Apple M5 chip (apple.com)

Article summary

Apple has announced the M5 chip, which delivers a significant boost in AI performance and graphics capabilities compared to its predecessor, the M4. The M5 features a next-generation GPU architecture with a Neural Accelerator in each core, a faster Neural Engine, and higher unified memory bandwidth. The chip is designed to power Apple's new 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro devices. The M5 is built using third-generation 3-nanometer technology and offers improved power efficiency.

Main themes

• Apple M5 chip
• AI performance
• Graphics capabilities
• Neural Engine
• Memory bandwidth
• Apple devices

What commenters say

• The base model M5's memory bandwidth of 153GB/s may not be sufficient to run large language models efficiently.
• The M5's performance is impressive, but the memory capacity is a bigger limitation, with a maximum of 32GB.
• Some commenters believe that the M5's power efficiency and matmul acceleration make it a game-changer for running AI workloads, potentially even more cost-effective than high-end GPUs.
• Others argue that the M5's memory bandwidth is not a significant limitation for most use cases, and that the chip's performance is plenty for the average user.
• There is speculation about the potential performance of the M5 Pro and Max models, with some predicting significantly higher memory bandwidth and capacity.
• Some commenters are disappointed that Apple's software teams seem to be lagging behind their hardware teams in terms of innovation and support for certain features and frameworks.
• The M5's ability to run large language models locally is seen as a major advantage, but some commenters are skeptical about the practical applications and limitations of this capability.
• There is discussion about the trade-offs between model size, quantization, and performance, and how these factors impact the usability of AI models on the M5 chip.