Apple M1 Ultra: Apple’s Latest Flagship SoC for the Desktop

Back in 2020, when Apple announced its first desktop silicon, the M1, it drew some criticism initially for not sharing hard figures to substantiate its performance claims. These claims, however, mainly were borne out as the reviews of the M1 poured in. And subsequently, the following year witnessed the release of two more SoCs, the M1 Pro and M1 Max, built on the same architecture.

While both the M1 Pro and the M1 Max shared the same architecture, the M1 Max had some distinct advantages. It packed double the amount of GPU cores as the M1 Pro, which allowed it to deliver nearly twice the performance of the M1 Pro in terms of GPU and media performance.

Just when these performance figures of the M1 Max gave the impression that Apple had reached the peak of optimization of the M1 architecture, it released the fourth and final SoC in the M1 family of chips at its spring Peek Performance event, the M1 Ultra.

What does the release of the M1 Ultra mean for the future of personal computing on the Mac? Let’s dive in to find out.

What’s At the Core?

Unlike previous iterations of M1 that grew in size and introduced more transistors and cores to push the boundaries of performance, the M1 Ultra comes across as a subtle upgrade on the surface as it does neither of those things. Until you peak inside, that is.

At its core, the M1 Ultra has the year-old M1 Max—two M1 Max, in fact. Essentially, what Apple has done with its latest M1 SoC this time around is combine the die of two M1 Max chips into one large SoC using what’s known as UltraFusion, Apple’s custom packaging architecture.

Apple claims the UltraFusion technology to be the first of its kind. But truth be told, it’s really not much different than the chiplet-based design approach currently adopted by AMD and Intel for their current CPU lineups. Even so, this still doesn’t negate UltraFusion’s apparent advantages in the development of the M1 Ultra.

According to Apple, the UltraFusion technology uses a silicon interposer to connect the two M1 Max chips together across more than 10,000 signal points. This results in an ultra-high, 2.5TB/s of bandwidth (with significantly low latency) that’s more than four times the bandwidth of conventional multi-chip interconnect technology, which allows the M1 Ultra to be recognized and used as a single chip by the operating system and applications.

Additionally, the benefits of this single design approach also extend to software developers: they now won’t have to rewrite code for the M1 Ultra and get to take advantage of its performance readily.

Apple M1 Ultra: Specifications

With a unified design approach on the M1 Ultra, Apple has virtually put twice the hardware components of the M1 Max inside the M1 Ultra. This ideally means a significant performance gain over previous M1 chips, given the number of transistors (114 billion) on board, combined with the powerful 20-core CPU and a whopping 64-core GPU.

Apple M1 Ultra: CPU

Talking about the CPU, the 20-core CPU on the M1 Ultra splits into 16 high-performance (Firestorm) cores and four high-efficiency (Icestorm) cores. Essentially, this is twice as many cores as the M1 Max that has a 10-core CPU with eight high-performance cores and two high-efficiency cores, which already gives it a significant boost in performance compared to the M1.

Apple says this increase in core count allows the M1 Ultra to deliver 90 percent higher multi-threaded performance than the fastest 16-core PC desktop out there in the same power envelope. Not just that, it’s said to do this using 100 fewer watts, meaning there’s less energy consumption even in some of the most demanding tasks, which has it covered from the efficiency standpoint as well.

Apple M1 Ultra: GPU

Graphics performance also gets a boost on the M1 Ultra, thanks to its 64-core GPU. To put this GPU into perspective, this is eight times more than the M1 and four times that of the M1 Pro chip. With graphics, too, Apple has drawn a comparison directly with the highest-end desktop GPU, wherein it claims that the M1 Ultra delivers faster graphics performance while using 200 fewer watts of power.

Apple M1 Ultra vs Competition

While Apple didn’t explicitly name the high-end CPUs and GPUs, it’s drawing these conclusions during the presentation; its press release suggests that the CPU in question here is the Intel Core i9-12900K, while the GPU is the Nvidia GeForce RTX 3090. Of course, how much these “jump in performances” and “gains over competitors” translate to in the real world will only come to light once an M1 Ultra-powered Mac is put through its paces. However, in the meantime, we have Geekbench scores for the M1 Ultra, which just happened to have leaked out a few hours after Apple’s announcement, which should give us some idea about M1 Ultra’s performance.

According to these results, the M1 Ultra gets a single-core score of 1793 and a multi-core score of 24055. This looks like a significant jump over the 1152 CPU and 19951 GPU scores of the 28-core Intel Xeon W chip that’s currently powering Apple’s high-end Mac Pro.

Much like CPU and GPU, the unified memory architecture on the M1 Ultra also gets a performance boost. It now offers double the bandwidth of the M1 Max (400GB/s) at 800GB/s, which Apple claims to be ten times more than what’s available with current PC desktop technology.

Besides an increase in bandwidth, the M1 Ultra also supports higher unified memory and can be configured up to 128GB. This is twice the capacity of the M1 Max and four times that of the M1 Pro, and to some extent, this should ease out GPU-intensive workloads, such as those involving heavy rendering or working with 3D geometry significantly.

Apple M1 Ultra: NPU

Lastly, when it comes to neural processing, the M1 Ultra comes equipped with a 32-core Neural Engine that can run up to 22 TOPS and speed through various ML operations. It can even improve video encoding and decoding throughput with ProRes, which can prove handy for those who shoot videos in ProRes on their iPhones.

Bringing Hardware and Software Together

One of the things Apple did really well with the launch of its first silicon—the M1—was to integrate it tightly with its operating system, which resulted in an increase in performance and efficiency—compared to the Intel-powered Macs—on its then-current OS, macOS Big Sur.

Its latest macOS release, macOS Monterey, builds upon the same idea. In fact, it’s already keeping up well with the M1-powered Macs and is said to be capable of leveraging the full potential of even the all-new M1 Ultra.

Of course, like before, being a part of the M1 silicon family, the M1 Ultra also offers the ability to run iPhone and iPad apps on the Mac. In addition, apps that haven’t yet been updated to Universal can also still run seamlessly on the M1 Ultra with Apple’s Rosetta 2 technology.

How Do I Get to Experience the All-New M1 Ultra Chip?

When Apple ditched Intel and launched its first silicon in 2020, it seemed focused on improving the performance of its MacBooks at the time, with the long-term aim to move its entire Mac lineup to Apple silicon.

While the latter may still take some time, Apple appears to have succeeded in achieving the former goal. So much so that incremental upgrades to the M1 in the form of the M1 Pro and M1 Max reached a point where it peaked the die size of the M1 and forced Apple to combine two chips together to power its next generation of desktop computers.

One such workstation is the Mac Studio, which was announced by Apple during the Peek Performance event. Mac Studio comes with both M1 Max and M1 Ultra, with the one powered by the M1 Ultra boasting up to 3.8 times faster CPU and up to 4.5 times graphics performance than the current fastest 27-inch iMac. It starts at $1999 in the US and Rs 189900 in India and will go on sale on March 18.