Supermicro Launches Advanced X14 Server Platforms

The new X14 servers aim accelerate AI and high-performance computing.

Supermicro has introduced its latest X14 server platforms, designed to deliver performance for compute-intensive workloads and applications. These servers are built on a newly redesigned architecture that leverages next-generation technologies to maximize computational efficiency and reduce costs in environments requiring high performance, such AI training, high-performance computing, and complex data analytics.

The new X14 systems are much more powerful than their predecessors, incorporating support for up to 256 performance cores (P-cores) in a single node. Combined with support for MRDIMMs at speeds up to 8800MT/s, this means a new level of memory performance, which is crucial for data-heavy tasks and large-scale AI models. Additionally, these systems are compatible with the latest GPU technologies, including next-generation SXM, OAM, and PCIe GPUs, allowing for substantial acceleration in AI and compute workloads.

The redesigned X14 servers also feature new 10U and multi-node form factors, which allow for higher GPU densities and increased CPU capabilities. The updated memory configuration includes 12 memory channels per CPU and MRDIMMs, delivering up to 37% better memory performance compared to traditional DDR5-6400 DIMMs. Enhanced storage interfaces support higher drive densities, and several systems now come with integrated liquid cooling solutions to improve efficiency and manage thermal loads more effectively.

Supermicro’s X14 family includes more than ten new systems across three primary categories, each tailored for specific workloads.

• GPU-optimized platforms are designed for applications that demand pure performance and enhanced thermal capacity, like large-scale AI training, large language models (LLMs), generative AI, 3D media, and virtualization.
• High compute-density multi-node servers, such as the SuperBlade and the new FlexTwin, utilize direct-to-chip liquid cooling to significantly increase the number of performance cores in a standard rack, optimizing for environments where space and power efficiency are critical.
• Hyper rackmount servers offer flexible I/O and storage configurations in traditional form factors, making them suitable for scaling up and out in evolving data center environments.

Overview of New Supermicro X14 Performance-Optimized Systems

The upcoming Supermicro X14 performance-optimized systems are tailored to specific high-performance use cases, including:

GPU-Optimized Systems: These platforms are engineered for maximum performance, and are specifically designed to handle demanding AI and HPC tasks. They support the highest-wattage GPUs, such as the latest-generation SXM5 and SXM6 GPUs, making them ideal for large-scale AI training, LLMs, generative AI, and other GPU-intensive workloads. These systems are available in both air-cooled and liquid-cooled configurations, allowing for flexibility depending on the operational environment and cooling requirements.

PCIe GPU Systems: Focused on providing maximum GPU flexibility, these systems can accommodate up to 10 double-width PCIe 5.0 accelerator cards in a thermally-optimized 5U chassis. This design is particularly suitable for workloads such as media rendering, collaborative design, simulation, cloud gaming, and virtualization, where high GPU count and flexibility are critical. The architecture allows these servers to maintain optimal performance under heavy GPU loads, making them versatile for various compute-heavy applications.

Intel Gaudi 3 AI Accelerator Systems: These are the first AI server based on the Intel Gaudi 3 accelerator to hit the market. Powered by Intel Xeon 6 processors, theses Accelerator Systems aim to enhance efficiency and reduce the cost of large-scale AI model training and inferencing. It features eight Intel Gaudi 3 accelerators on an OAM universal baseboard and integrates six OSFP ports for scalable networking. The open platform supports a community-based, open-source software stack, eliminating software licensing costs and providing flexibility for AI developers.

SuperBlade Systems: Featuring 6U high-performance, a density-optimized, energy-efficient build, the new SuperBlade Systems system maximize rack density with up to 100 servers and 200 GPUs per rack. This platform is optimized for AI, HPC, and other compute-intensive workloads. Each node can be configured with either air cooling or direct-to-chip liquid cooling, which maximizes efficiency and ensures the lowest Power Usage Effectiveness (PUE) with the best TCO. It also offers connectivity with up to four integrated Ethernet switches with 100G uplinks and front I/O, supporting networking options up to 400G InfiniBand or Ethernet per node.

FlexTwin Systems: The new FlexTwin architecture is built for maximum compute power and density in a multi-node configuration, supporting up to 24,576 performance cores in a 48U rack. Designed for HPC and other compute-intensive workloads, each node features direct-to-chip liquid cooling, which helps maximize efficiency and reduce CPU thermal throttling. The architecture also supports HPC Low Latency front and rear I/O with flexible networking options up to 400G per node, making it an excellent choice for dense computing environments where performance and cooling efficiency are paramount.

Hyper Rackmount Systems: The X14 Hyper is Supermicro’s flagship rackmount platform, focused on performance for demanding AI, HPC, and enterprise applications. These systems support single or dual socket configurations with the ability to include double-width PCIe GPUs, offering maximum workload acceleration capabilities. Available in both air-cooled and direct-to-chip liquid-cooled models, the Hyper systems are optimized to support top-tier CPUs without thermal limitations, reducing cooling costs for better overall data center efficiency.

Supermicro is also expanding its liquid cooling solutions to complement the X14 product line. These solutions include in-house developed cold plates for CPUs, GPUs, and memory, as well as comprehensive rack-level cooling systems designed to reduce thermal throttling and enhance energy efficiency.

Availability

All systems are now available for early access through Supermicro’s Early Ship Program.

Supermicro X14

Engage with StorageReview

Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | TikTok | RSS Feed