Apple presents the M6 ​​and M5 Ultra processors, the most powerful for local AI processing

The M5 Ultra uses UltraFusion to connect two dual-die M5 Max to form a quad architecture. Together, these low-latency, high-bandwidth interconnects allow all four chips to operate as a single unified processor.

The processor has a CPU with up to 36 cores, made up of twelve super cores and 24 performance cores, offering up to 1.25 times higher single-thread performance and up to 1.3 times higher multi-thread performance compared to the M3 Ultra.

Regarding the GPU, this component integrates up to 80 cores, with a Neural Accelerator in each of them, which multiplies the maximum processing capacity for AI by up to 4.5 compared to the M3 Ultra.

The M5 Ultra also includes a 32-core Neural Engine and a multimedia engine with hardware H.264 and HEVC, four ProRes encoding and decoding engines, and hardware-accelerated AV1 decoding.

In addition, it offers up to 512 GB unified memory and 1.2 TB/s of unified memory bandwidth, which is 50 percent more than the M3 Ultra. This allows you to store huge data sets in local memory, increase the rate of tokens per second, and run large language models with hundreds of billions of parameters on the device itself.