“M3 Ultra is the pinnacle of our scalable system-on-a-chip architecture, aimed specifically at users who run the most heavily threaded and bandwidth-intensive applications,” said Johny Srouji, Apple’s Senior VP of Hardware Technologies.
The M3 Ultra offers a 32-core CPU, featuring 24 performance cores and 8 efficiency cores, providing:
• 1.5x the performance of the M2 Ultra
• 1.8x the performance of the M1 Ultra
• Up to 2.6x faster GPU performance than M1 Ultra
Apple says its Neural Engine, built for AI tasks, can process large language models (LLMs) with over 600 billion parameters directly on-device. The AI acceleration engines in M3 Ultra are optimised for Apple Intelligence, boosting AI-driven workflows in image processing, video rendering, and computational design.
With up to 512GB of unified memory, M3 Ultra surpasses most workstation GPUs, making it ideal for AI development, 3D rendering, and high-resolution video editing. The memory bandwidth exceeds 800GB/s, ensuring seamless multitasking and faster data transfer.
Thunderbolt 5 integration enables:
• 120Gbps data transfer speeds (2x Thunderbolt 4)
• Support for multiple Mac Studio systems in parallel processing workflows
• Connectivity for eight Pro Display XDR monitors
M3 Ultra incorporates Apple’s custom UltraFusion interconnect, allowing 2.5TB/s low-latency communication between chips. The dedicated media engine supports:
• 4 ProRes encode/decode engines
• Real-time playback of up to 22 streams of 8K ProRes 422 video
• Hardware-accelerated mesh shading and ray tracing for ultra-realistic graphics
Security has also been enhanced, with Apple’s Secure Enclave, hardware-verified secure boot, and runtime anti-exploitation protections.