近日，英伟达（NVIDIA）首席执行官黄仁勋宣布，公司下一代AI超级芯片平台“Rubin”已全面投产。作为继Blackwell之后的又一重大突破，Rubin平台将采用更先进的制程工艺和架构设计，进一步提升人工智能模型训练与推理的性能与能效。黄仁勋在GTC 2025大会上表示，Rubin不仅支持万亿参数级大模型的高效运行，还将集成新一代NVLink互连技术和增强型Transformer引擎，显著加速多模态AI任务处理。此外，Rubin平台将与Grace CPU深度协同，构建更强大的AI计算系统，满足数据中心、云服务商及科研机构对算力日益增长的需求。这一发布标志着英伟达持续巩固其在AI硬件领域的领导地位，并为生成式AI、自动驾驶、科学计算等前沿应用提供坚实基础。
Recently, NVIDIA CEO Jensen Huang announced the full-scale production of the company’s next-generation AI superchip platform, ‘Rubin.’ Following the breakthrough Blackwell architecture, Rubin leverages even more advanced manufacturing processes and architectural innovations to significantly boost performance and energy efficiency for AI model training and inference. Speaking at GTC 2025, Huang highlighted that Rubin not only supports efficient execution of trillion-parameter-scale large models but also integrates next-generation NVLink interconnect technology and an enhanced Transformer Engine, dramatically accelerating multimodal AI workloads. Moreover, the Rubin platform will work in deep synergy with the Grace CPU to deliver powerful AI computing systems tailored for data centers, cloud service providers, and research institutions facing ever-growing computational demands. This launch underscores NVIDIA’s continued leadership in AI hardware and provides a robust foundation for cutting-edge applications such as generative AI, autonomous driving, and scientific computing.
原创文章，作者：admin，如若转载，请注明出处：https://ktez.cn/9047.html