Apple Debuts M6 and M5 Ultra, Its First 2nm Chip and First Quad Die Ultra

Apple today introduced two new Apple silicon chips, the M6 and the M5 Ultra, marking the company's move to a 2 nanometer manufacturing process and its first Ultra chip built from four dies.

The M6 debuts inside the new Mac mini, while the M5 Ultra powers the new Mac Studio. Both chips were announced alongside the hardware they ship in, and both center on faster on device AI.

M6 Moves to 2nm

M6 is Apple's first chip built on a 2 nanometer process, which packs more transistors into a smaller die for gains in both speed and efficiency. Apple describes it as the biggest per generation leap across the M series to date.

The chip has a 12 core CPU, up from 10 cores on the M5. That complex is split into 2 super cores, 4 performance cores, and 6 efficiency cores. Apple says the super cores deliver the world's fastest single threaded performance, with multithreaded performance up to 1.2x faster than M5 and up to 2.4x faster than M1.

The GPU also grows to 12 cores, again two more than M5, with a Neural Accelerator built into each core. Apple claims a nearly 30 percent jump in peak GPU compute for AI compared to M5, and more than 8x compared to M1.

A Dual Neural Engine

The headline addition on M6 is a Dual 16 core Neural Engine. Apple says the pairing delivers up to 2x the peak compute of previous generations, and that system frameworks can put both engines to work at the same time so apps run models faster.

M6 supports up to 32GB of unified memory and up to 170GB/s of memory bandwidth. That bandwidth figure is a 10 percent increase over M5 and 2.5x the M1.

For background, Apple skipped an M5 generation for the standard Mac mini entirely. The M4 model arrived in October 2024, and reports had tied the jump straight to M6 to industry wide memory shortages that reshaped Apple's roadmap through 2026.

M5 Ultra Goes Quad Die

The M5 Ultra is the more dramatic of the two. Apple calls it its most powerful chip ever, and it is the first M series chip to use a quad die architecture.

Where past Ultra chips fused two Max dies, the M5 Ultra uses next generation UltraFusion technology to connect two dual die M5 Max chips, forming four dies in total. Apple says UltraFusion pushes inter die bandwidth past 4.4TB/s and increases connection density by more than 6x, letting the four dies behave as a single processor.

The chip scales up to a 36 core CPU, made up of 12 super cores and 24 performance cores. Apple puts single threaded performance up to 1.25x higher than the M3 Ultra and multithreaded performance up to 1.3x higher.

The Biggest GPU Apple Has Built

M5 Ultra tops out at an 80 core GPU, with a Neural Accelerator in every core. Apple claims up to 4.5x the peak GPU compute for AI compared to M3 Ultra, and more than 6x the M1 Ultra. On graphics, the company cites up to 40 percent faster performance than M3 Ultra, helped by a second generation Dynamic Caching setup, hardware accelerated mesh shading, and third generation ray tracing.

The chip carries a 32 core Neural Engine and a beefed up Media Engine with four ProRes encode and decode engines, plus hardware accelerated AV1 decode.

Memory is where the M5 Ultra makes its case for AI work. It supports up to 512GB of unified memory and delivers 1.2TB/s of memory bandwidth, which Apple says is 50 percent higher than the M3 Ultra. That pool is large enough to hold LLMs with hundreds of billions of parameters entirely on device.

For Developers

Apple noted that its developer frameworks, including Core AI, Core ML, Metal, and Xcode, tap directly into both chips. Developers can use the Dual 16 core Neural Engine on M6, or the Neural Accelerators and 512GB memory pool on M5 Ultra, to run and fine tune large models locally. Apple's tools automatically spread work across the CPU, GPU, and Neural Engine.

Both chips are shipping now inside the new Mac mini and Mac Studio, which are available to pre order today and begin arriving September 22.