Apple has abandoned Intel, and the progress in producing its own processors for Mac products has been smooth. The M2 chip, based on the Arm architecture, combines powerful CPU and GPU cores, providing a perfect blend of mid-range to high-end performance for the entire Mac product lineup. This is particularly beneficial for products like MacBook and Mac Mini, which lack the space for powerful graphics cards but can accommodate high-end M1 and M2 chips with powerful integrated graphics, such as the M2 Max. The M2 Pro, M2 Max, and M2 Ultra are truly impressive.
Why don’t Intel and AMD do the same thing and manufacture large chips with a large number of CPU and GPU cores? After all, they both have excellent CPU and GPU technologies themselves, and they have more control over the broader processor market.
01
Intel and AMD can manufacture chips similar to the M2
If Intel and AMD were willing, they could manufacture chips similar to the M2 Max and M2 Ultra. Although the M2 Pro, Max, and Ultra differ significantly from the products we use in PCs, all Apple did was create a large chip containing CPU and GPU cores. The only substantial difference between Intel and AMD’s long-standing production of CPUs with integrated graphics, such as the Core i9-13900K or Ryzen 9 7950X, and the Apple M2 series is that the larger M2 chips come with larger GPUs.
Intel and AMD have produced chips similar to the larger M2 processors, even before the launch of the M1. Intel’s Kaby Lake G series combined a four-core Kaby Lake CPU with a mid-range Vega GPU. While Kaby Lake G didn’t succeed commercially, it ultimately provided a mainstream product at the time, with a potent CPU and GPU combination.
One of AMD’s main product lines is, of course, the APU, which is essentially a marketing term used to describe AMD’s gaming CPUs with integrated graphics. AMD’s most successful APU series since the launch of the Xbox One and PS4 a decade ago has been the APUs for gaming consoles. The latest gaming console APUs, found in the Series X and PS5, combine 8-core Zen 2 CPUs with mid-range RDNA2 GPUs, putting them at a similar level to at least the M1 Pro and M2 Pro, and possibly even some high-end chips.
02
The M2 has drawbacks that Intel and AMD may not favor
While the M2 series is indeed impressive for Apple, it’s not a flawless processor. It has some drawbacks compared to Intel and AMD chips, primarily related to memory, core count, and manufacturing, which can significantly impact both performance and pricing.
One peculiar aspect of Apple M processors is their memory configuration. Typically, CPUs and lightweight integrated GPUs don’t require such high bandwidth, so Intel and AMD tend to pair their mainstream chips with smaller 128-bit memory buses. However, GPUs demand a significant amount of memory bandwidth, which is why dedicated GPUs are paired with GDDR VRAM and have more extensive memory buses. A 128-bit bus is usually considered the minimum and reserved for very small GPUs.
However, Apple equips its M chips with very powerful GPUs, which require more memory bandwidth than usual. While the M2 has a 128-bit wide bus, the M2 Pro and M2 Max feature 256-bit and 512-bit wide buses, respectively. And since the M2 Ultra combines two M2 Max chips, it boasts a massive 1024-bit bus bandwidth. These buses occupy a substantial amount of space, accounting for approximately 13% of the M2 Max and Ultra’s size, solely dedicated to the memory buses.
All the space allocated for the memory bus creates a domino effect. The limited space for accommodating more CPU and GPU cores results in the M2 Pro and M2 Max underperforming in terms of performance per square millimeter. For example, the M2 Ultra in the Mac Studio lags far behind the Core i9-13900K and Ryzen 9 7950X in benchmark tests like Cinebench, especially in multi-threaded performance. These CPUs are much smaller in size because they aren’t hampered by the massive memory subsystem required by large integrated GPUs.
Attempting to combine large CPUs, large GPUs, and large memory systems into a single chip has manufacturing implications. The estimated chip size is 550 square millimeters, making the M2 Max very large, and the M2 Ultra is the largest consumer chip ever, exceeding 1000 square millimeters in size. The cost of producing these on TSMC’s 5nm node is undoubtedly astronomical.
03
Providing highly customized chips is not the typical approach of Intel or AMD
Apart from the hardware issues introduced by the larger versions of the M2 chip, there is also a fundamental difference in business models. Apple stands in stark contrast to Intel and AMD: it produces processors for its special products. Meanwhile, Intel and AMD are producing chips that support almost all PCs worldwide, excluding Mac. Pursuing such a vast market means that customization is a disadvantage, offering each company a radically different set of incentives when designing hardware.
If Intel and AMD were to attempt to manufacture large processors similar to the M2, a major issue would be motherboards. Traditional mainstream x86 CPUs are often small, but to accommodate chips like the M2 Max/Ultra, Intel and AMD would need to release new motherboards with numerous slots and extensive VRM levels. This might also require eight memory slots, all of which would need to be fully populated with fast modules to achieve good GPU performance. This is very costly and cumbersome.
Apple can sidestep all of this because the M2 is designed specifically for what Apple wants to produce in terms of computers and what its customers want to buy. Intel and AMD can’t do that because we like having a variety of CPUs that we can put into personalized yet ultimately similar PCs, where RAM, storage, and GPUs are plug-and-play. Imagine if you had to upgrade your 1000 square millimeter processor just because you wanted faster graphics processing; if you thought the RTX 4090 was expensive, you might go bankrupt with a single upgrade.
However, it goes both ways, and the M2 has some clear advantages. For example, while Apple’s CPUs and GPUs are relatively weak compared to high-end products manufactured by Intel, AMD, and Nvidia, the M2 product stack has high-end encoding performance across the board. Apple can justify adding these encoders to its chips because many people use Apple devices for video editing. However since Intel and AMD CPUs are general-purpose chips, this high-end encoder doesn’t make sense, especially when high-end GPUs can compensate for this deficiency.
In processor design, the hardware and different business models are intertwined for these three companies. If it were just about hardware, technology, and specifications, we might see Intel and AMD also attempt to manufacture processors similar to the M2 series. But Apple is a computer design company that manufactures its processors, while Intel and AMD are processor design companies that sell processors to other companies.
04
In certain situations, chips similar to the M2 from Intel and AMD could potentially thrive.
While the traditional PC realm may not be the most conducive space for Intel and AMD to introduce large processors with a multitude of CPU and GPU cores like the M2, I believe there are other areas where such chips could find success. One of the recent innovations from Intel and AMD is small chips, which would make building processors like the M2 easier. Intel and AMD could simply use their generic small chips and combine them in such a way as to create this large dedicated processor (manufacturing it with small chips would also likely be more cost-effective).
Gaming consoles are an obvious place where such chips could be relevant, but NUCs, OEM desktops, or even laptops might be better suited for M2-like processors. After all, these devices are already customized from the outset, so the benefit of using hardware designed to be compatible with many other components isn’t as significant. I don’t necessarily think it needs to become a reality, but I could be wrong, and I would love to see a gaming version of the Mac Mini or Mac Studio.
Related:
Disclaimer: This article is created by the original author. The content of the article represents their personal opinions. Our reposting is for sharing and discussion purposes only and does not imply our endorsement or agreement. If you have any objections, please contact us through the provided channels.