Ai Hardware

Nvidia Vera Rubin NVL72 beats Blackwell by 7x in DeepSeek inference test: Inference tests on a 1.6-trillion parameter DeepSeek model show Nvidia's Vera Rubin NVL72 delivers up to seven times better token throughput per megawatt compared to Blackwell. This performance exceeds CEO Jensen Huang's previous claim of a threefold improvement for models ranging from 1T to 3T parameters.

Hardware • Techmeme

China Mobile Cloud debuts GPU-neuromorphic inference system: China Mobile Cloud and its partners have introduced a domestic GPU plus neuromorphic heterogeneous mixed-inference system for large models. Tested on DeepSeek V4 Flash, the system reportedly delivers approximately twofold output and energy gains alongside a 40% reduction in operating expenses.

Hardware • Pandaily

CXMT reportedly starts HBM3E risk production with samples sent to T-Head: ChangXin Memory Technologies has reportedly started risk production for HBM3E, with early samples undergoing qualification at Alibaba's T-Head and Cambricon. Yields, volumes, and final specifications remain undisclosed. Optimistic timelines suggest commercial production could begin around 2027.

Hardware • Pandaily