Hanshu Tech unveils uHBM and uLPU inference architecture
新智元 · wechat · 2026-08-31
Hanshu Tech unveiled the uHBM® and uLPU™ inference architecture to address the weight transfer bottleneck in LLM Decode stages. By integrating Persistent MRAM with matrix-vector compute on a single die (Weight-Resident Compute), it achieves 24TB/s in-die read bandwidth, avoiding repeated weight transfers across storage interfaces.
More from Embodied
- Tesla FSD Anticipates Swerve to Avoid Debris: User Story — surmenok · 2026-08-31
- Robot That Puts Your Shoes Away When You Get Home Launches in September — chris_j_paxton · 2026-08-31
- AI-powered intelligent plastic ducks hint at a future of droid toys — VoidStateKate · 2026-08-31
- NVIDIA's Kimodo Turns Text Into Full-Body 3D Human and Robot Motion — maier_ak · 2026-08-31
- Microducks robot sells over $2.5M in first 24 hours — _akhaliq · 2026-08-31
- Zhongqing's Shift from Hardware to Embodied AI Brains — 量子位 · 2026-08-31