Two-Phase Cooling System Performance Under Different Operating Scenarios (ITherm 2025)

Overview

This document is a paper by Pritish Parida and Timothy Chainer, presented at IEEE ITherm 2025 and supported by ARPA-E award DE-AR0001577. The study used IBM's component and system models and the Sim2Cool tool to study above-ambient pumped two-phase cooling of a multi-server rack rejecting heat to dry coolers.

The model compares refrigerants: R1234ze(E), R1234ze(Z), R1233zd(E) and R1336mzz(Z). Scenarios include a typical meteorological year in several US cities, different facility-side set points, internal and external pump and dry-cooler fan speeds, and equal flow versus equal exit vapour quality. The metrics evaluated are thermal resistances, pumping power, cooling COP and junction temperatures. This framing shows that the study treats both component and system scale.

The main result in the abstract is that medium-pressure (for example R1234ze(Z)) and high-pressure (for example R1234ze(E)) refrigerants performed better than low-pressure ones such as R1336mzz(Z), at both component and system level. The work targets water-free, energy-efficient cooling of dense AI packages. The underlying question is whether a two-phase system running with dry coolers can keep adequate temperature margin through the year, answered with a model-based approach.

Because this record rests on the abstract rather than the full text, numerical results, model validation and assumptions are not covered. The results are model outputs based on US climate data; the same findings should not be assumed to hold in other climates such as Turkey's. It still offers a guiding framework on which parameters should be evaluated when choosing a refrigerant and designing dry-cooler-based two-phase cooling.

Key points

1. Refrigerants compared include R1234ze(E), R1234ze(Z), R1233zd(E) and R1336mzz(Z).
2. Scenarios: typical meteorological year in several US cities, different facility-side set points, pump and dry-cooler fan speeds, equal flow vs equal exit vapour quality.
3. Medium-pressure (e.g. R1234ze(Z)) and high-pressure (e.g. R1234ze(E)) refrigerants performed better than low-pressure ones such as R1336mzz(Z).
4. Metrics evaluated: thermal resistances, pumping power, cooling COP and junction temperatures.
5. The study covers above-ambient pumped two-phase cooling of a multi-server rack rejecting heat to dry coolers.
6. The target is water-free, energy-efficient cooling of dense AI packages.
7. Metadata-only (abstract) record; the full text was not read.

Using it in design

• Record the refrigerant (with pressure class and GWP) on two-phase CDU objects.
• In climate analysis, evaluate dry-cooler-only two-phase systems over typical-year weather to verify junction-temperature margins.
• Compare refrigerant choice not only by thermal performance but also by pressure class and environmental impact.
• Recompute results with your own climate data; do not assume US city results apply directly.

Who should read it

Thermal designers and research engineers evaluating liquid and two-phase cooling systems.

Publication details

Authors

Pritish Parida; Timothy Chainer

Publisher

IEEE

Year

2025

Document number

ITherm 2025 (ARPA-E award DE-AR0001577)

Event

IEEE ITherm 2025

Document language

English

Source type

Standard

Access

Citation only

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation