Apple unveiled the M5 Ultra and M6 chips, which power new Mac Studio and Mac Mini models. The M5 Ultra focuses on demanding computing workloads such as 3D rendering, visual effects, and running advanced AI models, while the M6 delivers increased AI processing performance for users and developers who rely on running models locally.
M5 Ultra: A Quad-Chip Design
Apple describes the M5 Ultra as its most powerful chip to date. The processor combines two dual-die M5 Max chips into a single quad-die design, making it the first Apple processor to use this approach, according to the announcement. The M5 Ultra includes up to 36 CPU cores and up to 80 GPU cores, as well as unified memory bandwidth of 1.2 terabytes per second.
This memory bandwidth represents a 50% increase over the M3 Ultra, which Apple launched only about a year and a half ago. These capabilities target users who need to move large amounts of data between the processor and memory while performing tasks such as complex rendering, producing visual effects, and running advanced AI models.
M6 Boosts On-Device AI Performance
The M6 is based on a 2-nanometer manufacturing process and features a new CPU architecture, two additional CPU and GPU cores, and a dual neural engine consisting of 16 cores, along with higher unified memory bandwidth.
Apple says the M6's CPU cluster consists of 12 cores, two more than the M5. The chip also delivers, according to the company, an increase approaching 30% in peak GPU processing performance for AI workloads compared with the M5. This is reflected particularly in faster prompt processing when using large language models on-device.
What Changes for Developers?
Apple says developers will be able to use these processors and its software frameworks to run and fine-tune large AI models locally on Mac devices. Apple's developer tools and frameworks distribute workloads among the CPU, GPU, and neural engine.
The software environment also enables the use of Apple Foundation Models, access to Apple Intelligence capabilities through App Intents, or the operation of AI models owned by the developer. This makes the device itself an environment for developing and running AI workloads, rather than relying entirely on cloud services.
Why Does This Announcement Matter?
The announcement's practical value is not limited to increasing the number of cores or memory speed; its primary focus is moving more AI workloads onto the device. This may give developers greater control over their data and operating environment, particularly for tasks they do not want to send to the cloud. However, the source does not provide independent measurements of power consumption or a comprehensive comparison with competitors' processors, nor does it clarify the limits on the model sizes that can be run locally or provide details about device availability in each market.
The new Mac Mini devices are available for preorder starting at $899, with shipping set to begin after September 22. The source did not mention a price or specific availability date for the Mac Studio model equipped with the M5 Ultra chip.