Supermicro, Inc. serves as a full-service IT provider for AI/ML, HPC, cloud, storage and 5G/edge. The solutions provider has added new systems with the highest performance GPUs, multi-node configurations, and rack-mountable designs to its Rapids-AP). A new, industry-leading family of workload-optimized servers designed to meet the needs of modern data centers, enterprises and service providers. Following the launch of the efficiency-optimized X14 servers equipped with Xeon 6700 series processors (using efficiency cores) in June 2024, the additional system models launched enable the Supermicro X14 series to have the highest computing density and computing power, providing The industry’s broadest range of optimized servers, supporting everything from high-demand AI, high-performance computing (HPC), media and virtualization to energy-efficient edge applications, scale-out cloud-native and microservices applications, and more Diverse workloads.
Supermicro President and CEO Liang Jianhou said: “The Supermicro X14 system has been completely redesigned to support the latest technologies, including next-generation CPUs, GPUs, MRDIMMs with high bandwidth and lowest latency, PCIe 5.0, and EDSFF E1.S and E3. S storage. Not only do we now offer more than 15 system product lines, we can also use these designs to create custom solutions, paired with our complete rack integration services and in-house developed liquid cooling solutions.”
Approved customers can obtain complete, deployable systems early through Supermicro’s Early Ship program or perform remote testing through Supermicro JumpStart .
These new Supermicro X14 systems feature a completely redesigned architecture, including new 10U and multi-node form factors that support next-generation GPUs and higher CPU core densities, as well as upgraded versions with 12 memory channels per CPU Memory slot configuration, and new MRDIMM with 37% higher memory bandwidth than DDR5 DIMM.
The new Supermicro X14 family includes several new systems, several with new architectures and divided into three categories for specific workloads:
- A GPU-optimized platform designed for pure performance and enhanced cooling to support the latest technology and highest wattage GPUs. The system architecture is built from the ground up from the most basic level and is suitable for large-scale AI training, large language models (LLM), generative AI, 3D media and virtualization applications.
- High-compute-density multi-node models, including the new FlexTwin™, SuperBlade® and GrandTwin®, can increase efficiency by sharing resources such as power and cooling. Select models use direct-to-wafer liquid cooling technology to maximize density without compromising performance.
- The market-proven Supermicro Hyper rack-mount platform combines single or dual-socket architecture, flexible I/O and traditional form-factor storage configurations to help enterprises and data centers scale vertically as workloads evolve. and horizontal expansion.
Supermicro’s new highest-performance X14 systems support the new Intel Xeon 6900 series processors with Performance Cores. In addition, X14 efficiency-optimized systems, launching in June 2024, support Intel Xeon 6700 series processors with Efficiency Cores. These systems provide sockets for the Intel Xeon 6900 series processors with Performance Cores, due in the first quarter of 2025, and the Intel Xeon 6700 series processors with Efficiency Cores, providing additional flexibility to enable Systems can be optimized in terms of performance per core or performance per watt.
Ryan Tabrah, vice president and general manager of Intel With the launch of the new series, Supermicro will be able to introduce these solutions that are suitable for AI. and new performance-optimized CPUs for compute-intensive workloads, providing customers with more choices.”
When configured with Intel Xeon 6900 series processors equipped with performance cores, Supermicro systems can support new FP16 instructions on the built-in Intel® AMX accelerator to further enhance AI workload performance. Each CPU in these systems is equipped with 12 sets of memory channels, supporting DDR5-6400 and MRDIMM up to 8800MT/s, and CXL 2.0. It also supports high-density, industry-standard EDSFF E1.S and E3.S NVMe hardware. Disc provides wider support.
Supermicro Liquid Cooling Solutions
Supermicro continues to complement its expanded X14 product portfolio with rack-level integration and liquid cooling capabilities. At the same time, Supermicro leverages its industry-leading global manufacturing capabilities, extensive rack-level integration and test facilities, and a comprehensive set of management software solutions to design, build, test, validate and deliver in just weeks Complete data center solution for any size.
Supermicro also offers complete liquid cooling solutions developed in-house, including heat sink plates for CPUs, GPUs and memory, as well as cooling distribution units, cooling distribution manifolds, hoses, connectors and cooling water towers. Liquid cooling technology can be easily incorporated into rack-level integration to further improve system efficiency, reduce the occurrence of overheating and throttling, and reduce the total cost of ownership (TCO) and total cost of the environment (TCE) of data center deployments.
Also Read: Vultr Enhances AI Cloud Inference with AMD MI300X
New Supermicro’s highest-performance X14 systems support Intel Xeon 6900 series processors with performance cores . These systems include:
GPU Optimized – The highest-performance Supermicro X14 system is designed for large-scale AI training, large language models, generative AI and HPC, supporting eight latest generation SXM5 and SXM6 GPUs. These systems are available with air or liquid cooling technology.
