Apple unveils M5 Pro and M5 Max with new Fusion Architecture and AI focus

Apple's latest Mac chips mark a structural shift in its silicon roadmap, introducing a new dual-die "Fusion Architecture" for Pro models and sharpening its focus on on-device AI, even as higher prices and unanswered performance questions temper early reactions.

New Fusion Architecture reshapes Pro and Max design

Apple on March 3 introduced the M5 Pro and M5 Max alongside updated MacBook Pro systems, describing them as its most advanced chips for professional laptops. At the center of the announcement is what Apple calls a new "Fusion Architecture," which combines two silicon dies into a single system-on-a-chip.

Unlike earlier Pro and Max generations, which largely scaled up a single die with more cores, the M5 Pro and M5 Max divide responsibilities across chiplets. According to Apple and analysis from Ars Technica, one die houses the 18-core CPU, 16-core Neural Engine, and key I/O controllers, while a second die focuses mainly on graphics, media engines, and memory bandwidth.

Apple says the approach allows it to scale performance while preserving its unified memory model and power efficiency. The M5 Pro supports up to 307GB/s of memory bandwidth, while the M5 Max doubles that to up to 614GB/s. The Max variant also doubles GPU resources compared with the Pro, offering up to 40 GPU cores and dual media engines.

A third type of CPU core introduced

Another notable change is the introduction of a third CPU core category. In addition to "super" cores — Apple's new branding for what were previously known as performance cores — and efficiency cores, the M5 Pro and M5 Max add a new middle-tier performance core optimized for multithreaded workloads.

Apple says the new 18-core CPU combines six super cores designed for peak single-threaded speed with 12 performance cores aimed at improving throughput in heavily parallel tasks. The company claims up to 30% faster pro workload performance compared with the M4 Pro, and up to 2.5 times higher multithreaded performance versus M1 Pro and M1 Max.

Ars Technica notes that while earlier M4 Pro and M4 Max chips featured more traditional "big" cores on paper, the M5 generation relies on architectural gains and a rebalanced core mix to deliver overall improvements. Independent benchmarks were not available at launch.

AI positioned as a central design goal

Across Apple's announcement, AI performance is framed as a primary objective. The M5 Pro and M5 Max integrate Neural Accelerators within each GPU core and pair them with a faster 16-core Neural Engine. Apple says the new GPUs deliver more than four times the peak AI compute of the prior generation, targeting demanding workflows such as large language model development, code compilation, and 3D rendering.

The chips also include updated media engines supporting hardware acceleration for H.264, HEVC, AV1 decode, and ProRes, as well as Thunderbolt 5 controllers integrated directly on chip.

Industry coverage has emphasized that the AI positioning extends beyond marketing language, reflecting a broader shift across the PC industry toward on-device inference and developer-focused AI tooling.

Pricing and open questions

While Apple characterizes the M5 Pro and M5 Max as a "monumental leap" in Apple silicon, industry commentary has also focused on rising Mac prices accompanying the refresh. Several outlets noted higher starting prices across updated MacBook configurations, framing the M5 rollout as both a technological step forward and a costlier upgrade cycle.

Ars Technica also pointed to open questions around how the new architecture will affect potential future Ultra chips, which in past generations were created by fusing two Max dies. With the Pro and Max chips already adopting a chiplet approach, it remains unclear whether Apple will continue that formula or design distinct dies for a future M5 Ultra.

For now, the M5 Pro and M5 Max represent Apple's most significant internal redesign of its Pro-class silicon since the launch of Apple silicon in 2020 — a move that underscores a dual strategy of architectural experimentation and intensified competition around AI performance in high-end laptops.

Source: Compiled by DIGITIMES Asia, March 2026

Article edited by Joseph Chen