Single-phase immersion liquid cooling is a highly efficient cooling method for CPUs. In order to address the limitations of temperature uniformity and cooling efficiency associated with centralized liquid supply in traditional cooling under fluctuating computing loads, a novel jet device was designed. To quantitatively evaluate its performance, an experimental platform based on a dual-CPU server was established for investigating the effects of various operating strategies on heat dissipation performance, temperature uniformity, energy consumption, and efficiency. The results show that while traditional liquid cooling offered improvements over air cooling by reducing CPU core temperature by 2~9 ℃, cutting server power consumption by 16.6~21.9%, and increasing Data Center Infrastructure Efficiency (DCIE) by 4~9%, it fails to effectively control the inter-CPU temperature difference. After the jet device is activated, the CPU core temperature is further reduced by 5~13 ℃, the local heat transfer coefficient of the CPU is improved by approximately 1.5 times, with a maximum value of 3 553.8 (W·m?2·°C?1), and the maximum temperature difference between CPUs is decreased by 8 ℃. Fluctuations in load has minimal impact on the average temperature of the measuring points on CPU heat sink fins, and temperature uniformity is greatly improved. The coolant temperature across measuring points is highly uniform, averaging approximately 24.5 ℃ with a root mean square deviation near zero. The equivalent thermal resistance of the CPUs is reduced to as low as 0.12~0.13 ℃/W, and the efficiency indexηis increased by 20 to 26 times. It is concluded that the proposed jet device achieves a successful balance between improving cooling performance and enhancing overall system efficiency.