A system for AI Data Centers to thrive in

The rapidly growing AI technology is pushing the limits of every Data Center. Demanding more storage, better cooling, faster compute times, networking and services. Supermicro has created what they call Data Center Building Block Solutions, DCBBS which simplifies all of that without sacrificing speed, compute power, uptime and flexibility. Well, let's dive in and take a closer look.

Supermicro’s innovative and cutting-edge solutions for complex AI Data Center approaches the complicated AI infrastructure by using a modular building block approach. Think component-level, system-level and rack-level, where you begin with a system and build your way up to a rack and beyond. This creates a serviceable and scalable system, that also offers direct liquid cooling (DLC) as a total package solution.

System level Building Blocks

Supermicro have long been at the top of the market, using Building Block architecture to provide a ton of optimization in favor for your own project requirements. It is important to fine tune the system bill of materials (BOM) because it establishes the balance of computing resources for the entire data center. Supermicro's DCBBS starts here at the system level and gives unbeatable customization options, with the freedom to choose individual sub-components.

• Fine tune every part of your resources down to CPUs, GPUs, DIMMs, drives, NICs, and more.
• Optimal layout thanks to multiple system design
• First to try cutting-edge technology
• Advanced liquid-cooling with cost savings
• Scalable and specialized systems for AI compute
• Reduces supply chain challenges

Rack and Cluster Building Blocks

After systems are defined, it's time to integrate them into rack-level building blocks, the very backbone of your cluster. The cable routing is optimized for reduced airflow obstruction. Need to service one of the servers? You can easily access via ports in the front, away from hot air. Scalable units enable 256-node clusters and beyond. Optimal networking configuration for a non-blocking network topology for fast node-to-node communication.

The block part of this system ensures a simpler way of building large AI clusters, such as 256 system node clusters. These "scalable units" consist of groups of systems, interconnected with rail-optimized network topology, that can be further multiplied to achieve the desired cluster size.

DLC-2 - Next-gen Liquid Cooling

A massive data center is bound to produce a good amount of heat. To support each individual hardware during intense workloads cooling is a must. Supermicro's Direct-to-Chip Liquid Cooling (DLC) cold plates keeps the temperature down when it gets hot. Supermicro provides a total solution for DLC infrastructure and is an important part of the DCBBS. Supermicro claims the new DLC-2 will save you up to 40% in power consumption compared to air cooling. It will also save you up to 40% in water savings.

To understand their customers, Supermicro has replicates the environment during manufacturing process. This way, the infrastructure as a whole unit is well tested and validated. Supermicro's global services and support offers a full portfolio of service level building blocks such as datacenter design, solution validation and professional onsite deployment.

If you'd like to read more, then click this link to learn about Supermicro's DLC-2 Direct Liquid-Cooling Solution.

DLC-2 Key Features

• Direct-to-chip liquid cooling
• Cold plates for CPUs, GPUs, DIMMs, drives, NICs, and more.
• Up to 40% power savings
• Up to 40% water savings
• Up to 98% System Heat Capture
• Quiet Data Center with ~50dB
• Up to 60% space savings due to 2.5x compute density compared to air-cooled systems