AI高計算能力サーバーの熱放散、8W / mkを超える高熱伝導率インターフェース材料を使用

The promotion of ChatGPT technology has further promoted the popularity of high-power application scenarios such as AI computing power. By connecting a large number of corpora to train models and achieve scene functions such as human-computer interaction, a large amount of computing power is required behind it. Synchronization consumption is greatly improved. With the continuous and rapid improvement of chip performance, the problem of heat dissipation has become more prominent.

In order to ensure the stable operation of the server, the operating temperature of high-performance ARM SoC (CPU + NPU + GPU), hard disk and other components should be controlled within the allowable range, so as to effectively ensure that the server has better working ability and longer working life. Due to higher power densities, heat dissipation through advanced thermal management material systems is critical to meeting new functionality standards.

When the AI high-computing server is working, its internal devices will generate a lot of heat, especially the server chip. In view of the heat conduction requirements between the server chip and the heat sink, We recommends 熱伝導性材料 above 8W/mk (thermal pads, heat conduction gel, heat conduction phase change materials), which have high thermal conductivity and good wettability. It can better fill the gap, effectively transfer heat from the chip to the radiator quickly, and then cooperate with the radiator and fan to keep the chip at a low temperature and ensure its stable operation.