Apple's M5 Chips Crush AI Video Tasks

The M5 series marks a significant architectural shift with the introduction of "Fusion Architecture," which bonds two dies into a single system-on-a-chip for the Pro and Max versions. This design allows for a substantial increase in core counts, with both the M5 Pro and M5 Max featuring an 18-core CPU. This new architecture boosts multithreaded performance by up to 30% over the M4 series. A key innovation in the M5 is the inclusion of a Neural Accelerator within each of the up to 40 GPU cores. This design, combined with a faster 16-core Neural Engine, delivers over four times the peak GPU compute for AI tasks compared to the M4. This distributed approach to AI processing is a departure from relying solely on the central Neural Engine and accelerates tasks like on-device transcription and AI-driven image generation. For video workflows, these hardware advancements translate into significant speed improvements. In Blackmagic's DaVinci Resolve Studio, video effects work is up to 5.4 times faster on the M5 Max compared to the M1 Max, and three times faster than the M4 Max. AI image generation is reported to be nearly eight times faster than the M1 Max. This leap is supported by a substantial increase in unified memory bandwidth, which reaches up to 614 GB/s on the M5 Max, crucial for handling large language models and high-resolution video. The increased on-device processing power directly challenges cloud-based workflows common in newsrooms. Local processing offers significant advantages in speed by eliminating network latency and enhances security and privacy by keeping sensitive footage off remote servers. For newsrooms, this means faster turnaround times for breaking news, with AI models running directly on a reporter's laptop to handle tasks like automated transcription, metadata tagging, and creating rough cuts. This shift impacts infrastructure decisions, moving from a predictable cloud operational expense (OpEx) model to an upfront capital expenditure (CapEx) for hardware. While on-premise or on-device infrastructure requires a larger initial investment, it can offer long-term cost benefits for high-utilization workloads and provides greater control over hardware and data sovereignty. Software developers are increasingly optimizing their applications for this on-device power. Apple's own Final Cut Pro shows a significant performance advantage on Apple Silicon, with exports of complex 4K projects being multiple times faster than Adobe Premiere Pro on the same hardware. Video editing platforms are integrating AI to automate repetitive tasks like applying transitions, adding captions, and reformatting content for different social media platforms.