在算力竞赛中,华为祭出了「以量换质」的非常规战术。最新技术白皮书显示,由384颗昇腾910C NPU组成的CloudMatrix 384集群,在运行深言科技6710亿参数的R1大模型时,算力表现已超越英伟达H800系统。这相当于用4倍能耗换来1.7倍性能——在电力资源充沛的中国市场,这种trade-off正在被重新定义。
Related Articles
- Huawei releases new tool to get Chinese firms around crushing HBM export blacklist — new UCM software claims up to 22x throughput gain and 90% latency reduction for traditional cache hierarchies in AI workloads3 days ago
- GPU Servers Boost Data Center Speed4 days ago
- US reportedly plans to curb sales of AI GPUs to Malaysia and Thailand to prevent smuggling to Chinaabout 1 month ago
- Dell and CoreWeave Show Off First NVIDIA GB300 NVL72 Rackabout 1 month ago
- Nvidia's Arm chips rapidly gain share in server market as AI booms — Nvidia's Arm-powered GB200 servers surge as market reaches a record $95 billion in the first quarterabout 1 month ago
- Nvidia reportedly books entire server plant capacity through 2026 to build Blackwell and Rubin AI servers, pushing out other potential customersabout 2 months ago
- Alchip responds to report that Nvidia is reportedly keeping prized NVLink tech closely guarded (Updated)about 2 months ago
- How Hard Can It Be?2 months ago
- Nvidia to boost AI server racks to megawatt scale, increasing power delivery by five times or more3 months ago
- Nvidia GB200 production ramps up after suppliers tackle AI server overheating and liquid cooling leaks3 months ago