Arm-based NVIDIA RTX Spark is redefining PCs for the agentic era

Arm Newsroom's Chris Bergey, Executive Vice President of the Edge AI Business Unit, explains how Arm-based NVIDIA RTX Spark is redefining personal computing for the agentic era. The article frames this as a fundamental shift: for decades, PCs were designed around people using applications, but agentic AI changes that model — apps become autonomous, with users prompting agents to observe, reason, plan, and act across complex multi-step workflows. This evolution creates demand for two broad classes of AI-enabled computing: efficient performance for all-day battery life and AI-enhanced productivity, and extreme performance for advanced AI workloads, content creation, software development, gaming, and local inference — both of which RTX Spark addresses through its integrated Arm CPU and Blackwell GPU architecture.

According to NVIDIA VP Kaustubh Sanghani, "Agents require tight integration and optimization of GPU, CPU, and memory architectures to deliver responsive, efficient on-device AI experiences." RTX Spark achieves this by tightly coupling an Arm-based Grace CPU with NVIDIA's Blackwell RTX GPU and unified memory via NVLink-C2C, combining NVIDIA's leadership in accelerated computing and AI to reinvent personal computing. Arm emphasizes that DGX Spark and RTX Spark together show how the market is expanding beyond traditional PCs toward a broader class of platforms — DGX Spark for compact AI development systems and RTX Spark bringing AI acceleration to Windows PCs for creators, developers, and gamers.

The article highlights that the next wave of PC experiences — agentic code generation, multi-stage reasoning, planning, dynamic workflows, agentic RAG, and personal AI assistants with continual task execution — requires more CPU and GPU cores than ever before. As AI models use increasingly large numbers of tokens, cost-per-task has been rising significantly. Arm argues that the path to reducing these costs requires large on-device models that use tokens more efficiently, yielding lower and predictable costs while ensuring data privacy for users and enterprises. RTX Spark, with its 1,000+ TOPS of AI performance capable of running 120B-parameter models locally, represents what Arm calls "one of the clearest examples of the emerging agentic PC era, powered by the Arm compute platform."