Vera: High Throughput for AI Applications

NVIDIA has introduced a significant innovation with the Vera CPU, specifically designed for the needs of AI agents. These agents do more than just generate answers; they seek information, query databases, and write and test code. This requires effective collaboration between CPU and GPU processes, enabling AI agents to operate efficiently without unnecessary wait times.

With the growing deployment of AI agents in businesses, there is an increasing need for powerful CPUs that allow concurrent execution of multiple tasks. The limitation of available energy makes it even more critical to choose infrastructures that deliver maximum performance per watt consumed. NVIDIA's Vera CPU offers both high performance and energy efficiency, making it possible to perform a wide range of agent tasks with a single CPU, without the need to mix various x86 components.

Data shows that the Vera CPU outperforms the fastest and most energy-efficient x86 CPUs in a comparison within a 100-megawatt data center. It appears that Vera can process up to 1.6 times more agents per watt compared to the fastest x86 CPU and is 2 times faster with optimal energy efficiency compared to the most energy-efficient x86 CPU.

Several partners have already achieved positive results with the Vera CPU. For example, Perplexity performed agentic coding tasks 1.5 times faster with Vera than with traditional CPUs, and DeepInfra achieved up to 2.2 times the speed of the best x86 processor in agentic orchestration. Other successful applications include those from ClickHouse, DuckLabs, and Daytona, all reporting significant performance improvements when using Vera in production environments.

Read the full article from Nvidia.