Rumours have it that Apple’s M5 chips may return to the more traditional format of separate CPU and GPU design. It’s an interesting bit of hearsay, given how deeply invested Apple has been since the launch of the M1 chips back in 2020.
According to Ming-Chi Kuo, an analyst renowned for his fairly reliable predictions and rumours of Apple products, the M5 chip lineup could ditch the Unified Memory architecture in favour of the conventional split component application. Additionally, the next-generation chip and all its variants will be fabricated on the TSMC N3P process node.
Apple M5 series chip
1. The M5 series chips will adopt TSMC’s advanced N3P node, which entered the prototype phase a few months ago. M5, M5 Pro/Max, and M5 Ultra mass production is expected in 1H25, 2H25, and 2026, respectively.
2. The M5 Pro, Max, and Ultra will utilize… https://t.co/XIWHx5B2Cy— 郭明錤 (Ming-Chi Kuo) (@mingchikuo) December 23, 2024
The Unified Memory architecture is one of the main appeals of the M-Series silicon. To put it simply, the chip basically shares a memory pool between CPU and GPU cores, which has proven to be incredibly power efficient, especially in the performance-per-watt segment for MacBooks. For the record, Apple isn’t the only company to make use of the Unified Memory technology; Sony’s PlayStation 5 consoles also make use of it, albeit with a different custom AMD APU.
Getting back on track, Ming-Chi Kuo says that the M5 Pro, Max, and Ultra chips will utilise server-grade SoIC packaging, using something called 2.5D packaging, which will improve production yields and thermal performance with the split CPU+GPU design.
Another reason for the change in architecture is an unsurprising one: AI inferencing and processing. Given how the medium has exploded in popularity, especially in Cloud Computing, it makes sense that Apple will want its M5 Chips to be more than capable in the field.
The M5 silicon is expected to find its way into Mac products by late 2025, with the M5 Ultra expected to make its appearance by 2026.
