The rise of agentic AI is changing the way enterprises approach infrastructure. While AI infrastructure planning initially focused heavily on training workloads, the growing importance of inference introduced a different set of requirements. Agentic AI is taking that evolution further by turning a single request into a dynamic sequence of retrieval, tool calls, code execution and result generation, creating workloads that can vary significantly from one execution to another.
This shift is placing greater emphasis on flexibility across the infrastructure stack, including the CPUs supporting agentic AI workflows. AMD is positioning its 6th Gen AMD EPYC 9006 “Venice” server CPUs to address these varied requirements, with a new white paper evaluating the processors across general-purpose, enterprise, cloud-native, AI and high-performance computing workloads.
AMD EPYC Targets Diverse Compute Requirements
The broader workload approach reflects the changing nature of enterprise infrastructure. Databases, virtualization, analytics, web services, technical computing and AI have traditionally required different system configurations. Agentic AI increasingly brings several of these workload profiles together within a single workflow.
In SPECrate 2026 Integer testing, the AMD EPYC 9996 server CPU delivered 1.2 times the per-core performance of an Nvidia Vera-based platform and 2.24 times its platform-level performance. AMD testing across enterprise and cloud-native workloads, including server-side Java, OpenSSL, MongoDB, Redis, NGINX and transaction processing, reported performance gains ranging from 2.4x to 3.7x.
The processor also demonstrated performance advantages in high-performance computing workloads. Against the Intel Xeon 6980P processor, the AMD EPYC 9996 delivered between 1.8x and 3.13x performance advantages across molecular dynamics, materials modelling and weather forecasting.
Balancing Performance and Compute Density
As agentic workloads can change their compute requirements during execution, infrastructure needs to accommodate both latency-sensitive tasks and high levels of concurrent processing. Strong loaded per-core performance can help accelerate latency-sensitive workloads, while higher core density can support concurrency and improve rack-level throughput.
AMD’s EPYC portfolio is designed to provide different processor profiles for these requirements. The “Venice” portfolio comprises four families built on a common software foundation, spanning 8-core processors for edge deployments to 256-core flagship processors and rack-scale AI host nodes.
This allows different stages of an agentic AI pipeline to use processor configurations suited to their specific requirements while maintaining a common software environment.
Performance at the Rack Level
Beyond individual server performance, AMD is highlighting the potential impact of processor selection at the data-center level. In a modeled 100-kilowatt rack, the AMD EPYC 9996 is estimated to deliver 3.4 times the throughput of a Vera-based platform.
The comparison highlights the importance of balancing performance, core density and power constraints as enterprises scale AI infrastructure. Rather than designing infrastructure around a single workload profile, the growing diversity of AI-driven applications calls for systems that can adapt to different types of compute demand.
The “Venice” processors are already in production, with major OEM platforms on track to launch and leading cloud providers expected to begin deploying them later this year. As agentic AI introduces increasingly varied and dynamic workloads, AMD is positioning its EPYC portfolio as a flexible CPU foundation for the infrastructure supporting this next phase of AI adoption.
𝐒𝐭𝐚𝐲 𝐢𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐰𝐢𝐭𝐡 𝐨𝐮𝐫 𝐥𝐚𝐭𝐞𝐬𝐭 𝐮𝐩𝐝𝐚𝐭𝐞𝐬 𝐛𝐲 𝐣𝐨𝐢𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 WhatsApp Channel now! 👈📲
𝑭𝒐𝒍𝒍𝒐𝒘 𝑶𝒖𝒓 𝑺𝒐𝒄𝒊𝒂𝒍 𝑴𝒆𝒅𝒊𝒂 𝑷𝒂𝒈𝒆𝐬 👉 Facebook, LinkedIn, Twitter, Instagram