NVIDIA's Vera BlueField-4 STX Storage Processor Accelerates Data Tasks by Up to 3.67x

iconKuCoinFlash
Share
This is the logo of the coin.
NVDA
----
AI summary iconSummary
NVIDIA’s Vera BlueField-4 STX storage processor accelerates data tasks by up to 3.67x, with MetaEra reporting significant improvements in encryption, compression, and data recovery. The chip delivers 1.43x higher encryption throughput, 3.26x faster Reed-Solomon recovery, and 3.67x faster CRC32C checks. Designed for distributed storage environments, it leverages 88 Armv9.2 cores to reduce CPU, power, and cooling requirements in AI storage systems. Decentralized storage solutions can benefit from enhanced efficiency and performance.
ME AI News: NVIDIA has released the latest benchmark results for its Vera BlueField-4 STX storage processor, showing that its Vera CPU-based native AI storage platform delivers significant performance improvements over traditional x86 CPUs in critical tasks such as encryption, compression, data integrity verification, and recovery. NVIDIA states that as AI Agent applications scale, storage systems must continuously handle enterprise knowledge bases, long-term memory, KV caches, tool invocation data, and model-generated outputs—tasks where traditional CPUs are increasingly becoming performance bottlenecks. The benchmarks show Vera CPU outperforming comparable x86 CPUs across multiple storage tasks: - AES-128 encryption throughput improved 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 verification performance improved 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 combining compression and encryption, overall throughput improved by up to 3.21x. The Vera CPU features NVIDIA’s proprietary Olympus core architecture, incorporating 88 Armv9.2-compatible CPU cores supporting 176 threads, along with Scalable Coherency Fabric (SCF) and SOCAMM2 LPDDR5X memory systems, delivering up to 3.4 TB/s interconnect bandwidth and up to 1.2 TB/s memory bandwidth. NVIDIA emphasizes that AI factories rely not only on GPUs for model inference but also require high-performance CPUs and storage infrastructure to support AI Agents during tool invocation, data retrieval, and task processing. By integrating Vera CPU capabilities directly into the storage data path, the BlueField-4 STX helps reduce CPU resource, power, and thermal burdens on native AI storage platforms. NVIDIA notes that future AI workloads will demand higher concurrency, larger context sizes, and greater data processing requirements—and the Vera architecture is designed to enhance the协同 efficiency of compute, storage, and AI inference infrastructure within data centers. (Source: BlockBeats)
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.