A20 chip preview: What the M6 reveals for the iPhone 18 Pro

Apple’s unveiling of the M6 chip on Aug. 25 offers the clearest glimpse of the forthcoming A20 chip that will power the iPhone 18 Pro. The M6, built on TSMC’s 2nm N2 process, mirrors the architectural path Apple has taken for its A‑series, meaning the iPhone’s silicon will inherit the same transistor density gains.

A20 chip performance expectations

Analysts project the A20 chip to be up to ↑ 18% faster and roughly 10% more power‑efficient than the A19 Pro, a ratio that mirrors the M6’s edge over the M5. This leap stems from the denser 2nm node and the shift to wafer‑level chip‑on‑wafer‑memory (WMCM) packaging, which shortens data paths between CPU, GPU and Neural Engine.

Neural Engine evolution

The M6 introduces a dual 16‑core Neural Engine, allowing system frameworks to address both engines simultaneously for up to double the peak AI throughput. The iPhone’s Neural Engine has stayed at 16 cores since the A17 Pro, so the A20 chip may adopt a single or dual configuration depending on thermal headroom.

“Apple’s move to WMCM is a game‑changer for on‑device AI,” said a senior analyst at Reuters.

GPU improvements follow the same pattern: the M6 adds 50% higher geometry rates, a refreshed shader core, and hardware‑accelerated ray tracing. Historically, such gains translate directly to the A‑series, so the A20 chip should deliver noticeably smoother gaming and AR experiences.

Memory bandwidth climbs to 170 GB/s on the M6, a 10% rise over the M5’s 153.6 GB/s. Apple appears to be using a faster grade of LPDDR5X rather than moving to LPDDR6, suggesting the A20 chip will hover around 85 GB/s—enough for modest AI workloads but still constrained by on‑device model size.

The media engine retains hardware‑accelerated H.264, HEVC, ProRes and ProRes RAW, and adds AV1 decode but no AV1 encode, mirroring the M5. Given the tight schedule, an AV1 encoder on the A20 chip is unlikely.