Partager

Étude numérique de l’amélioration des performances thermiques dans les centres de données à haute densité d’énergie à l’aide du refroidissement liquide par immersion.

Computational study of thermal performance enhancement in high-energy density data center through immersive liquid cooling.

Numéro : 2374

Auteurs : KIM J., LEE H.

Résumé

The escalating capacity and demand in data centers, driven by advancements in information technology such as artificial intelligence and machine learning, necessitate increased computing power. The heightened performance and energy consumption of central processing units and graphics processing units in data centers pose challenges to traditional air-cooling methods. Fan-based air cooling exhibits limitations with low cooling capacity and high energy consumption. Consequently, there is a growing focus on alternative cooling methods, particularly hybrid cooling with cold plates and immersion cooling using dielectric fluid. Immersion cooling relies on an electrically insulating and thermally stable dielectric fluid to directly cool electronic components.
This study seeks to compare different data center cooling techniques using the computational fluid dynamics simulation tool. A server model, featuring 2 central processing units and 16 memories, is employed for validation purposes and stacked on top of each other for imitation of the server rack. A simplified porous media model is used to replace the heatsink. The investigation involves a comparison of diverse performance parameters including central processing unit temperature, memory temperature and thermal resistance where immersion cooling emerges as the most efficient. Particularly, under controlled conditions of inlet fluid temperature and power consumption, the average central processing unit temperature in the immersion cooling is found to be 102.58°C and 22.89°C lower than that observed in air cooling and hybrid cooling, respectively, at 800 W central processing unit thermal design power. Further comparison is conducted for various central processing unit thermal design power and memory thermal design power.

Documents disponibles

Format PDF

Pages : 9 p.

Disponible

Gratuit

Télécharger

Détails

  • Titre original : Computational study of thermal performance enhancement in high-energy density data center through immersive liquid cooling.
  • Identifiant de la fiche : 30033156
  • Langues : Anglais
  • Source : 2024 Purdue Conferences. 19th International Refrigeration and Air-Conditioning Conference at Purdue.
  • Date d'édition : 17/07/2024

Liens


Voir d'autres communications du même compte rendu (51)
Voir le compte rendu de la conférence

Indexation