End-to-End Performance Comparison in Real-World Applications

This study presents an end-to-end performance comparison between two distinct hardware architectures: an Intel-based MacBook Pro equipped with DDR4 memory and an Apple M3 MacBook Pro featuring unified memory. The Intel system utilizes a 2.4 GHz 8-core Intel Core i9 processor (boostable to 5.0 GHz) with a 16 MB shared L3 cache, while the Apple M3 Max integrates a 16-core CPU and a 40-core GPU with a 400 GBps unified-memory bandwidth. In both environments, we measured and compared the inter-module latency across a complete video-processing pipeline, from video input to video output. The results highlight the architectural differences between discrete memory systems and unified memory architectures, particularly in how memory bandwidth, cache sharing, and data-transfer mechanisms affect real-time video processing performance. This evaluation provides practical insight into the impact of memory architecture and CPU-GPU integration on the end-to-end latency of real-world multimedia applications.