DLC improvements

Supermicro is announcing several improvements to its Direct Liquid Cooling (DLC) solution that incorporate new technologies for cooling various server components, accommodate warmer liquid inflow temperatures, and introduce innovative mechanical designs that enhance AI per watt.

The Supermicro DLC-2 solution reduces data centre power consumption by up to 40% compared to air-cooled installations. These advanced technologies enable faster deployment and reduced time-to-online for cutting-edge liquid-cooled AI infrastructure. Additionally, the total cost of ownership decreases by up to 20%. The comprehensive cold plate coverage of components allows for lower fan speeds and fewer required fans, significantly reducing data centre noise levels to approximately 50dB.

"With the expected demand for liquid-cooled data centres rising to 30% of all installations, we realised that current technologies were insufficient to cool these new AI-optimised systems," said Charles Liang, president and CEO of Supermicro. "Supermicro continues to remain committed to innovation, green computing, and improving the future of AI, by significantly reducing data centre power and water consumption, noise, and space. Our latest liquid-cooling innovation, DLC-2, saves data centre electricity costs by up to 40%."

Supermicro aims to save 20% of data centre costs and apply DLC-2 innovations as part of data centre building block solutions to make liquid-cooling more broadly available and accessible.

A significant component of the new liquid-cooling architecture is a GPU-optimised Supermicro server, which includes eight NVIDIA Blackwell GPUs and two Intel Xeon 6 CPUs, all in just 4U of rack height. This system is designed to support increased supply coolant temperatures. This unique and optimised design incorporates cold plates for CPUs, GPUs, memory, PCIe switches, and voltage regulators. This design reduces the need for high-speed fans and rear-door heat exchangers, thereby lowering cooling costs for the data centre.

The new Supermicro DLC-2 solution stack supports the new 4U front I/O NVIDIA HGX™ B200 8-GPU system, and the in-rack Coolant Distribution Unit (CDU) has an increased capacity of removing 250kW of heat generated per rack. The Supermicro DLC-2 solution also utilises vertical coolant distribution manifolds (CDMs) to remove hot liquid and return cooler liquid to the servers for the entire rack. The reduced rack space requirements enables more servers to be installed, increasing computing density per unit of floor space. The vertical CDM is available in various sizes, precisely matching the number of servers installed in the rack. The entire DLC-2 solution stack is fully integrated with Supermicro SuperCloud Composer software for data centre-level management and infrastructure orchestration.

The efficient liquid circulation and nearly full liquid-cooling heat capture coverage, at up to 98% per server rack, allow for an increase in the inlet liquid temperature at up to 45°C. The higher inlet temperature eliminates the need for chilled water, chiller compressor equipment cost, and additional power usage, saving up to 40% of data centre water consumption.

Combined with liquid-cooled server racks and clusters, DLC-2 also offers hybrid cooling towers as well as water towers as part of data centre building blocks. The hybrid cooling towers combine the features of standard dry and water towers into a single design. This is especially beneficial in data centre locations with strong seasonal temperature variation to reduce usage of resources and costs further.

Supermicro serves as a comprehensive one-stop solution provider with global manufacturing scale, delivering data centre-level solution design, liquid-cooling technologies, networking, cabling, a full data centre management software suite, L11 and L12 solution validation, onsite deployment, and professional service and support. With production facilities across San Jose, Europe, and Asia, Supermicro offers unmatched manufacturing capacity for liquid-cooled rack systems. This ensures timely delivery, reduced total cost of ownership (TCO), and consistent quality.

Data Centre Building Block Solutions (DCBBS) simplify and shorten buildouts

Supermicro is also announcing Data Centre Building Block Solutions, the easiest solution to overcome the immense complexities of outfitting liquid cooled AI factories with all critical infrastructure components, including servers, storage, networking, rack, liquid cooling, software, services, and support. As an expansion of Supermicro's System Building Block Solutions, DCBBS adopts a standardised, yet flexible solution architecture, vastly expanded in scope to handle the most demanding AI data centre training and inference workloads, enabling easier data centre planning, buildout, and operation – all while reducing cost.