PCIe GPU – Designed for maximum GPU flexibility, its thermally optimized 5U chassis supports up to 10 double-wide PCIe 5.0 accelerator cards. These servers are ideal for AI inference, media, collaborative design, simulation, cloud gaming and virtualization workloads.
Intel® Gaudi® 3 AI accelerator – Supermicro also plans to launch the industry’s first AI server equipped with Intel Gaudi 3 accelerator and Intel Xeon 6 processor. This system is expected to improve the efficiency of large-scale AI model training and AI inference and reduce costs. The system features eight Intel Gaudi 3 accelerators mounted on an OAM universal baseboard and six integrated OSFP ports to enable cost-effective, scale-out networking and includes an open platform that supports Community-based open source software stack without subsequent software licensing costs.
SuperBlade® – The 6U high-performance, density-optimized and energy-efficient SuperBlade in the Supermicro X14 series maximizes rack density, enabling up to 100 servers and 200 GPUs per rack. Each node is optimized for AI, HPC, and other compute-intensive workloads and features air or direct-to-die liquid cooling to maximize efficiency and achieve the lowest power usage effectiveness (PUE) and best overall Cost of Ownership (TCO). At the same time, each node also has up to four integrated Ethernet switches, supporting 100G uplinks and front-end I/O, providing flexible choices of up to 400G InfiniBand or 400G Ethernet.
FlexTwin™ – The new Supermicro X14 FlexTwin architecture is designed specifically for HPC, delivering high energy efficiency and delivering maximum computing power and density in multi-node configurations, while enabling a 48U rack to host up to 24,576 performance cores. Each node is optimized for HPC and other compute-intensive workloads, and uses direct-to-chip liquid cooling technology to maximize efficiency, reduce the occurrence of CPU overheating and frequency reduction, and has HPC low-latency front- and rear-end I/O O, supports a series of flexible network options up to 400G.
Hyper – X14 Hyper is Supermicro’s flagship rack-mount platform designed to deliver maximum performance for high-demand AI, HPC and enterprise applications, and supports dual-width PCIe GPUs in single or dual-slot configurations for maximum workloads Load acceleration. Available in air-cooled and direct-to-wafer liquid-cooled models, this platform can support top-end CPUs without thermal constraints, thereby reducing data center cooling costs and improving efficiency.
In addition, the Supermicro X14 efficiency-optimized system now available is powered by Intel Xeon 6700 series processors with efficiency cores. These systems include:
SuperBlade® – Supermicro’s high-performance, density-optimized and energy-efficient multi-node platform optimized for AI, data analytics, HPC, cloud and enterprise workloads. Model specifications include a 6U chassis with 10 or 5 nodes, or an 8U chassis with 20 or 10 nodes, allowing each rack to be configured with up to 34,560 Xeon computing cores.
Hyper – flagship high-performance rack-mounted server designed for horizontal expansion of cloud workloads. It has high storage and I/O flexibility to provide customized models to meet different application needs.
CloudDC – an all-in-one platform for cloud data centers, using OCP Data Center Modular Hardware System (DC-MHS), with flexible I/O and storage configurations and dual AIOM slots (supports PCIe 5.0; complies with OCP 3.0 Standard) to achieve maximum data transfer volume.
Petascale Storage – Achieve industry-leading All-Flash storage density and performance through EDSFF E1.S and E3.S hard drives, providing unprecedented capacity and performance in 1U or 2U models.
WIO – Cost-effective architecture with flexible I/O configurations that optimize acceleration, storage and networking alternatives to accelerate performance, increase efficiency and perfectly match specific enterprise applications.
BigTwin® – 2U 2-node or 2U 4-node platform delivers superior density, performance and serviceability through dual processors per node and hot-swappable tool-less design. New models of these systems are ideally suited for cloud, storage and media workloads and include support for E3.S drives for superior density and throughput.
GrandTwin® – Designed for single-processor performance and memory density, with front-facing (cold aisle) hot-swappable nodes and front or rear I/O for easy maintenance. The current model also supports E1.S hard drives, which have better storage density and transfer capacity.
Hyper-E – Delivers the power and flexibility of our flagship Hyper series, optimized for deployment in edge application environments. Features of edge-focused models include a short chassis and front-mounted I/O, making Hyper-E suitable for edge data centers and telecom cabinets. These short-body systems can support up to three sets of high-performance GPU or FPGA accelerator cards.
Edge/Telco – Offers high-density processing performance and compact form factor, optimized for installation in telecom cabinets and smart data centers. These systems are available with optional DC power configurations and optimized operating temperatures up to 55°C (131°F).
Source: PRNewswire