MacBook Pros : Apple’s Back to the Future

The first impression of the new MacBook Pro is that you have stepped back in time. The computer is remarkably powerful, insanely powerful for a laptop, but when you first pick it up, it seems like you have jumped back in time.

The computer has ports – loads of sensible ports, which would not be important to mention except it has been years since we could plug an SD card directly into a card reader on the MacBook Pro, or headphones and there is even a Magsafe power plug. After years of dongles, you can plug an HDMI monitor in directly! Add to this the computer is a beast. It is big, it is heavy and solid. If you never opened it, you could be forgiven for thinking it was 2015 (or even 2006). But the market for this computer is not the fashion-forward sleek designer crowd. Apple built these new laptops with function firmly in front of form, something decidedly un-Apple-ish.

Thankfully the innovation, build quality and Apple moments of delight are all present when you open the lid and switch it on. Without a doubt the machine is impressive, the screen is remarkably brighter. It is a 1000 nits / 1600 nits peak HDR screen – twice that of an iMac.  It has one HDMI port for displays and three Thunderbolt 4 ports in total. The sound system, web camera (now HD 1920×1080) and the new keyboard are all delights to use. But the real magic is the chipset and memory design. Without the M1 performance, the machine would lack substantial interest for professional users. With the new M1 Pro and M1 Max, combined with the GPU and ML engines, the MacBook Pro becomes very interesting.

Slots and SD card readers!

In simple terms, the M1 Max with its integrated graphics solution is similar in grunt to an NVIDIA RTX 3080 mobile but with lower power consumption, more memory, less noise and higher throughput. It is not the same as an RTX 3080 since it has no dedicated ray-tracing hardware and of course, no CUDA. But the new MacBook Pro runs fast, and thanks to some very clever power management, it runs fast on batteries for a much longer time than almost all PC laptops. The Metal graphics can drive up to four displays from MacBook Pro—and charge it at the same time.

Professional video editors using Final Cut Pro on the new MacBook Pro can perform tasks never before possible on a Mac, like the ability to play seven streams of 8K ProRes or colour-grade 8K HDR video with the new Liquid Retina XDR display — a first for any notebook. Final Cut Pro also introduces a new Object Tracker that uses machine learning to automatically detect faces and objects, and match their movement to create beautiful, cinema-quality titles and effects on any Mac. With the dedicated ProRes support, an editor can export ProRes video over 5x faster than before. Logic Pro, Apple’s pro music creation software, now comes with a complete set of tools for authoring spatial audio music — allowing anyone to mix and export their songs in Dolby Atmos.

These machines are VFX workhorses but not across all applications. Adobe, Blackmagic Design(DaVinci Resolve), Maxon (C4D) and several other companies have full M1 native applications but other applications sadly rely on CUDA and assume NVIDIA cards. For example, Foundry Nuke is not supported (and Katana never ran on MacOS), Autodesk Flame family of products is not supported on the M1 Chip. Autodesk Maya 2022 works but under Rosetta 2 mode.

Interestingly, Cinema 4D may leverage off their tight relationship with Apple to gain market share from Maya, as Maxon seems to be even more focused on innovating C4D to high-end users and freelancers. C4D coupled with the M1 performance is able to handle even more complex and data-dense projects.

Unified memory is a critical aspect

The new M1 Chips

Apple has two new processors for these MacBook Pros, the M1 Pro and the M1 Max. These Apple-designed ARM-based chips are much larger versions of the M1 launched last year, and importantly have their CPU and GPU memory combined and shared.

The new M1 chips are a ‘system-on-a-chip’ (SoCs) design which incorporates CPU, GPU, and NPU, leveraging a unified memory space to maximize efficiency and performance. M1 Pro packs in 33.7 billion transistors, more than 2x the amount in M1. The M1 Max is the larger of the two SoCs, with 57 billion transistors, 10-core CPU with eight performance cores, and two efficiency cores, a 32-core GPU, and up to 64GB of memory. Each of those GPU cores comes with 16 Execution Units, each of those containing 8 ALUs. That makes for a 4,096 ALU GPU within the M1 Max.

At the launch event Apple indicated that this was similar in performance to the one of the highest-end GPU in the PC laptop market but using up to 100 watts less power. The M1 Max is four times larger than the first M1 chip and also four times faster. This shows at the artists level, compared to a 13″ M1 MacBook, the 16″ M1 Max does feel at least twice (or more) as fast, but then it is significantly bigger and more expensive.

The M1 Max comes with 400GB/s of memory bandwidth and access to up to 64GB of RAM. So while the Terraflops performance of the M1 Max might be closer to the NVIDIA RTX 2080, (M1 Max ~10.44TFlops vs. RTX2080 10TFlops), the system benefits from being able to access a lot of memory very very quickly. M1 Max offers a higher-bandwidth on-chip fabric and doubles the memory interface compared with M1 Pro at 200GB/s or nearly 6x the memory bandwidth of the original M1. This allows M1 Max to be configured with up to 64GB of fast unified memory. With its unparalleled performance, M1 Max is the most powerful chip ever built for a pro Laptop.

