Supermicro, Inc., a full-service provider of IT solutions in AI/ML, high-performance computing, cloud, storage, and 5G/ edge, today announced the launch of a series of workload-optimized GPU servers and workstations. Supports the new NVIDIA RTX PROTM 6000 Blackwell Server Edition GPU. Supermicro has a broad line of servers optimized for NVIDIA Blackwell PCIe Gpus, enabling more enterprises to support LLM inference and fine-tuning, agent AI, visualization, graphics and rendering, and virtualization applications with accelerated computing. Supermicro's multiple GPU optimization systems are certified by NVIDIA to ensure support and compatibility with NVIDIA AI Enterprise, which simplifies the development and deployment of AI for manufacturing production.

Supermicro's President and CEO, Kanhou Liang, said, "Supermicro leads the industry with its multi-application optimized GPU server portfolio and its ability to deploy a wide range of enterprise environments with extremely short lead times. Our support for NVIDIA RTX PRO 6000 Blackwell Server Gpus provides a new level of performance and flexibility to enable customers to deploy the latest accelerated computing technologies, from the data center to intelligent edge applications. "Supermicro's extensive line of PCIe GPU optimization products also supports 2-way and 4-way NVIDIA NVLink™ configurations of NVIDIA H200 NVL Gpus, maximizing the inference performance of today's most advanced AI models and accelerating high-performance computing workloads."

The NVIDIA RTX PRO 6000 Blackwell Server Edition is a general-purpose GPU optimized for AI and graphics workload performance. Compared to the previous generation of NVIDIA L40S, this GPU has significantly enhanced performance, including faster GDDR7 memory and twice the memory capacity, and a PCIe 5.0 interface to enable faster communication between the GPU and the CPU. The NVIDIA RTX PRO 6000 Blackwell Server Edition also features new multi-instance GPU (MIG) performance, which allows a single GPU resource to be shared among up to four fully isolated instances. In addition, Supermicro's GPU-optimized systems support NVIDIA Supernics, such as NVIDIA BlueField®-3 and NVIDIA ConnectX®-8. NVIDIA Quantum InfiniBand and NVIDIA Spectrum Ethernet enable optimal infrastructure scaling and GPU clustering performance.

Bob Pette, vice president of Enterprise Platforms at NVIDIA, said: "The NVIDIA RTX PRO 6000 Blackwell Server Edition is the ultimate data center GPU built for AI and visual computing, delivering unprecedented acceleration for the toughest workloads. The NVIDIA RTX PRO 6000 Blackwell Server Edition further expands Supermicro's rich lineup of NVIDIA accelerated systems to accelerate almost all types of AI development and inference workloads."

In addition to Enterprise NVIDIA RTX PRO 6000 Blackwell Server Edition, Selected Supermicro workstations will also support the new NVIDIA RTX PRO 6000 Blackwell Workstation Edition and NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition, These most powerful professional-grade Gpus are suitable for workloads such as AI processing and development, 3D rendering, and media and content creation.

The Supermicro system product family supporting the new GPU is as follows:

NVIDIA RTX PRO 6000 Blackwell Server Edition

5U PCIe GPU - Highly flexible and thermally optimized architecture that supports up to 10 Gpus in a single air-cooled chassis. This series of systems features a dual CPU and PCIe 5.0 expansion for high-speed connections. Applicable workloads include AI inference and fine-tuning, 3D rendering, digital avatar, scientific simulation, and cloud gaming.

NVIDIA MGX™ - This GPU optimization system is based on NVIDIA's modular reference design and can accommodate up to 4 Gpus in a 2U model or 8 Gpus in a 4U model for applications such as industrial automation, scientific modeling, high performance computing, and AI inference.

3U Edge-Optimized PCIe Gpus - Compact body for edge data center deployments, supporting up to 8 double-width Gpus or 19 single-width Gpus per system. Applicable workloads include electronic design automation (EDA), scientific modeling, and edge AI reasoning.

SuperBlade® - A density-optimized, energy-efficient multi-node architecture for maximum rack density and up to 120 Gpus per rack.

Rack-mounted workstations - Provide workstation performance and flexibility in rack-mounted models with high density and high security to meet the organization's need for centralized resource utilization.

NVIDIA RTX PRO 6000 Blackwell Workstation Edition and NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition

Vertical Workstations - a range of desktop and rack-mounted models designed for AI, 3D media and simulation applications, ideal for AI developers, creative studios, educational institutions, field offices and laboratories.

Support for existing Gpus in other markets such as H200/H100 NVL, L40S, L4:

4U GPU Optimization - Supports up to 10 double-width Gpus and can be configured in Single Root and Dual Root configurations. Vertical GPU servers can support up to four double-width Gpus.

1U and 2U MGX™ - This compact GPU optimization system is based on an NVIDIA modular reference design and supports up to four double-width Gpus.

1U and 2U rack platforms - including flagship performance Hyper and Hyper-E, as well as CloudDC optimized for cloud data centers, support up to four double-width Gpus or eight single-width Gpus.

Multi-processor - Features 4 - and 8-slot architectures to maximize memory and I/O density, and can support up to 2 double-width Gpus in a 2U case and up to 12 double-width Gpus in a 6U case.

Edge System - A compact Box PC that can support 1 double-width GPU or 2 single-width Gpus.