Apple’s decision to shift from Intel’s x86 silicon to its own Arm-based designs has proven successful, with no signs of regret. This bold move was announced in June 2020, and now, at WWDC 2023, Apple is proudly introducing the M2 Ultra, their most expansive and powerful chip to date. In fact, the M2 Ultra is approximately double the size of the M2 Max.
Apple is able to achieve this by utilizing its UltraFusion technology, which connects the dies of two M2 Max chips, effectively doubling the overall performance. This approach mirrors what Apple did in the previous generation with the M1 Ultra, where a pair of M1 Max chips were interconnected within the same package using the innovative UltraFusion die-to-die interconnect technology.
The M2 Ultra by Apple is built using an advanced second-generation 5-nanometer process. With an impressive 134 billion transistors, it surpasses the M1 Ultra by an additional 20 million. The M2 Ultra showcases a unified memory architecture, boasting an astonishing 192GB of memory capacity, representing a remarkable 50 percent increase compared to the M1 Ultra. Additionally, it delivers exceptional memory bandwidth of 800GB/s, which is double the amount found in a single M1 Max die.
“M2 Ultra delivers astonishing performance and capabilities for our pro users’ most demanding workflows, while maintaining Apple silicon’s industry-leading power efficiency,” said Johny Srouji, Apple’s senior vice president of Hardware Technologies. “With huge performance gains in the CPU, GPU, and Neural Engine, combined with massive memory bandwidth in a single SoC, M2 Ultra is the world’s most powerful chip ever created for a personal computer.”
Apple states that the CPU of the M2 Ultra outperforms the M1 Ultra by 20 percent. Furthermore, the larger GPU of the M2 Ultra demonstrates up to a 30 percent increase in speed, while its Neural Engine is claimed to be up to 40 percent faster. Additionally, Apple asserts that the media engine of the M2 Ultra is twice as capable as that of the M2 Max, particularly when it comes to ProRes acceleration.
Concerning the potential latency resulting from combining two dies, Apple has addressed this with the UltraFusion interconnect, featuring a custom-built packaging technology that utilizes an interposer attaching the dies through more than 10,000 signals. According to Apple’s data, this arrangement provides over 2.5TB/s of “low-latency” interprocessor bandwidth. Notably, software recognizes the package as a single chip rather than two chips merged together.
In terms of the design, the M2 Ultra features a 24-core CPU, comprising 16 high-performance cores and 8 high-efficiency cores. It is coupled with either a 76-core GPU (with a 60-core option available). Additionally, the Neural Engine boasts 32 cores and achieves an impressive 31.6 trillion operations per second, making it 40 percent faster than the M1 Ultra.
Apple has made the M2 Ultra available for order today in their refreshed Mac Studio and Mac Pro systems, with shipping scheduled for next week.
Via Apple