微软的新微流体冷却技术将AI芯片温度提高65%, 提升性能和效率。 Microsoft’s new microfluidic cooling tech slashes AI chip temps by 65%, boosting performance and efficiency.

微软为人工智能芯片开发了一个原型微氟化物冷却系统,该系统使用直接嵌在芯片上的微小通道来循环冷却剂,比传统方法将峰值温度降低65%。 Microsoft has developed a prototype microfluidic cooling system for AI chips that uses tiny channels etched directly onto the chip to circulate coolant, reducing peak temperatures by up to 65% compared to traditional methods. 该技术在Redmond校园进行测试,能够提高性能,更安全地超时超时工作,通过允许芯片在不减速的情况下跑得更热,从而提高数据中心的运作效率。 The technology, tested at its Redmond campus, enables higher performance, safer overclocking, and more efficient data center operations by allowing chips to run hotter without throttling. 它支持3D堆叠等先进设计,并提高能源效率,有可能降低电网需求和废热。 It supports advanced designs like 3D stacking and improves energy efficiency, potentially lowering grid demand and waste heat. 该系统正在纳入微软定制的AI芯片,包括Maia加速器和钴CPU,并可能有益于更广泛的行业。 The system is being integrated into Microsoft’s custom AI chips, including the Maia Accelerator and Cobalt CPU, and may benefit the broader industry. 虽然它仍在发展之中,但它是朝向解决AI硬件日益增长的热量挑战迈出的重要一步。 While still in development, it represents a major step toward solving AI hardware’s growing thermal challenges.