Apple has expanded its desktop lineup with the introduction of the M6 system on a chip in the Mac mini and the M5 Ultra in the Mac Studio, delivering architecture-level updates across compute, graphics, and on-device artificial intelligence.
M6: 2nm Process and Re-Architected Compute
The M6 marks Apple’s transition to a 2-nanometer manufacturing process, increasing transistor density to improve processing throughput and power efficiency. The chip incorporates a newly configured 12-core CPU complex comprising 2 super cores, 4 performance cores, and 6 efficiency cores. A
ccording to technical specifications, this setup delivers up to 1.2 times faster multithreaded performance compared to the M5 and up to 2.4 times faster performance than the original M1.
On the graphics and AI side, the M6 features:
- Dual 16-core Neural Engine: Delivers up to 2x peak compute over previous generations, allowing system frameworks to run parallel model execution across both engines.
- 12-core GPU with Neural Accelerators: Embeds dedicated Neural Accelerators into each core, increasing peak GPU compute for AI by roughly 30 percent over the M5 and more than 8x over the M1.
- Updated Graphics Architecture: Incorporates updated shader core architecture, Dynamic Caching, hardware-accelerated ray tracing, and a 50 percent increase in geometry rates.
- Unified Memory: Supports up to 32GB of unified memory with bandwidth reaching 170GB/s—a 10 percent gain over the M5.
M5 Ultra: Quad-Die Architecture for Professional Workloads
Targeted at high-demand professional, scientific, and enterprise AI workflows, the M5 Ultra utilizes next-generation UltraFusion packaging technology. By interconnecting two dual-die M5 Max chips, the processor functions as a single unified quad-die SoC with inter-die bandwidth exceeding 4.4TB/s.
Key hardware specifications for the M5 Ultra include:
- CPU: Up to a 36-core configuration (12 super cores and 24 performance cores), yielding up to 1.25x faster single-threaded and 1.3x faster multithreaded performance over the M3 Ultra.
- GPU & Acceleration: Up to 80 GPU cores with integrated Neural Accelerators, second-generation Dynamic Caching, mesh shading, and third-generation ray tracing, providing up to 40 percent faster graphics rendering compared to the M3 Ultra.
- Dedicated Media Engines: Hardware-accelerated support for H.264, HEVC, AV1 decode, and four ProRes encode/decode streams.
- Memory and Bandwidth: Capacity of up to 512GB of unified memory with 1.2TB/s of bandwidth, enabling local execution of large language models containing hundreds of billions of parameters.
Framework and Developer Integration
Both chips integrate with Apple’s software stack, including Core AI, Core ML, Metal, and Xcode. The toolsets automatically distribute compute tasks across the CPU, GPU Neural Accelerators, and Neural Engines, allowing developers to execute, fine-tune, and deploy large models locally using Apple Foundation Models, App Intents, or custom architectures.