The first Apple event of this 2022 has left us with great news in a wide range of products, not just in terms of phones iPhone SE 2022 Edition long-awaited, but also in the ipad air and on new desktops mac studio.
However, perhaps the big surprise was the introduction of a new Apple Silicon processor that positions itself as the most powerful in the family, the M1 Ultra. It’s above its M1 Pro and M1 Max predecessors, with significantly more CPU and GPU cores.
However, the higher performance of this new chip seems to have a trick, since what Apple did was join two previous generation chips together to double the numbers and increase performance.
How does this translate into power dissipation and energy efficiency effects? Let’s see some technical data provided by the company in order to analyze what this architecture that Apple calls UltraFusion consists of.
What are the differences between the M1 Ultra, M1 Pro and M1 Max?
As Apple explained at its event, thanks to the new UltraFusion architecture, the company manages to integrate two M1 Max chips into an SoC package in a way that reaches new levels of performance and capabilities. It was called the M1 Ultra.
In this way, the new M1 Ultra chip debuts in a new team launched by Apple, which is none other than the Apple Mac Studio, a team very oriented towards creative professionals who demand power, keeping intact performance per watt.
Inside, it packs a total of 114 billion transistors and can be configured with up to 128 GB of high-bandwidth unified memory, providing a total of 20 CPU cores and up to 64 GPU cores.
To get an idea of the increase in performance, which can sometimes reach 60% in certain tasks compared to the predecessor, we must take into account the data of the M1 Max chip, which has 10 CPU cores and 32 GPU cores.
What is chip fusion?
We’ve already mentioned that to create the M1 Ultra, Apple combines two M1 Max chips with UltraFusion, the custom-designed packaging architecture. We’ve seen this before in other microarchitectures, so it’s nothing new.
What’s true is that the most common way to increase CPU performance is usually by connecting two chips through a motherboard, using a dual socket, or even scaling a number. It’s common to see this in server setups and workplace
This often leads to latency issues, as well as reduced bandwidth and increased power consumption. According to Apple, these factors are eliminated by integrating the two chips into a single SoC chip package.
With UltraFusion, Apple uses a middleman that connects chips to more than 10,000 signals, providing approximately 2.5 TB/s of bandwidth, more than four times the multi-chip connection technology used by other architectures.
How does the operating system behave?
If the M1 Ultra stands out for anything, it’s because it appears to the operating system as if it were a single chip, so the software recognizes it as such, as long as you have the latest version of macOS Monterey on your computer.
Developers don’t have to rewrite code to benefit from performance. Development technologies like Metal let you get the most out of the chip, and Core ML optimizations use the new 32-core neural engine to run machine learning models.
How will this chip take advantage of virtualization environments? What performance will it offer running Windows 11 in virtual mode? We look forward to answering all of these questions, which we will do once we can test the new gear. Mac Studio.
