Search
On July 21st, NVIDIA unveiled that, according to the test results from its AI cloud platform, DeepInfra, the Vera CPU exhibits a remarkable improvement in orchestration speed. Specifically, under AI agent workloads, it can achieve speeds up to 2.2 times faster than those of other CPU products. Additionally, it supports a significantly higher number of concurrently running agents, up to 1.6 times more. The DeepInfra platform handles close to 5 trillion Tokens weekly, with roughly 30% of this volume stemming from agent applications. NVIDIA emphasized that as AI agents increasingly shoulder responsibilities in inference, planning, and tool invocation, the scheduling efficiency of CPUs during model invocation will emerge as a pivotal factor influencing the performance of AI infrastructure.
-
C114 Communication Network -
Communication Home
