**[NVIDIA Launches AI Storage "Accelerator": Encryption, Compression, and Recovery Speeds Increased by Up to 3.67x]** BlockBeats News, August 4th, NVIDIA unveiled the latest benchmark results for the Vera BlueField-4 STX storage processor, showing that its AI-native storage platform based on the Vera CPU architecture delivers significant performance improvements in critical tasks such as encryption, compression, data integrity checks, and recovery compared to traditional x86 CPUs. NVIDIA stated that as the scale of AI Agent applications expands, storage systems need to continuously handle enterprise knowledge bases, long-term memory, KV caches, tool invocation data, and model generation results. Traditional CPUs are increasingly becoming performance bottlenecks in the data processing pipeline. Tests revealed that Vera CPU outperformed x86 CPUs in multiple storage tasks: - AES-128 encryption throughput increased by up to 1.43x, decryption throughput by up to 1.29x. - Reed-Solomon data recovery performance improved by up to 3.26x. - CRC32C data integrity check performance increased by up to 3.67x. - Compression throughput improved by up to 3.29x, decompression performance by up to 1.72x. - In multi-stage storage workflows involving compression + encryption, overall throughput increased by up to 3.21x. The Vera CPU utilizes NVIDIA's proprietary Olympus core architecture, featuring 88 Armv9.2-compatible CPU cores, supporting 176 threads, and equipped with Scalable Coherency Fabric (SCF) and SOCAMM2 LPDDR5X memory systems. It delivers up to 3.4TB/s interconnect bandwidth and up to 1.2TB/s memory bandwidth. NVIDIA emphasized that AI factories not only rely on GPUs for model inference but also require high-performance CPUs and storage infrastructure to support AI Agents in tool invocation, data retrieval, and task processing. By integrating Vera CPU capabilities into the storage data path, BlueField-4 STX helps AI-native storage platforms reduce CPU resource consumption, power usage, and thermal pressure. NVIDIA stated that future AI workloads will bring higher concurrency, larger context scales, and greater data processing demands. The Vera architecture is designed to enhance the collaborative efficiency of computing, storage, and AI inference infrastructure within data centers. [Original Link]