TechPrufer
Get started
NVIDIA
NVIDIA logoNVIDIA · Feature trail

NVIDIA Vera Storage Processor Performance

3 posts · Jun 2026 – Aug 2026

What this trail covers

This feature thread tracks the evolution of NVIDIA's storage processing capabilities, focusing on the performance and efficiency gains provided by the NVIDIA Vera BlueField-4 STX Storage Processor. Initial efforts focused on understanding the demands of AI-native storage for agentic AI workloads, which require high throughput for operations like encryption, compression, integrity checking, and data recovery. Subsequent developments have demonstrated the Vera CPU's ability to significantly outperform x86 CPUs in agentic workload execution, particularly in sandboxed code execution, data retrieval, and orchestration tasks. The Vera CPU's design emphasizes high per-core performance, high concurrency, and energy-efficient memory bandwidth, making it a critical component for scaling agentic AI and reinforcement learning in AI factories. The Olympus core, with its advanced branch prediction and out-of-order execution, coupled with a high-bandwidth LPDDR5X memory subsystem and the NVIDIA Scalable Coherency Fabric, delivers superior performance for these demanding workloads.

All NVIDIA trails
  1. NVIDIA Vera Storage Benchmarks: Faster Encryption, Compression, Integrity Checking, and Recovery for AI-Native Storage | NVIDIA Technical BlogThis post benchmarks the NVIDIA Vera BlueField-4 STX Storage Processor, demonstrating its performance advantages over x86 CPUs for AI-native storage workloads. It details throughput gains of up to 1.43x for AES-128 encryption, 3.26x for Reed-Solomon recovery, 3.67x for CRC32C integrity checking, and 3.29x for compression. The benchmarks highlight how Vera's architecture, with its 88 Armv9.2 cores, Spatial Multithreading, and high-bandwidth memory, addresses the dual demands of sustained per-core performance and system-wide bandwidth for storage primitives. The post emphasizes that these improvements enable storage platforms to secure, protect, validate, and compress more data with less CPU and power overhead, increasing storage-processing throughput and efficiency.
  2. NVIDIA Vera CPU: Olympus Cores Built for Maximum Single-Thread Performance in Agentic AI | NVIDIA Technical BlogThis post details the NVIDIA Vera CPU architecture, specifically the Olympus core, designed for maximum single-thread performance in agentic AI workloads. It highlights the core's front-end optimizations (advanced branch prediction, 10-wide decode), mid-core for exposing parallelism (wide rename/allocation, large reorder buffer, dependency breaking), execution engine for diverse operations, and cache subsystem with graph prefetcher for irregular memory accesses. It also discusses NVIDIA Spatial Multithreading for improved sandbox responsiveness and the Scalable Coherency Fabric for high on-die bandwidth. The post emphasizes the co-design approach with the Vera Rubin platform and the importance of CPU performance for agentic AI responsiveness and AI factory throughput.
  3. NVIDIA Vera CPU Boosts AI Factory Throughput to Accelerate Agentic Workloads | NVIDIA Technical BlogThis post details how the NVIDIA Vera CPU architecture, with its 88 Olympus cores, monolithic compute die, unified cache, and Scalable Coherency Fabric, directly addresses CPU-bound bottlenecks in agentic AI workloads. It quantifies performance improvements in Reinforcement Learning (RL) training, showing 1.8x faster sustained per-core performance leading to higher policy gradient quality and environment rollout completion rates. For interactive agentic deployments, Vera CPU achieves 40% lower peak loaded latency and over 3x the per-core memory bandwidth at less than half the power of traditional x86 CPUs, enabling predictable latency and higher GPU utilization by minimizing KV-cache evictions and recomputation. The post highlights how these CPU-side optimizations are critical for maximizing overall AI factory throughput and adhering to service level agreements.
  4. NVIDIA Vera CPU Sets a New Standard for Agentic Workloads in AI Factories | NVIDIA Technical BlogThis post introduces the NVIDIA Vera CPU and its suitability for agentic AI workloads. It details how the Vera CPU, with its 88 NVIDIA Olympus cores and 1.2 TB/s LPDDR5X memory bandwidth, addresses the increasing importance of CPU performance in the agentic AI loop. The post highlights the Olympus core's architectural improvements (wide front end, advanced branch prediction, deep out-of-order execution) and the memory subsystem's low latency and high bandwidth. It quantifies the performance gains, showing over 1.8x higher sandbox performance compared to x86 architectures in agentic workloads. The article also discusses the shift in AI factory metrics from 'cores per dollar' to 'tokens per dollar' and how Vera CPU's design aligns with this new paradigm.