Expanding HPC Possibilities with Supermicro and NVIDIA Vera CPU Solutions
From versatile air-cooled systems to high-density liquid-cooled racks, Supermicro expands its HPC portfolio with NVIDIA Vera CPU solutions designed to address the evolving demands of scientific computing and AI.
As the high-performance computing (HPC) community gathers for SC26, the industry's premier conference for supercomputing, networking, storage, and analysis, the demands placed on modern computing infrastructure continue to evolve. From scientific simulations and computational fluid dynamics to molecular modeling and large-scale data analytics, researchers and enterprises increasingly require computing platforms that deliver high performance, memory bandwidth, and energy efficiency at scale.
While GPUs have transformed accelerated computing, CPUs remain a critical foundation of HPC infrastructure. Beyond supporting accelerator-based workloads, CPUs continue to play a central role in scientific simulations, data processing, workload orchestration, and applications that rely on complex execution patterns and efficient memory access.
Building on its established portfolio of HPC-optimized solutions, including the highly successful SuperBlade® and Twin product lines, Supermicro continues to expand its offerings to address the industry’s increasingly diverse computing requirements. With the introduction of air-cooled and liquid-cooled NVIDIA Vera CPU server solutions, which are both being showcased at Supermicro’s booth at SC26, Supermicro brings additional architectural choices to its HPC portfolio, complementing its existing platforms.
NVIDIA Vera: Expanding Beyond Agentic AI into HPC
While HPC and agentic AI serve very distinct computational objectives, both place significant demands on the underlying CPU infrastructure. From large-scale scientific simulations and ensemble modeling to concurrent AI agents and iterative execution workflows, all these applications benefit from high core counts, strong per-thread performance, substantial memory bandwidth, and the ability to move data around efficiently. As such, NVIDIA Vera, although designed to address the emerging CPU role in agentic AI workloads, introduces several architectural capabilities that render it highly relevant to traditional HPC applications.
Each NVIDIA Vera CPU features:
- 88 NVIDIA Olympus cores and 176 threads, supporting highly parallel computational workloads.
- Up to 1.2 TB/s of memory bandwidth through SOCAMM LPDDR5X memory, providing substantial bandwidth for memory-intensive applications.
- NVIDIA Spatial Multithreading, designed to improve workload isolation and predictable execution through partitioned core resources.
- Second-generation NVIDIA Scalable Coherency Fabric, enabling communication between CPU cores, cache, memory, and I/O.
For example, scientific applications involving sparse computations, irregular memory access, and large-scale data processing can be constrained by the rate at which data moves between memory and CPU cores. Vera's high-bandwidth memory architecture combined with a monolithic die for consistent memory access provides an opportunity to address these bottlenecks, while its CPU architecture supports applications involving complex control flows and computationally intensive operations.
Scaling HPC Infrastructure with the Supermicro Liquid-Cooled NVIDIA Vera CPU Rack
For organizations operating at the scale of large HPC clusters and supercomputing facilities, compute density, energy efficiency, and cooling infrastructure become increasingly important considerations. At this scale, liquid-cooled solutions enable higher-density configurations while providing a more direct method of removing heat from high-performance components.
Supermicro's liquid-cooled NVIDIA Vera CPU Rack extends the Vera architecture to rack scale, delivering high-density CPU computing infrastructure designed for demanding data center environments.
The rack contains:
- 256 NVIDIA Vera CPUs, delivering a total of 22,528 CPU cores and 45,056 threads.
- Up to approximately 300 TB/s of aggregate theoretical peak CPU memory bandwidth, providing substantial memory bandwidth across the rack.
- Up to 400 TB of SOCAMM2 LPDDR5X memory, supporting memory-intensive workloads and large-scale data processing.
From a networking standpoint, the rack ships with the liquid-cooled NVIDIA Quantum-X800 Q3300 InfiniBand switch featuring 144 ports of 800 Gb/s InfiniBand, and each compute tray contains NVIDIA ConnectX™-9 SuperNICs fully optimized for scaling HPC workloads. For demanding applications, including climate modeling, molecular dynamics, and other next-generation scientific workloads, the NVIDIA Quantum-X800 platform supports NVIDIA SHARP technology, helping reduce communication bottlenecks and improve performance and scalability across large-scale scientific computing environments.
Deployment of the Supermicro Vera CPU Rack can be accelerated by leveraging Supermicro’s Data Center Building Block Solutions (DCBBS). DCBBS offers customers a total solution with all critical computing and cooling infrastructure, alongside on-site services and support, accelerating both time-to-market and time-to-online of these powerful computing solutions.
Flexible HPC Deployments with Supermicro Air-Cooled NVIDIA Vera Systems
While liquid-cooled infrastructure enables high-density computing at rack scale, air-cooled systems continue to play an important role in HPC environments where deployment flexibility, ease of integration, and compatibility with existing data center infrastructure are key considerations.
Complementing its liquid-cooled Vera CPU Rack, Supermicro offers a portfolio of air-cooled NVIDIA Vera CPU solutions in 1U and 2U form factors. These systems bring the architectural capabilities of NVIDIA Vera to traditional server configurations, providing organizations with additional flexibility to address diverse HPC requirements without the need for dedicated liquid-cooling infrastructure.
From scientific modeling and engineering simulations to large-scale data processing and workload management, Supermicro's air-cooled Vera CPU solutions provide a versatile foundation for a broad range of HPC applications. Whether expanding existing compute clusters or deploying dedicated resources for research and engineering workloads, these systems offer organizations additional choices to scale their computing infrastructure according to their performance requirements and deployment needs.
Expanding the Future of HPC and AI Infrastructure
As the demands of HPC and AI continue to evolve, Supermicro remains committed to expanding its portfolio of high-performance, energy-efficient computing solutions to address the diverse requirements of modern data centers. Building on its established leadership in HPC infrastructure, Supermicro's new NVIDIA Vera CPU solutions introduce additional architectural and deployment choices, enabling organizations to explore new possibilities in scientific computing, AI, and other compute-intensive applications.
From flexible air-cooled servers to high-density liquid-cooled racks, these new solutions reflect Supermicro's continued investment in delivering scalable, application-optimized infrastructure for both traditional and emerging workloads.
Attendees at SC26 can explore Supermicro's air-cooled and liquid-cooled NVIDIA Vera CPU solutions on display at the Supermicro booth and learn more about how these platforms are expanding the possibilities for HPC and AI infrastructure.
Learn more about Supermicro's NVIDIA Vera CPU solutions:
- Supermicro NVIDIA Vera CPU Rack – Product Page
- Supermicro NVIDIA Vera CPU Rack - Datasheet
- Supermicro NVIDIA Vera CPU Solutions – Press Release
- Supermicro HPC Solutions – Portfolio Overview
- Supermicro Air-Cooled NVIDIA Vera CPU Solutions:
Subscribe to Data Center Stories
By clicking subscribe, you consent to allow Supermicro to store and process the personal information submitted above to provide you the content requested.
You can unsubscribe from these communications at any time. For more information on how to unsubscribe, our privacy practices, and how we are committed to protecting and respecting your privacy, please review our Privacy Policy.
