Intel’s Gaudi3 AI Accelerator to Launch with Innovative Two-Phase Liquid Cooling System, Developed in Partnership with Vertiv

In the rapidly evolving field of artificial intelligence (AI), Intel’s Gaudi3 AI accelerator is set to make a significant impact when it launches in 2024. Recognizing the increasing power demands of AI, Intel has partnered with Vertiv, a data center infrastructure solutions provider, to develop an innovative two-phase liquid cooling system. This collaboration aims to address the growing need for efficient heat removal and sustainability in the face of escalating AI power consumption.

Two-Phase Liquid Cooling for Gaudi3 AI Accelerator

At the core of the Gaudi3 AI accelerator is a pumped two-phase cooling system, jointly developed by Intel and Vertiv. This cutting-edge solution is capable of removing up to an impressive 160kW of heat, utilizing water temperatures ranging from 17°C to 45°C. The two-phase liquid cooling system excels at efficiently transferring heat away from the electronics, ensuring optimal performance and reliability for the AI accelerator.

The advantages of two-phase liquid cooling

Two-phase liquid cooling is an effective method for heat removal in high-performance computing. It enables the coolant fluid to boil while in contact with hot electronics, extracting heat more efficiently than traditional cooling methods. By harnessing the thermal properties of the coolant, this advanced cooling system optimizes performance, enabling faster processing speeds and preventing overheating.

P2P Cooling System for Air-Cooled Data Centers

In addition to the pumped two-phase cooling system for the Gaudi3 AI accelerator, Vertiv has also developed a P2P (pumped refrigerant-to-pump refrigerant) cooling solution specifically designed for air-cooled data centers. This alternative system can handle heat loads of up to 40 kW, offering a practical option for cooling requirements in a range of computing environments.

Benefits of Vertiv’s P2P refrigerant-based cooling solution

Vertiv’s P2P cooling system is not only highly efficient but also offers additional advantages for data center operators. By implementing this refrigerant-based solution, customers can enable heat reuse and utilize free air cooling. The system’s efficiency leads to reduced power consumption and lower water usage, contributing to sustainability goals and environmental preservation.

Intel’s Gaudi processors for AI training

Intel’s Gaudi processors are specifically designed for AI training, positioning the company as a strong competitor to industry leader Nvidia. With significant computing power and advanced capabilities, Gaudi processors are poised to drive AI advancements across various sectors, from autonomous vehicles to healthcare and financial services.

Growing Power Demands of AI

The power demands of AI are escalating rapidly, making energy consumption a significant concern. According to projections, the AI sector is expected to consume as much energy as the entire Netherlands by 2027. As the AI industry continues to expand, it becomes imperative to address the energy requirements and seek sustainable solutions to mitigate the environmental impact.

Two-Phase Liquid Cooling and Challenges

While two-phase liquid cooling is an effective heat removal method, it has faced challenges related to the use of per- and polyfluoroalkyl substances (PFAS), known as “forever chemicals.” PFAS are considered a health risk and require alternative solutions to ensure the safety and sustainability of cooling systems. The collaboration between Vertiv and Intel aims to explore and develop alternatives that are both effective and environmentally friendly.

Vertiv and Intel Collaborate for Cooling Innovation

Vertiv and Intel’s partnership signifies a commitment to driving technological innovation in cooling solutions for next-generation AI accelerators. By addressing the increasing thermal design power and heat flux requirements, this collaboration aims to meet the evolving needs of customers and support the advancement of AI technology. Additionally, the partnership aligns with sustainability goals, contributing to reduced carbon footprints and increased energy efficiency.

Intel’s Gaudi3 AI accelerator, combined with Vertiv’s two-phase liquid cooling system, represents a significant breakthrough in the field of AI technology. The collaboration between these industry leaders not only enhances the efficiency and performance of the Gaudi3 accelerator but also enables customers to achieve their sustainability objectives. As the power demands of AI continue to rise, innovative solutions like two-phase liquid cooling will play a crucial role in driving the future of AI while minimizing its environmental impact. With cutting-edge advancements in cooling technology, the partnership between Vertiv and Intel is poised to shape the landscape of AI computing and pave the way for a sustainable future.

Explore more

Ethereum Plans Major Glamsterdam Upgrade for Late 2026

Ethereum developers are currently finalizing the specifications for the Glamsterdam hard fork, which represents the next major milestone in the network’s ongoing evolution toward a more scalable and efficient global computer. This upcoming transition is not merely a routine update but a comprehensive overhaul of several critical components that have defined the network since its inception. By addressing long-standing technical

How Does Databricks CustomerLake Redefine the Agentic CDP?

The landscape of customer data management is currently undergoing a seismic transformation as the traditional boundaries between storage, analysis, and execution are being dismantled by the rise of the Data Intelligence Platform. For years, enterprises have struggled with the fragmentation tax, which represents the hidden cost of moving, cleaning, and syncing customer information across dozens of disconnected marketing clouds and

KDE Releases Plasma 6.7 with Per-Screen Virtual Desktops

The sheer complexity of contemporary digital workspaces often leads to a phenomenon where users feel overwhelmed by the literal lack of physical and virtual boundaries across their hardware. For years, the traditional approach to virtual desktops treated all connected displays as a singular, unified canvas, meaning that switching a workspace on one screen would force a transition on all others

Is the Fixed-Price AI Subscription Model Sustainable?

The rapid expansion of generative artificial intelligence has fundamentally transformed the digital landscape, yet the industry remains tethered to a subscription-based pricing model that may soon prove mathematically impossible to sustain. While the initial wave of adoption was fueled by the accessibility of flat-rate subscriptions, the underlying economics of massive compute clusters suggest a growing disconnect between user fees and

Will Agentic Automation Drive EMEA’s Autonomous Enterprise?

The transition from experimental artificial intelligence to deep-seated industrial application has reached a critical inflection point where simple task execution no longer suffices for the modern enterprise. As organizations across the Europe, Middle East, and Africa region navigate the complexities of a digital-first economy, the focus is pivoting toward Agentic Process Automation to bridge the gap between human intuition and