Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems

Q: What type of study is this?

This is a Experimental Study study.

Imran Latif; Muhammad Ali Shafique; Hayat Ullah; Alex C. Newkirk; Xi Yu; Arslan Munir

doi:10.48550/arxiv.2507.16781

July 22, 2025Open Access

Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems

View Full Paper

Ask AI

Bookmark

Share

Key Points

Liquid-cooled systems maintain GPU temperatures between 41-50 degrees Celsius, enhancing thermal stability.
Under load, liquid cooling yields 54 TFLOPs per GPU, 17 percent higher than air cooling at 46 TFLOPs per GPU.
Energy overhead is reduced with liquid cooling, improving performance per watt in data centers.
These findings highlight the sustainability benefits of liquid cooling in hyperscale data centers.

No abstract yet

We only have this paper's title and publication details.

Read it at the source

Implication

This research benchmarks large language models and vision-language models on liquid-cooled and air-cooled systems, showing energy benefits for data centers.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Latif et al. (2025) studied this question.

synapsesocial.com/papers/68d4759931b076d99fa6da9c https://doi.org/10.48550/arxiv.2507.16781

View Full Paper

Ask AI

Bookmark

Share

Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems

Key Points

No abstract yet

Implication

Discussion

Cite This Study

Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems

Key Points

No abstract yet

Implication

Discussion

Cite This Study

Keep reading

More related papers

Keep reading

More related papers