Apple doesn’t need LPDDR6, here’s how it will make iPhones faster

While the semiconductor industry and its direct rivals, driven by chips like the anticipated Snapdragon 8 Elite Gen 6 Pro, prepare to embrace the new memory standard LPDDR6, the Cupertino company chooses to take an alternative path.

The iPhone 18 Pro and iPhone 18 Pro Max, equipped with the A20 Pro processor, will retain LPDDR5X memory. This decision, in line with Apple’s historic habit of adopting new technologies at a slower pace, will not translate into a compromise on performance.

Recently surfaced information, including the valuable hints from Tata leaks and the technical schematics shared by Reptalica, explain how the next devices will be able to handle more intensive AI workloads through two important modifications.

iPhone 18 Pro, 6-channel RAM and wider data bus

Locally run AI models require a constant and massive data flow. An undersized memory infrastructure could risk creating severe bottlenecks, forcing the processor to idle while waiting for data to arrive.

To avoid this risk, the A20 Pro chip, which will be Apple’s first SoC produced with a 2 nm process, will abandon the traditional four-channel approach and a 64-bit bus. Instead, the new architecture will support a six-channel memory operating on a 96-bit bus.

This substantial widening drastically increases the maximum bandwidth, ensuring instant data transfer to the Neural Engine, which, in turn, will be designed with larger dimensions than previous generations.

The switch to WMCM format and efficiency

Supporting a six-channel LPDDR5X memory with a 96-bit bus requires a larger physical footprint on the board. This space requirement has driven California engineers to retire the old InFO-PoP packaging method in favor of the more capacious Wafer-Level Multi-Chip Module Packaging (WMCM).

Aside from hosting substantially higher data volumes, this new assembly technique brings with it an essential advantage.

Keeping the DRAM module physically separate from the A20 Pro’s main die, heat dissipation is much quicker and more effective.

A lower and well-managed operating temperature allows the processor to sustain high performance for extended periods, ensuring total smoothness in the complex AI operations without needing an immediate shift to LPDDR6.