Apple introduced two desktop processors, the M6 and M5 Ultra, positioning them as major advances in computing performance and artificial intelligence. The M6 will power a new Mac mini, while the M5 Ultra will appear in a new Mac Studio, according to the company.[1]
The M6 is Apple’s first chip manufactured using a 2-nanometer process. Its 12-core CPU combines two super cores, four performance cores and six efficiency cores. Apple says the processor provides up to 1.2 times the multithreaded performance of the M5 and up to 2.4 times that of the M1. The company also describes its CPU core as the world’s fastest, though the supplied announcement does not include independent testing.[1]
For graphics, M6 includes a 12-core GPU with a Neural Accelerator in every core. Apple claims this configuration raises peak GPU compute for AI by nearly 30 percent over M5 and by more than eight times over M1. Updated graphics technology includes Dynamic Caching, hardware-accelerated ray tracing and changes to the shader architecture. The company also reports a 50 percent increase in geometry processing rates.[1]
A Dual 16-core Neural Engine is intended to accelerate AI models running locally. System frameworks can use both engines at once, Apple said. The processor supports as much as 32GB of unified memory and offers bandwidth of up to 170GB per second, which Apple places 10 percent above M5 and 2.5 times above M1.[1]
M5 Ultra targets more demanding professional and AI work in the Mac Studio. It uses Apple’s UltraFusion technology to connect two dual-die M5 Max chips, creating the company’s first quad-die M-series system on a chip. Apple reports more than 4.4TB per second of inter-die bandwidth and says the links allow the four dies to operate as one processor.[1]
At its highest configuration, M5 Ultra has a 36-core CPU and an 80-core GPU. Apple claims up to 1.25 times the single-threaded performance and 1.3 times the multithreaded performance of M3 Ultra. It also says graphics performance can be up to 40 percent faster, while peak GPU compute for AI reaches 4.5 times the M3 Ultra level.[1]
The chip supports up to 512GB of unified memory with 1.2TB per second of bandwidth, 50 percent more than M3 Ultra. Apple says that capacity can keep large datasets and language models with hundreds of billions of parameters in local memory. A 32-core Neural Engine and dedicated hardware for H.264, HEVC, ProRes and AV1 workloads round out the design.[1]
Apple said its Core AI, Core ML, Metal and Xcode tools can access the chips’ specialized hardware. The performance comparisons and efficiency assertions in the announcement remain manufacturer claims rather than independently verified benchmark results.[1]
