MediaTek Dimensity 9600 Pro Unveiled: 2nm Node, LPDDR6, and AI Upgrades
by Kreso · NokiamobMediaTek has officially unveiled the Dimensity 9600 Pro, marking its flagship transition to TSMC’s cutting-edge 2nm process node. Built around ARM’s latest C-core architecture and housing over 33 billion transistors, the mobile platform leverages a 2+3+3 CPU layout to deliver generational leaps in both computational power and energy efficiency.
Dimensity 9600 Pro: Architectural Overhaul and Benchmark Gains
At the node level, moving to 2nm silicon yields a 14% performance boost and a 23% reduction in power consumption compared to 3nm designs. MediaTek builds on this foundation with a fourth-generation “all-big-core” CPU setup:
- Super Cores: 2x C2-Ultra cores clocked at 4.55GHz (2MB L2 cache each)
- Performance Cores: 3x C2 Pro cores at 4.35GHz (1MB L2 cache each)
- Efficiency Cores: 3x C2 Pro cores at 3.10GHz (512KB L2 cache each)
With 16MB of L3 cache, total cache reaches 34.5MB—a 21% increase in L2 capacity over the Dimensity 9500. This translates to a 17% jump in single-core performance (scoring 4,276 on Geekbench 6.4) and a 15% bump in multi-core tasks (12,650 points). Crucially, multi-core power consumption drops by 61% at matching peak workloads.
Dimensity 9600 Pro: World-First Support for LPDDR6 and UFS 5.0
Setting a new industry benchmark, the Dimensity 9600 Pro is the first mobile chipset to support next-generation LPDDR6 memory (33% faster than LPDDR5X) and UFS 5.0 storage, which doubles the sequential read/write throughput of dual-channel UFS 4.0.
Graphics duties are handled by the new G2-Ultra NX GPU, which delivers 27% higher peak performance, 18% faster ray tracing, and mobile gaming frame rates up to 185fps—all while cutting power draw by 24%. Camera processing is spearheaded by the Imagiq 1290 ISP, enabling 4K video recording at 240fps for slow-motion capture alongside hardware-level H.266 decoding.
On-Device Generative AI Breakthroughs
Anchored by an “AI-native” architecture, the chipset features a dual NPU configuration: the 10th-generation high-performance NPU 1090 and a low-power, always-on secondary NPU. The updated Generative AI Engine 3.0 doubles INT4 computing power, allowing the processor to run Mixture-of-Experts (MoE) models with up to 30 billion parameters entirely on-device. MediaTek’s new Tianji Intelligent Computing Fusion Architecture seamlessly orchestrates workloads across the CPU, NPU, and memory to prevent resource bottlenecks during intensive multitasking.