Apple's M4 Chip Praised for Performance, Panned for Repairability
A new iFixit teardown of the M4 MacBook Air praises the chip's raw performance, which is a boon for local AI and creative tasks. However, the review criticizes Apple's continued resistance to user-repairable or upgradeable designs, highlighting a point of friction for builders who value modular hardware.
The M4 chip is built on a second-generation 3-nanometer process, packing 28 billion transistors. Its 16-core Neural Engine is capable of 38 trillion operations per second, a significant jump intended to power on-device AI tasks and what Apple calls the fastest neural processing unit in any AI PC today. In multi-core benchmarks, the M4 chip shows up to a 30% performance increase over the M3. The new 10-core GPU architecture supports hardware-accelerated ray tracing and mesh shading, delivering up to four times the rendering performance of the M2 chip. This raw power is aimed squarely at accelerating creative workflows in photo and video editing, as well as 3D rendering. Despite the performance gains, the M4 MacBook Air received a 5/10 repairability score from iFixit. While ports and batteries are more accessible, the RAM and SSD are soldered to the logic board, making user upgrades impossible. This design choice locks builders into their initial configuration, removing the modularity valued in the PC building community. A primary point of friction is Apple's continued use of software locks and parts pairing. Replacing a display, even with a genuine Apple part from another MacBook, requires calibration with Apple's proprietary System Configuration tool to restore full functionality like True Tone, a significant hurdle for independent repair and refurbishers. This practice persists even as Apple has publicly shifted to support Right to Repair legislation and launched self-service repair programs. Critics argue these software limitations create a "repairability nightmare" that undermines the open repair ecosystem, regardless of the availability of manuals or individual parts. For AI developers, the M4 architecture presents a duality. The M4 Max chip can be configured with up to 128GB of unified memory, allowing it to handle large language models with nearly 200 billion parameters directly on the device. However, some developers note that AI software optimization still lags, with certain applications showing better performance on PCs with high-end NVIDIA GPUs. The performance scaling across the chip lineup creates a clear choice for creators. While the base M4 is sufficient for many creative tasks, professionals working with heavy 3D modeling, animation, or complex video rendering will need to opt for the more powerful M4 Pro or M4 Max chips to avoid performance bottlenecks.
