Daily Tribune

TECHTALKS

AMD brings ‘Venice’ CPUs

DT · Sep 30, 2026, 1:13 AM

AMD is pitching its 6th-generation EPYC 9006 ‘Venice’ server processors for agentic AI workloads that can shift between information retrieval, tool calls, code execution and other computing tasks. — Photo courtesy of AMD

AMD is positioning its 6th-generation EPYC 9006 “Venice” server processors for the growing demands of agentic artificial intelligence, where AI systems perform multiple tasks ranging from information retrieval and tool calls to code execution.

Unlike conventional AI workloads centered largely on training or inference, agentic AI can shift between different computing requirements during a single workflow, increasing demand for flexible infrastructure.

AMD said its EPYC portfolio allows data centers to match different processor configurations to general-purpose computing, enterprise applications, cloud-native services, AI and high-performance computing while maintaining a common software foundation.

A new AMD white paper evaluated its 6th-generation EPYC processors across workloads including Java, OpenSSL, MongoDB, Redis, NGINX, transaction processing, molecular dynamics, materials modeling and weather forecasting.

In AMD testing, the flagship EPYC 9996 delivered 1.2 times the per-core performance and 2.24 times the platform-level performance of an Nvidia Vera-based platform in SPECrate 2026 Integer tests.

AMD also reported performance gains ranging from 2.4 times to 3.7 times across selected enterprise and cloud-native workloads.

Against Intel’s Xeon 6980P, AMD said the EPYC 9996 delivered between 1.8 times and 3.13 times the performance across selected scientific computing workloads.

The results are based on AMD testing and modeling and may vary depending on system configurations and workloads.

The EPYC 9006 Series spans configurations for different computing environments, including high-core-count processors and AI host systems.

AMD said “Venice” processors are now in production, with major original equipment manufacturer platforms expected to launch and cloud providers set to begin deployments later this year.