expected Apple to advertise that 3x figure. Wasn't Apple's 7x graphics performance claims between the iPhone 4 and iPhone 4S heavily based on improved ALU performance? That they didn't and only claimed 2x for the A6 would seem to make a SGX554MP2 less likely.

543MP3 at 266?

Brian Klug is saying MP3 BTW.

[ltcommander.data]

Originally Posted by anexanhume

543MP3 at 266?

Brian Klug is saying MP3 BTW.

My mindset was always on a 4x clock speed differential between CPU and GPU so my original suggestion was SGX543MP3 @ 300MHz. Looks like I posted my speculation before Brian's too (7:10 pm EST vs 6:19PM PST). Nice to know my thoughts weren't completely unreasonable.

http://twitter.com/nerdtalker/status/246417961294909440

[Ailuros]

Originally Posted by ltcommander.data

My mindset was always on a 4x clock speed differential between CPU and GPU so my original suggestion was SGX543MP3 @ 300MHz. Looks like I posted my speculation before Brian's too (7:10 pm EST vs 6:19PM PST). Nice to know my thoughts weren't completely unreasonable.

http://twitter.com/nerdtalker/status/246417961294909440

There's nothing unreasonable about it in theory. MPs can scale in odd numbers and 300MHz should consume still less power than a MP2@400MHz f.e. Albeit if you'd want exactly twice the theoretical GPU performance of a MP2@200MHz, you'd rather go for a MP3@267MHz.

[Mariner]

An interesting iPhone 5 limitation which may be of interest to some American users:

http://www.anandtech.com/show/6295/w...-svdo-support-

I'm not exactly sure how much of a show-stopper this might be for most users but it appears it is another downside to the downsize.

[ltcommander.data]

Originally Posted by Mariner

An interesting iPhone 5 limitation which may be of interest to some American users:

http://www.anandtech.com/show/6295/w...-svdo-support-

I'm not exactly sure how much of a show-stopper this might be for most users but it appears it is another downside to the downsize.

I'm guessing most CDMA iPhone 5 users would be existing CDMA 3G users so it's kind of a psychological question of can you miss something you never had? Being a GSM user, although not a previous iPhone, I'd be annoyed without simultaneous voice and data. I don't use it all the time, but when I do use it, it's because it's natural to do so. Maybe existing 3G CDMA users have a usage paradigm where they wouldn't normally try simultaneous voice and data so it won't be such a big deal for them.

One way lack of simultaneous voice and data could be a big problem is if it breaks other iOS functionality, specifically turn-by-turn directions. I wonder how much map data/turn-by-turn directions Apple caches locally. If someone were driving with turn-by-turn and then took a phone call, hands-free of course, would turn-by-turn directions just cutoff which could leave the driver missing the next turn and staying on the same road oblivious? That would be a major problem.

[silent_guy]

I was about to blast Apple for this, how it is totally unacceptable, but then I realized that I've probably used the data during voice feature maybe 10 times in my 5 years of iPhone use. So I'll live.

[seeingwhite]

EE Times is now claiming the A6 is "a dual-core Cortex-A15 manufactured for Apple by Samsung Electronics Co. Ltd. in its 32-nm HKMG manufacturing process." They site Nomura Equity Research, but go on to say that Nomura Equity Research did not disclose the source of their "information."

http://eetimes.com/electronics-news/...ore-Cortex-A15

[Helmore]

Originally Posted by silent_guy

I was about to blast Apple for this, how it is totally unacceptable, but then I realized that I've probably used the data during voice feature maybe 10 times in my 5 years of iPhone use. So I'll live.

Going by what you're saying I assume you're on AT&T, which means that you won't even have this problem. You can do voice and data simultaneously on the AT&T iPhone 5. Or you're not even living in the US, which means this issue won't even come up.

[silent_guy]

Originally Posted by Helmore

Going by what you're saying I assume you're on AT&T, which means that you won't even have this problem. You can do voice and data simultaneously on the AT&T iPhone 5. Or you're not even living in the US, which means this issue won't even come up.

I'm switching to Verizon.

[Rys]

After some poking around with the new toolchain shipped with Xcode 4.5, it looks like the architecture target for A6 is armv7s, not armv7. If I compile for armv7s and ask the toolchain to just run the preprocessor and stop, I can see it defining support for VFPv4.

So that means Cortexes A5, A7 or A15 but definitely not (unless I'm missing something) Cortex A9.

[wco81]

Or could it be the A15 and A7? Isn't that what big.little is suppose to be?

[Rys]

Seems unlikely to be big.LITTLE if you assume they'd make more of a splash about something as architecturally advantageous as that. Then again, I'd have assumed they'd make a splash about not using A9 regardless.

[wco81]

IIRC, they didn't note the A9 cores either when they first moved to it.

Odd that they're so opaque about the SOC. They cited sRGB and they'll mention all the tech specs of the camera sensor.

[ltcommander.data]

Originally Posted by wco81

IIRC, they didn't note the A9 cores either when they first moved to it.

Odd that they're so opaque about the SOC. They cited sRGB and they'll mention all the tech specs of the camera sensor.

Maybe they feel the display and camera are more user facing components. Most consumers aren't going to understand what those specs mean, but they are going to experience/utilize them (see the wider gamut display, the change in focal length, etc.) so it's worthwhile to point out the specific details. Specific SoC specs are more important to and more directly utilized by developers, who may code differently to take advantage of the new VFP for example, whereas all users need to know is the general specs that the end result is their apps can run 2x faster or there is more power available for more complex graphics.

[anexanhume]

Originally Posted by wco81

Or could it be the A15 and A7? Isn't that what big.little is suppose to be?

A7 is too new to be in any IP.

Originally Posted by Rys

Seems unlikely to be big.LITTLE if you assume they'd make more of a splash about something as architecturally advantageous as that. Then again, I'd have assumed they'd make a splash about not using A9 regardless.

Brian Klug says you're bang on: https://twitter.com/nerdtalker/statu...96538104410112

[Exophase]

Guess that mostly settles it. Let's hope that the performance figures are due to Apple clocking the CPU a lot lower than the 1.5-2GHz we've been led to expect in upcoming SoCs, and not due to the IPC improvement being far below the expected. If it were anyone other than Apple releasing this I'd be concerned, but they have a history of low clock speeds in phones.

[Steven Lin]

XCode 4.5 will now default build binary with armv7/armv7s architecture which means they have a new cpu core other than Cortex A9.

armv7 ==>Cortex A8, Cortex A9
armv7f ==>Cortex A9MP
armv7s ==> Cortex A15.

[Ailuros]

Actually it's a very wild guess but I'd even prefer a custom designed "next generation" CPU than A15.