AMD & Red Hat: Addressing the Data Centre Challenge with AI

AMD and Red Hat have announced an expanded collaboration to address growing AI infrastructure requirements at Red Hat Summit.

Both companies outline plans to deliver solutions for data centres struggling with the dual challenge of maintaining traditional operations whilst accommodating intensive AI processing demands – centreing on AMD's hardware capabilities combined with Red Hat's software platforms.

AMD Instinct GPUs, the company's data centre graphics processing units designed for AI and high-performance computing, are now fully supported on Red Hat OpenShift AI – OpenShift AI being Red Hat's platform for developing and deploying machine learning (ML) models.

This integration provides processing power for AI deployments across hybrid cloud environments, which combine on-premises infrastructure with public cloud services.

How AMD and Red Hat are using AI to address data centre infrastructure challenges

The partnership addresses the fundamental challenge facing data centre operators. 

Traditional data centres dedicate most resources to established IT systems, leaving limited capacity for AI workloads that require substantial computational power.

The companies aim to resolve this through Red Hat's open-source software combined with AMD's high-performance computing portfolio.

“Fully realising the benefits of AI means that organisations must have the choice and flexibility to optimise their IT footprint for the rigors of scaling demand,” says Ashesh Badani, SVP and CPO at Red Hat.

“Our extended collaboration with AMD expands the spectrum of options for organisations seeking to ready their IT environments for an ever-evolving future, from modernising existing investments on a high-performing CPU architecture and virtualisation platform to preparing for production AI with next-generation hardware accelerators and open source AI technologies.”

The technical focus

AMD's x86-based processors, which use the dominant instruction set architecture for servers, will work alongside the company's GPU architectures through Red Hat AI software.

This combination targets optimised environments that reduce costs whilst maintaining production readiness for AI applications.

Testing conducted on Microsoft Azure ND MI300X v5 instances demonstrated AI inferencing capabilities for both small and large language models (LLMs) – and the tests showed successful deployment across multiple GPUs within single virtual machines, reducing the need to distribute workloads across multiple VMs and cutting performance costs.

The collaboration focuses on three technical areas:

Performance improvements on AMD GPUs involving upstreaming the AMD kernel library and optimising components including the Triton kernel and FP8 precision format.
Multi-GPU support enhancements targeting improved collective communication between processors and workload optimisation.
Ecosystem engagement including cross-collaboration with industry partners such as IBM

Red Hat and AMD are also working within the upstream vLLM community to drive these improvements – vLLM is an open-source library for fast LLM serving that enables efficient deployment of AI models in production environments.

The role of OpenShift Virtualisation 

Red Hat OpenShift Virtualisation is another component of the expanded partnership.

This feature within Red Hat's OpenShift container platform provides a pathway for organisations to migrate and manage virtual machine workloads using cloud-native application methodologies.

The virtualisation platform has been validated for AMD EPYC processors, the company's server CPU line designed for data centre applications.

These processors offer performance and power efficiency characteristics that enable deployment across hybrid cloud environments.

When organisations refresh legacy data centre infrastructure, this approach enables high infrastructure consolidation ratios – consolidation reduces total cost of ownership across hardware procurement, software licensing, and energy consumption.

This means that for data centre executives facing pressure to accommodate AI workloads whilst maintaining existing systems, the partnership offers a method to optimise current infrastructure whilst preparing for AI innovations.

Furthermore, the companies tested AI inferencing performance using vLLM software on AMD Instinct GPUs, demonstrating capabilities for both small and LLMs deployed across multiple GPUs on single virtual machines. 

This configuration reduces deployment complexity compared to multi-VM approaches whilst maintaining performance standards.

The testing results showed successful performance across Microsoft Azure's ND MI300X v5 instances, which feature AMD's MI300X accelerators designed for AI and high-performance computing workloads – demonstrating the technical feasibility of the integrated approach across cloud environments.

“As enterprise customer workloads grow more diverse and demanding, they require solutions that can scale,” says Philip Guido, EVP and CCO at AMD.

“By combining Red Hat's industry-leading open source platforms with world-class AMD Instinct GPUs and AMD EPYC CPUs, we're delivering the performance and efficiency customers demand to accelerate AI, virtualisation and hybrid-cloud innovation.”

Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. 

Discover all our upcoming events and secure your tickets today.

Also sign up to our free weekly newsletter for the latest insights and stories straight into your inbox.

AI Magazine is a BizClik brand