Supermicro Direct to Chip Liquid Cooling Solutions

Horizontal CDM

Cold Plates

The cold plates are placed on top of the CPUs, GPUs, and DIMMs and feature micro-sized channels through which coolant flows to cool down the chips with high efficiency. Supermicro cold plates are designed to reduce hot spots on the chip and achieve ultra-low thermal resistance.

GPU-Optimized

Supermicro GPU-optimized systems provide maximum acceleration for large-scale AI training, large language models, and generative AI applications. The GPU-optimized systems have been completely re-engineered to take advantage of the latest interconnect, memory, storage, and cooling technologies to ensure significant performance increases over prior generations. The modular design includes a dedicated GPU tray which houses an SXM5 (H100/H200) or SXM6 (B200 8-GPU) baseboard and is easily accessible from the cold aisle, simplifying installation and servicing. For maximum GPU density, a liquid-cooled 4U architecture can be integrated using Supermicro’s complete direct-to-chip liquid cooling solution, allowing up to 8 systems in a standard 48U rack for a total of 64 GPUs.

BigTwin®

Supermicro BigTwin systems provide maximum performance and serviceability in a multi-node architecture, with dual Intel® Xeon® 6700 or 6500 series processors per node and a hot-swappable tool-less design. Optimized for density (2U/4N) or storage (2U/2N), BigTwin systems with shared components can be up to 8% more power-efficient than standard 1U servers, with Supermicro’s Resource Saving Architecture of shared power and cooling reducing TCO and TEC. All BigTwin systems can be air-cooled, with liquid cooling options available to not only further reduce power consumption and noise levels, but also allow maximum compute density of up to eight 350W TDP CPUs in a 2U chassis.

FlexTwin™

Supermicro FlexTwin is a new 2U 4-node platform designed for maximum performance density and serviceability in a multi-node architecture, featuring support for the latest CPU, memory, storage, and cooling technologies. Purpose-built to support demanding HPC workloads including financial services, manufacturing, scientific research, and complex modeling, FlexTwin can be customized to suit specific HPC applications and customer requirements thanks to Supermicro’s modular Building Block architecture. Each hot-swappable FlexTwin node features direct-to-chip liquid cooling technology which can remove up to 90% of server-generated heat, reducing data center cooling costs compared to traditional air cooling and ensuring maximum compute performance by preventing thermal throttling under heavy load.

SuperBlade®

Supermicro’s GPU-enabled high-performance, density-optimized, and energy-efficient SuperBlade can significantly reduce initial capital and operational expenses for many organizations. SuperBlade utilizes shared, redundant components—including cooling, networking, power, and chassis management—to deliver the compute performance of an entire server rack in a much smaller physical footprint. For even greater compute densities, dual-processor and single-width blade configurations can support 350W TDPs with optional direct-to-chip liquid cooling. Liquid cooling in data center racks complements traditional air cooling to run at higher performance levels, while reducing the need for CRAC units and improving overall efficiency, lowering OPEX, TCO, and TCE.

Hyper

The Hyper series brings flagship performance to Supermicro’s rackmount servers, built to tackle the most demanding workloads along with storage & I/O flexibility for a wide range of applications. With up to 8 PCIe 3.0 expansion slots in a 2U chassis accommodating up to four double-width GPUs, enterprises can add GPUs, DPUs, and other accelerators to boost performance. With hot-swap direct-to-chip liquid cooling, Hyper supports dual processors up to 500W TDP and 12 hot-swappable NVMe drive bays.

Supermicro’s Rack Integration Services, Turnkey Cluster Level Liquid Solutions

Supermicro’s Rack Integration Services leverage application-optimized motherboards, chassis, cooling subsystems, networking components, cluster management tools, energy-efficient power supply technologies, and compact enclosures to design and develop customized and enterprise solutions. Supermicro understands the requirements of today’s fast-paced business environments and customer requirements. We offer an end-to-end integration service that helps customers reduce overhead, maximize efficiency and quality, and gain a competitive advantage through reduced time-to-results.

Supermicro works with leading organizations in all geographies to design, install, and test various liquid cooling solutions. The Supermicro process involves a rigorous set of phases that ensure the most optimized and tested solutions for environments where liquid cooling is required for maximum performance.