STUDI EKSPERIMENTAL KINERJA TERMAL SISTEM SINGLE-PHASE IMMERSION COOLING UNTUK DATA CENTER DENGAN VARIASI WORKLOAD KOMPONEN DAN LAJU ALIRAN FLUIDA

Daffa Raihan, Ir. Indro Pranoto, S.T., M.Eng., Ph.D., IPM., ASEAN Eng.

2026 | Skripsi | TEKNIK MESIN

Peningkatan kebutuhan komputasi pada data center menyebabkan konsumsi daya dan beban panas pada komponen elektronik semakin tinggi, terutama pada CPU dan GPU. Konsumsi energi data center diperkirakan dapat mencapai sekitar 4,5?ri total konsumsi listrik global, dengan sekitar 40% energi digunakan untuk sistem pendinginan. Selain itu, tingkat kegagalan perangkat elektronik dapat meningkat ketika temperatur melebihi 75 °C. Sistem fan cooling masih banyak digunakan, tetapi memiliki keterbatasan dalam menangani densitas panas tinggi karena umumnya hanya mampu menangani beban panas sekitar 37 W/cm², sedangkan sistem pendinginan berbasis cairan dapat mencapai sekitar 100 W/cm². Oleh karena itu, penelitian ini dilakukan untuk menganalisis kinerja termal sistem single-phase immersion cooling sebagai alternatif pendinginan data center dengan mempertimbangkan variasi workload komponen dan laju aliran fluida.

Penelitian dilakukan secara eksperimental menggunakan fluida dielektrik S3X. Variasi laju aliran fluida yang digunakan adalah 1 LPM, 1,75 LPM, dan 2,5 LPM, sedangkan variasi workload meliputi CID-GID, C50-GID, C100-GID, C50-G100, dan C100-G100. Parameter yang dianalisis meliputi temperatur CPU dan GPU, heat transfer coefficient, thermal resistance, heat absorbed, heat loss, pressure drop, daya pompa, Cooling Performance Index (CPI), dan Power Usage Effectiveness (PUE). Hasil pengujian kemudian dibandingkan dengan sistem fan cooling sebagai data pembanding.

Hasil penelitian menunjukkan bahwa peningkatan laju aliran fluida mampu menurunkan temperatur CPU dan GPU, meningkatkan heat transfer coefficient, menurunkan thermal resistance, meningkatkan heat absorbed, serta menurunkan heat loss. Dibandingkan dengan fan cooling, sistem single-phase immersion cooling menghasilkan performa pendinginan yang lebih baik. Pada kondisi C100-G100, temperatur CPU menurun dari 75,09 °C pada fan cooling menjadi 49,33 °C pada immersion cooling 2,5 LPM, sedangkan temperatur GPU menurun dari 79,97 °C menjadi 64,73 °C. Nilai PUE juga menurun dari 1,57 pada fan cooling menjadi 1,09 pada immersion cooling 1 LPM. Dengan demikian, sistem single-phase immersion cooling terbukti lebih efektif dibandingkan fan cooling, meskipun pemilihan laju aliran optimum tetap perlu mempertimbangkan keseimbangan antara performa termal dan konsumsi daya pompa.

The increasing demand for computing performance in data centers has led to higher power consumption and heat generation in electronic components, especially CPUs and GPUs. Data center energy consumption is estimated to reach approximately 4.5% of global electricity consumption, with around 40% of the energy used for cooling systems. In addition, the failure rate of electronic devices may increase when the temperature exceeds 75 °C. Fan cooling is still widely used; however, it has limitations in handling high heat density, as air cooling generally manages heat loads of around 37 W/cm², while liquid-based cooling systems can handle up to approximately 100 W/cm². Therefore, this study aims to analyze the thermal performance of a single-phase immersion cooling system as an alternative cooling method for data center applications by considering variations in component workload and coolant flow rate.

The study was conducted experimentally using S3X dielectric fluid. The coolant flow rates were varied at 1 LPM, 1.75 LPM, and 2.5 LPM, while the workload variations consisted of CID-GID, C50-GID, C100-GID, C50-G100, and C100-G100. The analyzed parameters included CPU and GPU temperatures, heat transfer coefficient, thermal resistance, heat absorbed, heat loss, pressure drop, pump power, Cooling Performance Index (CPI), and Power Usage Effectiveness (PUE). The experimental results were also compared with a fan cooling system as a baseline.

The results show that increasing the coolant flow rate reduced CPU and GPU temperatures, increased the heat transfer coefficient, decreased thermal resistance, increased heat absorbed, and reduced heat loss. Compared with fan cooling, the single-phase immersion cooling system provided better cooling performance. Under the C100-G100 condition, the CPU temperature decreased from 75.09 °C with fan cooling to 49.33 °C with immersion cooling at 2.5 LPM, while the GPU temperature decreased from 79.97 °C to 64.73 °C. The PUE also decreased from 1.57 with fan cooling to 1.09 with immersion cooling at 1 LPM. Therefore, singlephase immersion cooling is proven to be more effective than fan cooling. However, the optimum flow rate should still be selected by considering the balance between thermal performance and pump power consumption.

Kata Kunci : data center, fan cooling, single-phase immersion cooling, thermal resistance, PUE