Apple Silicon Macs - Apple's ARM SoC Tech Explained!

Jacob Jacob 21 July 2020
Apple Silicon Macs - Apple's ARM SoC Tech Explained!

I’m getting a lot of comments where people aren’t understanding the benefits of Apple Silicon, with some people claiming that they’ll ditch Apple once they transition away from Intel.

What a lot of people don’t understand is that Apple’s transition isn’t just about switching to ARM-based processors, there’s actually a lot more to it that will greatly benefit Apple’s future Macs. So in this article, I’m going to dive deep into how their upcoming SoC technology is gonna work, how it compares to current x86 processor technology, and why it makes sense for Apple.

Apple Silicon Macs - Apple's ARM SoC Tech Explained!Currently, Intel-based Macs like the 16” MacBook Pro have a CPU, a dedicated GPU, memory for each of them, and more recent Macs have a T2 chip which adds things like a storage controller and coprocessors. With Apple Silicon, all of that is going to be built into one system on a chip, or SoC.Apple Silicon Macs - Apple's ARM SoC Tech Explained! The main benefit is that all of those components, from the CPU to the graphics, the memory, the new neural engine and the machine learning accelerators will all be able to communicate with each other much more efficiently compared to before. So let me give you an example from intel-based Macs.

Current dedicated GPUs have to communicate with the CPU using the PCI express bus, which means there is added latency. Apple Silicon Macs - Apple's ARM SoC Tech Explained!I found some white papers from semanticscholar.org discussing this, so here’s a quote from a paper named efficient data communication between the CPU and GPU.

Limited by low bandwidth and high latency interconnects, larger page migration has longer delay, which may reduce the overlap of computation and transmission and cause serious performance decline.

So when there is more latency, like when you’re using PCI express for the GPU, the rest of the system like the CPU has to wait before it can continue processing, reducing overall performance. Here’s a quote from another page titled unlocking bandwidth for GPUs.

In the near future, systems will appear where software page migration is an optional choice and hardware cache-coherence can also support the GPU accessing CPU memory directly.

And this is exactly what Apple is doing with their new ARM-based Macs. Apple Silicon Macs - Apple's ARM SoC Tech Explained!If you look at this old diagram from 2015, you’ll see that at the bottom, the GPU and CPU have a cache coherent link which raises the bandwidth from 16GB/s to 100Gb/s, except in Apple’s case, the GPU will now have direct access to system memory on top of all of that. Apple Silicon Macs - Apple's ARM SoC Tech Explained!So basically, Apple’s thought process is that if they can greatly reduce the latency in communication between all of these components, then each component has less of a wait time before it can start its specific task, and system performance in general will go up thanks to increased efficiency. And this ultimately means that you can get better real-world performance with less RAW power, and that’s before accounting for software optimization, which Apple Silicon will greatly benefit from since Apple will directly control both the hardware and the software for the first time ever on a Mac.

That’s the exact same reason why the iPhone is the fastest smartphone in the world, while also getting great battery life. And this is also why our 2020 iPad Pro is now over 20% faster in terms of graphics performance just by upgrading to iPadOS 14 beta, so control over both the software and the hardware means incredible optimization.

People are comparing the A12Z chip from the 2020 iPad Pro to current Intel chips to see if it’s worth it, but this is actually a mistake. Apple Silicon Macs - Apple's ARM SoC Tech Explained!First of all, the processor side of the A12Z is basically unchanged from the A12X from 2018, since Apple chose not to improve it again since the competition is still nowhere near the 2018 iPad Pro. That chip is made using 7 nm tech built by TSMC, the leader in ARM chip manufacturing. Apple’s upcoming Mac family of SoCs will actually use the 5 nm fabrication process, and according to TSMC, that should bring a 15% performance improvement per watt. But it doesn’t stop there, TSMC has confirmed that they’ll be shipping 3 nm chips in 2022, and that should give us another 10-15% performance boost per watt.

