Supermicro AI Servers with AMD Instinct MI350 Series GPUs

Summary:

• Supermicro launches highly optimized AI solutions, based on AMD Instinct MI350 Series GPUs and AMD ROCm Software, delivering breakthrough inference performance and power efficiency.
• The H14 GPU solutions are powered by the 4^th Gen AMD CDNA architecture, delivering optimized performance and efficiency for large AI training models and high-speed inference workloads
• Large memory capacity with 2.304TB total HBM3e per 8-GPU server which delivers faster computation and more efficient scaling for AI, inferencing, and training.

Supermicro, Inc. is announcing that both liquid-cooled and air-cooled GPU solutions will be available with the new AMD Instinct MI350 series GPUs, optimized for unparalleled performance, maximum scalability, and efficiency.

The Supermicro H14 generation of GPU optimized solutions featuring dual AMD EPYC 9005 CPUs along with the Instinct MI350 series GPUs, are designed for organizations seeking maximum performance at scale, while reducing the total cost of ownership for their AI-driven data centers.

“Supermicro continues to lead the industry with the most experience in delivering high-performance systems designed for AI and HPC applications,” said Charles Liang, president and CEO, Supermicro. “Our Data Center Building Block Solutions® enable us to quickly deploy end-to-end data center solutions to market, bringing the latest technologies for the most demanding applications. The addition of the new AMD Instinct MI350 series GPUs to our GPU server lineup strengthens and expands our industry-leading AI solutions and gives customers greater choice and better performance as they design and build the next generation of data centers.“

The company’s H14 generation data center solutions offer unparalleled performance and flexibility for a wide range of AI, HPC, Cloud, and enterprise workloads. These systems are built on the firm’s building-block architecture to help enterprise customers efficiently upgrade and scale their workloads. The Supermicro GPU servers feature both the latest AMD EPYC CPU and Instinct GPUs which together offer a powerhouse family of AI-enabled solutions, adding to the company’s data center building block solutions (DCBBS) ensuring customer success from consultation to on-site deployment and support.

“Our new AMD Instinct MI350 series delivers up to 40 percent more tokens-per-dollar relative to the competition, while retaining the industry standard form factor for OEM infrastructure compatibility,” said Dr. Lisa Su, CEO and chair, AMD (Advanced Micro Devices, Inc.). “By combining these GPUs with Supermicro’s proven platforms, their customers can deploy fully integrated, air- or liquid-cooled racks built entirely on AMD technology, giving them the flexibility and efficiency they need to deploy robust AI solutions at any scale.“

Supermicro is expanding its proven liquid-cooled and air-cooled high-performance fabric 8-GPU system, leveraging the open multi-architecture industry standard OCP Accelerator Module (OAM) to support the latest Instinct MI350 series GPUs. For higher-density environments, the 4U liquid-cooled system with Instinct MI355X GPUs features the company’s new improved Direct Liquid Cooling (DLC) architecture incorporating new technologies for cooling various server components and reducing power consumption by up to 40%, unlocking higher performance per rack and supporting advanced cooling infrastructure at scale. Customers can choose the firm’s 4U liquid-cooled option for higher- density rack-scale deployments as well as the 8U option for air-cooled environments.

Designed to maximize computational throughput, memory bandwidth utilization, and energy efficiency to enable more power-efficient AI inference. these accelerated GPU servers with Instinct MI350 series will offer 288GB HBM3e/GPU, boosting 1.5x memory capacity compared to previous-gen of Instinct accelerator, 8TB/s bandwidth, and 1.8x petaflops of FP16 / FP8 compared to the previous-gen Instinct MI325X, so customers can process more data faster for their AI workloads.

“AI models aren’t just increasing in size; they’re demanding faster, more efficient infrastructure that can stay consistently deployable in real-world environments,” said Paul Schell, industry analyst, ABI Research. “By adding AMD MI350 series GPUs into its latest H14 generation, Supermicro demonstrates a commitment to providing scalable, high-performance and memory solutions that are optimized for both AI training and inference. With support for both liquid and air cooling, these systems offer the flexibility and efficiency that CSPs, neo-clouds and enterprises alike need to power the next wave of AI growth.”

These new GPU solutions are designed for powering AI at scale, across cloud service providers and enterprise partners, built on the 4th Gen AMD CDNA architecture, the Instinct MI350 Series delivers exceptional performance and energy efficiency for a wide range of workloads, from AI training and inference to complex scientific simulations. The AMD MI350 series will also feature new FP6 and FP4 data types to provide exceptional AI capabilities, handling larger models for AI deployments.

These Supermicro GPU systems supporting the AMD Instinct GPUs are available to order, including 4U Liquid-Cooled GPU, and 8U Air-Cooled GPU.

Resource:
Supermicro Servers with AMD Instinct MI350 Series GPUs