## Apple M5 vs M4: Key Differences ### Fabrication Process - **M4:** TSMC's second-generation 3nm process (N3E) - **M5:** TSMC's third-generation 3nm process (N3P) ### CPU Performance - Both have up to 10-core CPU (4 performance/super + 6 efficiency cores) - **M5:** Up to 15% faster multithreaded CPU performance - **M5:** Introduces "super cores" with improved front-end bandwidth, new cache hierarchy, and enhanced branch prediction ### GPU Performance - **M5:** 10-core GPU with Neural Accelerator in each core - **M5:** Up to 30% faster overall graphics performance vs M4 - **M5:** Up to 45% faster ray tracing performance vs M4 - **M5:** Enhanced shader cores, hardware-accelerated mesh shading, second-generation dynamic caching, and third-generation ray-tracing engine ### AI/ML Performance (Major Upgrade) - **M5:** Over 4× peak GPU compute performance for AI vs M4 - **M5:** Over 6× peak GPU compute performance for AI vs M1 - **M5:** 3.6× faster time to first token (LLM) vs M4 - **M5:** 1.8× faster Topaz Video Enhance AI processing - **M5:** 1.7× faster Blender ray-traced rendering - **M5:** 2.9× faster AI speech enhancement in Premiere Pro ### Neural Engine - Both: 16-core Neural Engine - **M5:** Faster design, accelerates Apple Intelligence and larger local models ### Memory Bandwidth - **M4:** 120 GB/s - **M5:** 153 GB/s (~27.5% increase) ### Memory Capacity - Both: Up to 32 GB unified memory ### Developer APIs - **M4:** Metal 3 developer APIs - **M5:** Metal 4 developer APIs with Tensor APIs to program GPU Neural Accelerators ### Power Efficiency - **M5:** Further efficiency gains at similar power ### Launch Dates - **M4:** Launched in May 2024 - **M5:** Announced October 15, 2025 ### Bottom Line For typical day-to-day usage (browsing, office work, media playback, basic editing), the difference between M4 and M5 is unlikely to be perceptible. The M4 was already a high-performance chip that routinely exceeds the demands of normal Mac and iPad workloads. However, for users whose workloads include on-device AI inference, complex 3D rendering, or other GPU-bound or memory-intensive tasks, the jump from M4 to M5 is material, with multi-fold speed-ups in certain AI operations. Sources: MacObserver, Wikipedia, MacRumors