There are rumors that the upcoming higher-end M5 Pro and M5 Max processors will be architected differently than that M5presumably to improve scalability. They are expected to debut in next wave of MacBook Pros.

This change in architecture means that it is easier and more efficient to speed up the chips, as well as reducing heat generation and possibly straining battery life. To achieve this feat, word is Apple switched from TSMC’s system-on-chip layout used in the M5 to a custom variation of TSMC’s system-on-integrated-chip, molding-horizontal layout, called SoIC-MH.

What has changed?

With the SoC architecture used by Apple for its M5everything except the memory is made wholesale on a die, which is connected to the system memory for a chip. The company is essentially scaling the Pro, Max and Ultra chips by integrating multiple dies.

So, for example, the GPU core counts of the Max and Ultra chips are usually twice that of the two lower versions. But that is often an inefficient and inefficient way to increase performance.

TSMC’s SoIC Manufacturing produces separate dies (or chiplets) for selected operation groups, which are then connected via small high-speed connections into a package and combined with memory for the final chip.

In the case of Apple, the expectation is that it will move the GPU to its own chiplet to make it easier to scale its performance independently of the CPU. This is important because the demand for tensor (for AI) and graphics processing power is growing rapidly, while the demand for high-powered CPUs is not so much at the moment.

SoICs generally have stacked chiplets, which allow for the smallest and, therefore, fastest possible interconnects. But Apple may be using a custom SoIC-MH layout, where they are placed next to the main die rather than stacked.

A potentially critical step

Without any solid information, it’s hard to predict how much this will affect performance.

For example, Apple may put fewer GPU cores in some system configurations rather than more, although I really doubt it wants to reduce performance. Additionally, the layout can change Apple’s CPU core count options, depending on what Apple chooses to do with the die opening space, or if the company chooses to shrink the die.

But when Apple announced the M1 Ultra for the Mac Pro, my first thought when I saw that it only uses integrated graphics is that it is not a good long-term strategy in chip architecture.

With each generation, I’m waiting for the shoe to drop: The M series still doesn’t support discrete graphics cards, only the integrated GPU on the chips, which means the company needs a better way to increase the power of the on-chip GPU over time. Apple has addressed the M5’s AI tensor processing needs by adding a neural accelerator to each GPU core, but the one-accelerator-per-core approach is also a slow road to take.

I don’t know if a GPU chiplet is the best solution to these problems, but hopefully it will allow Apple to cram as many GPU cores as possible, without compromising on fitting the fixed available die space. Instead, Apple could offer, say, a 14-inch MacBook Pro that could compete when it comes to heavy machine learning workloads.

If Apple really implements this kind of architecture, then we will definitely see the M5 Pro and Max systems at the Worldwide Developers Conference, the company’s annual conference of developers, because Apple needs to wake up developers about gaming (again) and AI (again).

And given the laptop price pressure from the severe lack of memory (among other obstacles), the ability to offer chip options with GPU and CPU performance broken in this way can help control costs – or at least make it easier for you to manage to make a choice within your budget from granular configuration options.