Resource Library

Intel® Advanced Matrix Extensions (Intel® AMX) Enhance vLLM Inference Through CPU-GPU Co-Serving: A Multi-Agent Architecture Leveraging Supermicro HGX B200 System

This IBM Redbook presents a validated reference architecture for an AI infrastructure that allows KV cache to be persistently stored, shared, and reused across requests, sessions, and GPU nodes. It consists of NVIDIA Dynamo for intelligent distributed KV cache management, IBM® Storage Scale Erasure Coding Edition (ECE) as the high-performance shared storage tier, Supermicro Petascale servers as the storage and networking foundation, and NVIDIA Spectrum-X Ethernet to tie it all together with the low-latency, high-bandwidth fabric that production AI inference demands.

This solution brief outlines the performance breakthroughs and business value of the myrtle.ai VOLLO™ acceleration stack running on a Supermicro AS -2015CS-TNR server, as showcased in the recently released STAC ML benchmark report.

Supermicro AI solutions featuring Arm AGI CPUs are built for the age of massive-scale agentic AI orchestration, delivering performance, efficiency, and density that maximizes the economics of rack-scale deployments. 

The rapid proliferation of AI agents is creating a need for energy-efficient systems that deliver the required results within the expected timeframe. While AI training has dominated recent discussions, the rising demand for inference and agentic workloads calls for energy-optimized systems that operate across diverse environments. The rise of agentic AI will require CPUs capable of orchestrating and managing continuously running agents. This inflection point will require significantly more CPUs at rack-scale density, while keeping energy consumption within the power envelopes of current enterprise data centers.

NVIDIA GB300 Performance. Supermicro Ingenuity. From Desk to Data Center.

New System is Designed for OCP ORv3 Racks for Increased Data Center Compatibility

Supermicro X14 Servers Give Users the Flexibility to Match Systems and Processors to Workloads

Introducing Supermicro's X14 Generation supporting Intel® Xeon® 6/6+ processors – the latest generation of proven platforms designed for maximum performance, efficiency and flexibility for AI, Cloud, Storage and 5G/Edge workloads.

High Performance, Density, and Efficiency with Resource-Saving Architecture – Supermicro’s SuperBlade® and MicroBlade® platforms are designed to meet the growing demand for higher performance per rack with industry leading density and efficiency. These multi-node systems maximize compute capacity while minimizing power consumption, cabling requirements, and operational overhead.

Verda and Supermicro’s deployment in Europe runs on the latest NVIDIA technologies and is powered by 100% Renewable Energy