iPhone 18 Pro: Everything M6 Tells Us About the A20 Chip

[MacRumors]

Apple announced the M6 chip last week, just ahead of its September iPhone event, providing the clearest indication yet of what the A20 will bring to the iPhone 18 lineup.

Apple silicon generations share much of their architecture across the A-series and the M-series, so a new A-series chip normally functions as a partial preview of the M-series chip that follows it. This year the order reversed; the M6 was revealed on August 25, ahead of the A20 Pro expected in the iPhone 18 Pro this month.

Recent reports place the A20 Pro at up to 18% faster and up to 30% more power efficient than the A19 Pro, which is broadly similar to the M6 over the M5.

Apple's First 2nm iPhone Chip

The M6 is Apple's first chip built on TSMC's 2nm (N2) process, with the second almost certain to be the A20. Being a smaller manufacturing process, N2 has higher transistor density, allowing Apple to pack more performance into a smaller area.

Apple has historically debuted new nodes on the A-series, since iPhone volume validates yields, so introducing N2 in an $899 Mac mini represents a change. As such, reports that the A20 and A20 Pro will move to N2 are now supported by a shipping product rather than analyst predictions. The consequence for the iPhone event is that the "Apple's first 2nm chip" claim is no longer available.

Dual Neural Engine?

The iPhone's Neural Engine has remained at 16 cores and approximately 35 trillion operations per second since the A17 Pro, spanning three generations. The M6 introduces a Dual 16-core Neural Engine, which Apple says system frameworks can address simultaneously for up to 2x the peak compute of previous generations.

The change corresponds with the reported move to WMCM packaging, which integrates memory onto the same wafer as the CPU, GPU, and Neural Engine to shorten the path between them. It is possible that die area and thermal constraints may limit a dual engine to the A20 Pro.

GPU AI Gains Are Now Incremental

Apple said the M5 delivered over 4x the peak GPU compute for AI compared to the M4, reflecting the introduction of Neural Accelerators within each GPU core. The M6 is rated at nearly 30% over the M5.

The A19 Pro recorded a comparable 4x improvement on the iPhone 17 Pro last September for the same architectural reason. Second-generation Neural Accelerators consequently appear to represent a refinement rather than another step change, suggesting the A20 Pro will not repeat the A19 Pro's headline figure.

Graphics Improvements Should Carry Across

The M6 adds 50% increased geometry rates, an updated shader core architecture, revised Dynamic Caching, and hardware-accelerated ray tracing. These are architectural changes and comparable graphics work has historically appeared across both the A-series and M-series within the same generation.

Memory Bandwidth Suggests a Modest A20 Gain

Memory bandwidth determines how quickly data moves between the memory and the processor, and it is the main constraint on running large language models on device. The M6 provides up to 170GB/s, which Apple describes as a 10% increase over the M5's 153.6GB/s.

That increase is small enough to indicate Apple is using a faster grade of the same LPDDR5X memory rather than moving to the newer LPDDR6 standard, which would have delivered a far larger jump. Reports indicate Apple is sourcing high-speed LPDDR5X from Samsung for the iPhone 18 lineup, and the M6 is consistent with that.

The implication for the A20 is that memory bandwidth is unlikely to improve dramatically. The A19 Pro is rated at 76.8GB/s, and a similar 10% uplift would take the A20 Pro to roughly 85GB/s. Any meaningful gain in how quickly the iPhone 18 Pro loads and runs on-device models is therefore more likely to come from WMCM packaging shortening the distance data travels than from the memory itself.

AV1 Encode Remains Absent

The M6 Media Engine supports hardware-accelerated H.264, HEVC, ProRes, and ProRes RAW, with AV1 decode but no AV1 encode, matching the M5. The omission on a new process node in a desktop chip makes its inclusion in the A20 Pro highly unlikely.

Article Link: iPhone 18 Pro: Everything M6 Tells Us About the A20 Chip

 

[BorisDG]

The A20 will likely focus on CPU upgrades, just like the A18 did. Since the A19 was the GPU leap, the next major graphics architectural change won't come until the A21 with it's Apple 11. Expect the A20 to stick with the Apple 10 GPU architecture introduced in the A19 Pro and M5.