"Supermicro's DCBBS enables clients to easily construct data centre infrastructure with the fastest time-to-market and time-to-online advantage, deploying as quickly as three months," said Charles Liang, president and CEO of Supermicro. "With our total solution coverage, including designing data centre layouts and network topologies, power and battery backup-units, DCBBS simplifies and accelerates AI data centre buildouts leading to reduced costs and improved quality."

DCBBS offers packages of pre-validated data centre-level scalable units, including a 256-node AI Factory DCBBS scalable unit, designed to alleviate the burden of prolonged data centre design by providing a streamlined package of floor plans, rack elevations, bill of materials, and more. Supermicro provides comprehensive first-party services to ensure project success, starting from consultation to on-site deployment and continued on-site support. DCBBS is customisable at the system-level, rack cluster-level, and data centre-level to meet virtually any project requirements.

Along with our DLC-2 technology, DCBBS also helps customers save up to 40% power, reducing 60% data centre footprint, and decreasing 40% water consumption, all of which leads to 20% lower TCO.

AI Factory Data Centre-Level Scalable Unit

The need for AI infrastructure continues to scale: AI training clusters require clusters of thousands of GPUs to develop foundation models. AI inference applications are also leveraging more test-time compute capacity by running multiple inference passes with a mixture of models to deliver greater intelligence. Supermicro's AI Factory DCBBS package fully equips data centres to tackle these rising AI computational requirements.

Solutions from Supermicro include up to 256 Liquid Cooled 4U Supermicro NVIDIA HGX™ system nodes, each system equipped with 8 NVIDIA Blackwell GPUs (2,048 GPUs in total), interconnected with up to 800Gb/s NVIDIA Quantum-X800 InfiniBand or NVIDIA Spectrum™ X Ethernet networking platform. The compute fabric is supported by elastically scalable tiered storage with high-performance PCIe Gen5 NVMe, TCO optimised Data Lake nodes, and resilient management system nodes for continuous uninterrupted operation.

System-Level, Rack-Level, and Data Centre-Level Customisation

Supermicro features a modular building block approach, composed of three hierarchical levels: the system-level, rack-level, and data centre level, giving customers unparalleled design options in determining a system-level bill of materials, down to selecting individual components, including CPUs, GPUs, DIMMs, drives, and NICs. System-level customisation ensures the ability to meet specialised hardware requirements for a particular data centre workloads and applications and allows for granular fine-tuning of data centre resources.

Supermicro aids in designing rack enclosure elevation layouts to ensure optimisation for thermals and cabling, giving customers the ability to select the type of rack enclosure, including 42U, 48U, and 52U configurations.

After the initial consultation with the customer, Supermicro delivers a project proposal tailored to a given data centre power budget, performance target, or other requirements.

Supermicro DLC-2

With liquid-cooled data centres growing from less than 1% of the market to an expected 30% within a year, Supermicro is driving the industry-wide adoption of DLC by helping customers achieve the challenge of needing to build new liquid-cooled data centres that can more efficiently remove heat.

DLC provides unmatched efficiency by capturing heat directly from the individual chips, including AI GPUs running at 1000W TDP and beyond. Liquid cooling infrastructure is planned and deployed at data centre scale, including the piping and facility-side liquid cooling tower for heat dissipation. Supermicro leads the industry in providing a total solution for direct-to-chip liquid cooling infrastructure, consisting of DLC systems, in-rack or in-row coolant distribution units, coolant distribution manifolds, cooling towards, and more.

Vertiv has been chosen by Polar as the primary supplier for its first modular AI-ready data centre...
Expanding data centre critical power product manufacturing capacity with new site: Bielsko-Biała,...
Shell Lubricants’ immersion cooling fluids have become the first to receive official...
EfficiencyIT has been awarded a Royal Warrant of Appointment by His Majesty King Charles III,...
The 260,000+ square-foot facility will blend manufacturing with modern offices, creating 215 new...
Schneider Electric designated a Vendor Champion in Canalys’ new report for channel management...
The Data Centre Energy Storage Market is set to grow from its current market value of more than $2...
Launches new transfer switch capabilities to keep critical equipment and operations up and running...