So if we take the current A12X chip and change nothing at all except for switching out the 7 nm chip for a 3 nm chip, the performance will go from a Geekbench 5 multi-core score of 4,606 to between 5,757 to 5,987.Apple Silicon Macs - Apple's ARM SoC Tech Explained! Those scores are much higher than Intel’s brand new 10 nm i7 quad-core chip from the 2020 13” MacBook Pro.Apple Silicon Macs - Apple's ARM SoC Tech Explained! and that’s coming from a much smaller chip that is passively cooled in an iPad Pro that’s 38% thinner than the iPhone 11 Pro.

What a lot of people don’t realize is that the Apple Silicon chips that will be used in Macs are going to be a brand-new family of chips, which means that Apple is most likely going to scale the size of the chip much larger now that they don’t have to fit it into a thin tablet. And they’re also going to allow it to use more power, since the batteries in Macs are much larger than the ones in the iPad Pro, and they’ll be able to add active fan cooling giving them a lot more thermal heatroom.

However, these Apple Silicon Macs are going to have better battery life since they’ll use asymmetric cores, meaning that they’ll have both high performance and efficiency cores. And those scores from before don’t factor in any other improvements made by Apple which could be memory speed improvements and further optimizations.

So that was just the future potential of the processor side of the Apple Silicon chip. Now let’s get into the graphics side. The graphics will be built-into the main chip, which like I mentioned earlier, will greatly reduce the latency in terms of communication with the CPU and the other components. So like I said, you can get more real-world performance with less RAW graphics power. If you don’t believe me, let me prove it to you.

We recently compared our Mac Pro to a Windows PC in terms of video editing. Apple Silicon Macs - Apple's ARM SoC Tech Explained!As far as RAW performance, you can see that the PC had over twice as much processor performance, and over 60% more graphics performance than the Mac Pro’s AMD Vega 2 GPU, so the PC is clearly much faster. However, when we tested video editing in Davinci Resolve 16 Studio, which is a cross-platform program, the Mac Pro was actually faster at exporting a 5min RAW 6K clip.Apple Silicon Macs - Apple's ARM SoC Tech Explained! And even while stabilizing a 1 minute 4K clip, the Mac Pro was yet again faster than the PC, showing that RAW performance doesn’t always mean more real-world performance.Apple Silicon Macs - Apple's ARM SoC Tech Explained! And this is without considering all of the technology that’s coming to Apple Silicon Macs.

But let me remind you again, the A12Z chip in the super thin iPad Pro currently has about 20% more graphics performance than Intel’s brand new 10th-gen i7 processor in the 2020 13” MacBook Pro.

Just imagine if they scaled the processor up and added the benefits from everything being packed into one main chip, like the unified memory architecture. Some users commented that this shared memory is too slow compared to the current GDDR6 memory in dedicated graphics chips, but they’re completely forgetting that Apple Silicon Macs will have their own dedicated on-chip Tile Memory, which could very well be the incredibly fast HBM2 memory.

So check out this example by Apple showing how developers can optimize the graphics to access the tile memory more often so that it won’t have to access the system RAM as often, reducing the use of DRAM bandwidth. Apple Silicon Macs - Apple's ARM SoC Tech Explained!So Apple’s goal is to streamline graphics processing to free up as much system memory bandwidth as possible, which will reduce the chances of a bottleneck, and then combined with incredibly low latency, every component in the entire system can be utilized much better than on a traditional x86 Intel processor.Apple Silicon Macs - Apple's ARM SoC Tech Explained! Like the T2 chip which currently takes care of HEVC transcoding, which will now have reduced latency since it’s built into the main chip instead of being somewhere else on the motherboard like it is in current Intel-based Macs.

These are the types of improvements that will allow Apple to achieve great real-world performance even if they have less RAW performance due to no longer using large dedicated graphics chips.

Let me finish off with two more major benefits of Apple Silicon, Macs will be even more secure than ever before now that they’re ditching Intel and it’s MANY hardware vulnerabilities like Meltdown, Spectre, Zombieload and much more. And a huge benefit to Apple Silicon is that we no longer have to deal with AMD graphics driver issues, the same issues that cause your 16” MacBook Pro to overheat when connected to a display.

With Apple’s ARM-based Macs, all of that is going away. But of course, this is a major transition and it won’t be perfect right out of the gate, since it could potentially take developers a year or two before they fully optimize their apps for Apple Silicon.

So hopefully you learned something new from this article.

Comments (0)

    No comments yet

You must be logged in to comment.