Server power supplies (PSUs) in a rack must balance high wattage, efficiency, redundancy, and scalability to reliably support modern data center workloads.

Key Considerations for Server PSUs

Wattage and Power Budget: Modern servers, especially those with CPUs and GPUs for AI or HPC workloads, can demand 800 W to over 3,000 W per server. Rack PSUs are often limited by AC inlet ratings, typically around 16–20 A at 240 VAC, which translates to 3,000–3,600 W per PSU. Using busbars or higher-rated inlets can increase power density and allow future expansion . Efficiency: High-efficiency PSUs reduce energy loss and heat generation. Look for 80 PLUS certification or higher, which ensures minimal energy wastage. Advanced designs using SiC MOSFETs or GaN HEMTs can further improve efficiency, thermal performance, and switching speed, enabling compact, high-density server power solutions . Redundancy and Reliability: Redundant PSUs are critical in server racks to maintain uptime. Dual or N+1 configurations allow one PSU to fail without affecting server operation, which is essential for enterprise and cloud environments . Form Factor and Integration: PSUs must fit the server chassis (1U, 2U, or blade form factors) and support modular cabling to improve airflow and reduce clutter. Reference designs, such as 1U 1.6 kW AC-DC PSUs, demonstrate how high efficiency and compact size can coexist . Digital Monitoring and Control: Modern PSUs often include digital monitoring for voltage, current, and temperature, enabling proactive management and integration with data center infrastructure management (DCIM) systems .

Advanced Technologies

• Wide-Bandgap Devices: Silicon Carbide (SiC) and Gallium Nitride (GaN) devices allow higher switching frequencies, lower losses, and smaller component sizes, improving PSU efficiency and power density .
• High-Voltage DC Architectures: Some data centers adopt 800 V HVDC PSUs to reduce conversion losses and improve energy efficiency for AI and HPC workloads .
• Phase-Shifted Full-Bridge Converters: Used in high-efficiency designs to minimize losses and support high output power in compact form factors .

Planning for the Row

When designing power for a row of servers:

1. Calculate Total Load: Include CPUs, GPUs, storage, and networking equipment. Use tools like data center power sizing calculators to model scenarios .
2. Allow Headroom: Include extra capacity for future upgrades and peak loads.
3. Implement Redundancy: Ensure N+1 or 2N PSU configurations to maintain uptime.
4. Optimize Cooling: Efficient PSUs reduce heat, lowering cooling requirements and improving overall PUE (Power Usage Effectiveness).
5. Monitor and Manage: Use digital monitoring to track PSU performance and detect anomalies early. By carefully selecting PSUs with adequate wattage, high efficiency, redundancy, and modern semiconductor technologies, a server row can achieve reliable, scalable, and energy-efficient operation suitable for AI, cloud, and high-performance computing workloads .