Odaily Planet Daily report: NVIDIA has released benchmark results for its Vera CPU, showing that its NVIDIA BlueField-4 STX storage processor significantly enhances encryption, compression, integrity verification, and data recovery performance in AI-native storage, helping businesses meet growing data processing demands in the era of Agentic AI.
NVIDIA stated that as AI agents perform knowledge retrieval, tool invocation, long-term memory management, and handle larger context windows, storage systems are no longer merely simple data read/write components but have become critical elements in the AI inference pipeline. Large volumes of data require encryption, compression, checksum verification, and recovery along the storage path—tasks typically handled by CPUs, which can become performance bottlenecks in AI infrastructure. Benchmark tests show that the BlueField-4 STX storage processor equipped with Vera CPUs achieves significant performance improvements over comparable x86 CPUs across multiple storage tasks: AES-128 encryption performance is improved by up to 1.43x.
AES-128 decryption performance improved by up to 1.29x
Reed-Solomon data recovery performance improved by up to 3.26x
CRC32C integrity verification performance improved by up to 3.67x
Compression performance improved by up to 3.29x
Decompression performance improved by up to 1.72x
Compression and encryption multi-stage storage process performance improved by up to 3.21x
NVIDIA stated that traditionally, scaling CPU storage processing capacity requires increasing the number of cores, power consumption, and cooling costs, while AI-native storage must maintain low latency under higher concurrency and larger data volumes. Vera enhances the processing capacity per CPU resource, enabling storage systems to support more AI agent workloads without significantly increasing infrastructure costs.