But it seems the AI is now the main focus for Apple - 96bit memory bus + bigger NPU.

 

[chars1ub0w]

iPhones are so fast nobody claims a sluggish device. The big thing is power efficiency. Phones run out of juice before the day is up so often that in some countries power bank rentals are everywhere.

 

[Nyx1]

What if A20 architecture soc the same with M7 in the future but also imagine Apple finally using lpddr6 rather they lpddr5x?
Using lpddr6 with 6-channel for A20 series will be amazing but the leaks already say lpddr5x sad but really hope got more core for CPU and GPU especially GPU 2nm already have more transistor density but always stagnant because the same way architecture without get more core meanwhile m series get more core even putting it on iPad pro.

 

[bigglow]

Looking forward to the upcoming Macbook Neo A20

 

[73CortinaV8]

I heard there's a dedicated on-die controller specifically for the Dark Cherry feature

 

[name99]

The A20 will likely focus on CPU upgrades, just like the A18 did. Since the A19 was the GPU leap, the next major graphics architectural change won't come until the A21 with it's Apple 11. Expect the A20 to stick with the Apple 10 GPU architecture introduced in the A19 Pro and M5.

But it seems the AI is now the main focus for Apple - 96bit memory bus + bigger NPU.

Nah, I don't agree. In fact this will be, I think, the single most interesting HW element of the intro - how will the chip be positioned.

My thinking is two-fold.
On the one hand, AI/LLM capabilities are clearly going to be the driving element of compute for at least the next few years. Not just in the sense of "use your phone to access an LLM" but in the sense that over the next few years we will evolve our way to a new UI and a new set of expectations for how app act and interact, all built around AI. We're at a position somewhat like the first iPhone of 2007, where a vast new world of options have opened up relative to the 2006 Apple world of mac and desktop-based UI.
All this means that over the next few years it makes sense to spend most of the area and power budget that opens up on the SoC for better AI capabilities. (And to be somewhat flexible in those capabilities, eg pursue both the ANE and GPU tracks, because we have no certain idea how the LLM algorithms will evolve over these next few years.)

On other other hand, after 15+ years of other CPU designers seriously lagging Apple, they appear to have, not exactly caught up, but at least to be taking Apple seriously and adopting most of the ideas that made Apple's CPUs so impressive. So in the recent past we've seen some competitive ST performance numbers from both Xiaomi and Mediatek.
How will (and should) Apple respond to this? They could treat this as an important race and pour resources (and area, and power) into it and stay substantially far ahead. But I don't think that, at this moment in time, that's the sensible direction. Instead, for now, work on providing the best local AI experience, which will mostly mean GPU and ANE (though, who knows, it could soon mean strange unexpected consequences like adding many more tiny CPU cores, if it turns out that phone agents really like to spawn lots of small short-lived CPU processes?).
So if Apple appreciates that, at this moment in time, it doesn't make sense to get into a CPU-pissing match with China and Taiwan, then the section of the intro that always talks about how great the new SoC is will need to be revisited to spend much less time on how great the new CPU is (maybe just some stuff about power and security) and much more time on how great the new ANE and GPU are.

The point is not that the CPU will now not be the best in the world; that will still be true.
The point is rather to start the mental repositioning, to get Apple customers, developers, and the whole world to appreciate and believe that the most important features of phone HW are no longer the CPU performance but some sort of "integrated AI" performance, ie to get people to understand the axis along which Apple will be competing (and positioning itself as number one) for the next few years.

 

[BorisDG]

What if A20 architecture soc the same with M7 in the future but also imagine Apple finally using lpddr6 rather they lpddr5x?
Using lpddr6 with 6-channel for A20 series will be amazing but the leaks already say lpddr5x sad but really hope got more core for CPU and GPU especially GPU 2nm already have more transistor density but always stagnant because the same way architecture without get more core meanwhile m series get more core even putting it on iPad pro.

My guess is that LPDDR6 will debut with A21. I think there is big chance that they do the same thing like M6 ie adding "Super Core" and "Performance Core". Let's say 1S/1P/4E cores config vs old 2P/4E setup.

 

[IIGS User]

The M/A processors really are marvels of engineering. The amount of compute power is staggering.