NVIDIA Vera CPU Boosts AI Factory Throughput to Accelerate Agentic Workloads
NVIDIA 推出 Vera CPU,旨在通过加速 CPU 间模型步骤的工作负载,提升 AI 工厂中智能体系统的吞吐量。智能体系统利用多步工作流将模型推理转化为行动,涵盖推理、工具使用、代码执行、检索、编排及结果处理等环节。随着智能体系统在 AI 工厂中的规模化扩展,其性能不仅依赖 GPU 加速,更取决于模型步骤间的 CPU 工作表现。
EVENT DOSSIER
NVIDIA releases Vera CPU, which uses Olympus cores to optimize single-thread performance to address challenges in intelligent agent clusters.
To address the challenges posed by unpredictable and highly variable workloads in intelligent factories due to the operation of Agentic AI, NVIDIA has introduced the Vera CPU. This processor is designed specifically for agent-based AI, with an Olympus architecture at its core, aimed at maximizing single-thread performance. As agents move more critical execution paths to the CPU, including concurrent execution of code in sandboxes, tool calls, context retrieval, and result analysis, CPU performance directly determines the response speed of individual agents and the overall throughput of the factory. The Vera CPU effectively improves the operational efficiency and scalability of agent systems in AI factories by accelerating CPU work between model steps.
企业 · 官方发布across 3 days
NVIDIA releases Vera CPU, which uses Olympus cores to optimize single-thread performance to address challenges in intelligent agent clusters.
NVIDIA introduces Vera CPU, designed to improve overall throughput of intelligent agent systems in AI factories by accelerating model-step workloads between CPUs.
NVIDIA releases Vera CPU, tailored for proxy-type AI. Its Olympus cores focus on maximizing single-thread performance, thereby determining the response speed of intelligent agents and factory throughput.
NVIDIA introduces Vera CPU, with GPUs running models and CPUs handling orchestration and tool execution to handle unpredictable workloads in intelligent factories.
NVIDIA 推出 Vera CPU,旨在通过加速 CPU 间模型步骤的工作负载,提升 AI 工厂中智能体系统的吞吐量。智能体系统利用多步工作流将模型推理转化为行动,涵盖推理、工具使用、代码执行、检索、编排及结果处理等环节。随着智能体系统在 AI 工厂中的规模化扩展,其性能不仅依赖 GPU 加速,更取决于模型步骤间的 CPU 工作表现。
NVIDIA 推出 Vera CPU,专为代理型 AI(Agentic AI)打造,其 Olympus 核心聚焦于最大化单线程性能。随着代理型 AI 将更多关键执行路径移至 CPU,智能体在沙箱中并发执行代码、调用工具、检索上下文及分析结果,CPU 性能直接决定单个智能体的响应速度及整体工厂吞吐量。
NVIDIA 推出 Vera CPU 以解决智能体 AI 集群挑战。在智能工厂中,GPU 运行模型,CPU 负责编排、工具执行及沙箱计算。与传统计算稳定的运行时不同,智能体工作负载具有不可预测且高度变化的特征。