We continue our article series on critical liquid cooling design and infrastructure updates and why liquid immersion cooling will save the data center. This week, we’ll outline liquid immersion cooling design considerations and explain modern immersion cooling technology.
Liquid immersion cooling design considerations
As the adoption of immersion cooling continues to grow, it is essential to consider the design aspects of these systems to maximize their effectiveness. Several key factors must be taken into account when designing an immersion cooling system:
- Direct-to-Chip Cooling: Direct-to-chip cooling involves applying a coolant directly to the heat- generating components, such as CPUs and GPUs. This method provides more efficient heat dissipation than traditional air cooling, allowing for higher densities and improved performance.
- Cold Plates and Immersion Options: Cold plates are used to target specific hot spots within the hardware, while immersion options include both single-phase and two-phase cooling systems. Single-phase immersion involves submerging components in a liquid that remains in a single state, while two-phase immersion involves a liquid that boils and re-condenses to manage heat. Each option has its advantages, and the choice between them depends on the specific needs of the data center.
- Dunking or Chassis Flooding: These are different approaches to immersion cooling, each with its own benefits. Dunking involves submerging the entire server in a tank of dielectric fluid, while chassis flooding involves flooding the server chassis with fluid. The choice between these methods depends on the design and operational requirements of the data center.
Understanding modern immersion cooling technology
Immersion tank design is a critical aspect of the overall cooling system, as it directly impacts the efficiency and performance of the cooling process. For example, modern immersion tanks are designed with several key features that maximize cooling efficiency and ensure reliable operation.
- Design Considerations in Hypertec: Hypertec specifically engineered a chassis to help the liquid flow from bottom to top. With less obstruction, the liquid is more efficient at dissipating the heat, resulting in cooler temperatures. Further, a key feature to look for are purpose-built immersion machines. The optimized component layout of these servers simplifies maintenance, maximizes thermal efficiency, and improves reliability.
- Retrofitted vs. Immersion-Born Servers: Not all servers are equal regarding immersion cooling. Retrofitted servers are air-cooled servers that have been modified for use in immersion cooling systems.
In retrofitted immersion-ready hardware, the liquid flow encounters obstacles as it navigates chassis housings and a backplane. Additionally, the absence of side openings, originally intended for air, further impedes the free flow.
While they can be effective, they often do not perform as well as immersion-born servers designed from the ground up for immersion cooling. Immersion-born servers, such as those offered by Hypertec, provide better performance, reliability, and cooling efficiency, making them the preferred choice for modern data centers.
- Optimization Within the Tank: Effective immersion cooling goes beyond just submerging servers in the fluid. Careful consideration of fluid dynamics, cable management, and component placement is required to ensure optimal cooling performance. Proper cable management, for example, is essential to prevent cables from obstructing fluid flow and reducing cooling efficiency. Similarly, strategic component placement ensures heat is dissipated effectively throughout the tank.
Special Section: Immersion-born GPU servers
GPUs are driving AI, so GPU-driven AI systems will be the next logical piece of componentry to be leveraged in immersion cooling. Hypertec’s commitment to innovation is exemplified in its Immersion-Born GPU Servers, particularly the Trident iG610R-G6 and iG615R-G6 models. These servers are specifically engineered for immersion cooling, making them a powerful choice for handling the most demanding workloads in modern data centers. For these servers, there are several market-impacting use cases. The applications include, but are not limited to:
- Classical Machine Learning: Train models like decision trees, support vector machines, and clustering algorithms on large datasets. PCIe-based GPUs can efficiently handle these tasks and benefit from immersion cooling for sustained performance.
- Moderate-Scale Deep Learning: Develop convolutional neural networks (CNNs) for image classification or recurrent neural networks (RNNs) for sequence modeling that do not require the highest-end interconnects of SXM or OAM GPUs.
- Edge AI Inference: Deploy trained models for real-time inference in applications like object detection in surveillance systems or anomaly detection in manufacturing processes.
- Batch Inference Processing: This method handles large volumes of data that need to be processed through AI models, such as tagging images or documents, where latency is less critical but throughput is important.
Innovative Heatsink Design
One of the standout features of Hypertec’s Immersion-Born GPU Servers is the advanced heatsink architecture. Unlike traditional air-cooled systems, these servers utilize a heatsink design optimized for the unique properties of dielectric fluids used in immersion cooling. The fins and overall heatsink structure are meticulously crafted to enhance heat dissipation, considering the fluid’s viscosity to maximize cooling efficiency.
Hypertec has also leveraged forced convection add-ons for these heatsinks, further improving the heat dissipation generated by high-performance GPUs. As GPU thermal design power (TDP) continues to rise, with some GPUs reaching up to 1000W TDP, these innovations are critical. The design ensures that the servers maintain optimal temperatures even under extreme loads, something that traditional air cooling would struggle to achieve.
The power of Immersion-Born GPU servers
Hypertec’s immersion-born GPU servers are not just about innovative cooling—they are about delivering raw power and reliability. Supporting the latest generation of PCIe GPUs, these servers are ideally suited for high- demand applications like machine learning, deep learning, HPC, and inferencing (inference training).
By being built from the ground up for immersion cooling, these servers eliminate the inefficiencies of retrofitted systems. They provide superior performance, greater reliability, and extended component lifespan while operating more energy-efficiently. For data centers looking to harness the full potential of GPU-driven workloads, Hypertec’s Immersion-Born GPU Servers represent a robust and future-proof solution.
Key design considerations
When implementing an immersion cooling system, it is important to consider several key factors to ensure the system meets the needs of the data center:
- Vendor Considerations: Selecting the right vendor is critical to the success of an immersion cooling implementation. Data centers should evaluate vendors based on their experience, product features, and support capabilities. Hypertec, for example, offers a range of immersion-born servers specifically designed for immersion cooling, providing a reliable and efficient solution for data centers. These include immersion-born GPU, compute, and storage servers.
- Review Process: Regular reviews and updates to the immersion cooling system are necessary to ensure it continues to meet the evolving needs of the data center. This includes evaluating the system’s performance, identifying areas for improvement, and making necessary adjustments to maintain optimal cooling efficiency.
- Challenging Vendors on Features and Capabilities: It is important to challenge vendors on the features and capabilities of their products to ensure they meet the specific needs of the data center. This includes evaluating the scalability of the system, its ability to handle different workloads, and its long-term sustainability.
- Serviceability and Maintenance: Easy access to components and simple maintenance processes are critical for minimizing downtime and ensuring the longevity of the immersion cooling system. Hypertec’s immersion-born servers, for example, are designed for easy serviceability, with tool-less PCIe card cages and other features that simplify maintenance operations.
Download the full report, Liquid Immersion Cooling Will Save the Data Center: Critical Design and Infrastructure Updates, featuring Hypertec, to learn more. In our next article, we'll discuss how to get started on a liquid immersion cooling journey.