Octane Renderer from Otoy on M1 Max

In the launch demos, Jules Urbah from Otoy showed an Octane 3D sequence from an upcoming CBS project. This Enterprise model was too large to load in almost any normal computer, being over 34.8 GB. But in the 64GB configured M1 Max it loads and still has room to spare. It is worth remembering that most PC laptops top out at 16GB of graphics memory.  With all of this graphics processing power and RAM, the M1 Pro and M1 Max make 3D modeling and rendering quick, and they handle deeply layered HDR video timelines, and complex filters and audio easily.

Naturally, the whole 64GB is only notionally available to the graphics engine, as the CPU still needs to use some, but this is still a huge leap in terms of what can be stored in memory. For developers and TDs, the new computers compile about four times after. Apple’s M1 range goes against the current trend of chiplets: separating a large chip into several smaller, distinct chiplets that are tightly interconnected. They do this to speed up data transfer and have a more unified memory approach.

Power Amps vs Watts/hrs

Unlike many PCs that only provide this power when plugged in, the new MacBooks can sustain this even when on battery. This is due to a combination of low power design, new cooling and very smart memory management. The 14-inch model delivers up to 17 hours of video playback, which is seven additional hours, while the 16-inch model gets up to a remarkable 21 hours of video playback, which is 10 additional hours — the longest battery life ever on a Mac laptop. It can do this because it can lower the CPU/GPU processor power requirements dramatically while playing video.

This is indicative of how power is managed. While all the components together would pull quite a lot of power, it is rare that every aspect of the machine needs maximum power at the same time. In video editing, one may need fast video processing of multiple streams of video, but much of the UI real estate of FCP, DaVinci or Premiere Pro does not draw power when one is scrolling a timeline. So while the edit window showing the current edit is updating and pulling power, the rest of the UI is being updated at a much lower frame rate, which a user will never notice, since it is static and often dark.

Apple learned many of these tricks from years of mobile computing. The iPhone requires variable frame rates and clever power usage, and the MacBook is inheriting all these years of trying to eke out lower power requirements. Simple electrical engineering points to a maximum amount of Amps that a battery in a laptop can provide. If you exceed this, then you have to plug in a power adapter, but assuming you can design a low enough set of performance computing engines, then the trick is to manage who gets what power when. While many people focus on battery size, just having more power storage or battery size is only half the battle. It is the size of the drain on this power reserve that normally causes a PC laptop to be plugged in, not that the laptop’s battery is low on reserves.

Why is Apple’s silicon so advanced?

There is no doubt that Apple benefits from the software and the hardware being developed together and that Apple just has extremely good chip designers and engineers. But there is much more to it than just the depth and strength of the Apple engineering team. Apple also has a much better supply chain management system. While NVIDIA is also leaping forward with massive GPU gains, they have suffered from supply chain problems and thus the availability of new graphics cards. This is something seemingly not affecting Apple which is expecting to ship the new MacBooks in quantity, almost immediately.

But while it might appear that NVIDIA is a good company to compare to, it is actually Intel that is the real story. Apple has moved from purchasing Intel chips and into a very tight and productive arrangement with the global manufacturing giant TSMC in Taiwan.  As the Economist magazine noted “It is hard to overstate the importance of the Taiwanese semiconductor maker to the world’s economy. TSMC dominates production of the world’s most sophisticated semiconductors and counts Apple and Qualcomm among its biggest customers.”  TSMC, founded in 1987, is the world’s most valuable semiconductor company and one of Taiwan’s largest companies. It is also the first company to commercialize extreme ultraviolet (EUV) lithography technology in high volume. Apple and NVIDIA both use TSMC chips, but Apple seems to have a tighter supply chain link, and even Intel is now expected to be a customer of TSMC’s chip production in 2022.

After years of struggling to keep up with competitors, Intel CEO Pat Gelsinger announced in March this year that Intel will be commissioning other companies to make chips for them. Prior to this, Intel had focused on making its own chips, but that changed with the establishment of their independent Intel Foundry Services business unit.

TSMC is significant for its current impressive 5nm technology and where it is heading. Process technology has been and will continue to be one of the most important aspects of computer chip performance and efficiency. TSMC is one of only a small number of cutting-edge chip manufacturers, along with Samsung, UMC, and Intel. But unlike Intel, TSMC is leading in 3nm lithography fabrication process.  Apple is a long-standing customer of TSMC, and used it, along with others, to produce chips for its iPhones, iPads and now MacBook Pros. The current M1 Pro and M1 Max are using the current 5nm fabrication, but TSMC’s new 3 nm process would provide even more performance and efficiency gains beyond the new M1 Apple chips. Just moving from 5nm to 3nm alone should mean performance gains of up to 15% and efficiency gains of up to 30% it has been estimated. Apple has not discussed a possible ‘M2’ 3nm chip but various industry reports indicate this is very much something on the cards. Meanwhile, TSMC is also discussing 2nm technology rollouts. This is important to VFX companies who want to know that any technology path has a long road map ahead.