After more than a week of living and breathing Nvidia’s RTX Spark chip, I’m impressed by how much AI, creative and gaming power it packs into such a tiny space

I recently had the opportunity to experience Nvidia's RTX Spark chip, a system on a chip (SoC) that combines an 18-core CPU and 5120 CUDA core Blackwell RTX GPU, or a 20-core CPU and 6144-core GPU for its more powerful version. This compact powerhouse is designed to deliver significant AI, creative, and gaming capabilities in a small package.

The RTX Spark is Nvidia's first SoC, competing with Apple's M series chips. Although Apple's silicon excels in performance, the Blackwell RTX GPU in the RTX Spark offers equally impressive gaming performance. The SoC has 20 cores within its CPU, with a 2,623 single-core performance score on Geekbench 7.1. However, its multi-core score of 21,899 significantly surpasses Apple's M5 chip, which scored 17,926.

This means the RTX Spark should provide enhanced performance for tasks such as content creation, software development, and AI.

Nvidia has also introduced a hybrid system for the RTX Spark, allowing users to choose between local models and cloud frontier models for hybrid agentic AI. This approach enables users to maintain control over their sensitive data while still benefiting from the power of frontier models when necessary. During testing, the RTX Spark performed well in agentic AI tasks, processing data analyst and software engineer workflows using the Qwen 3 8B local model.

The chip also demonstrated strong performance in generative image and video workflows using ComfyUI.

Beyond AI, the RTX Spark's creative capabilities are equally impressive. Nvidia has targeted this SoC at creative professionals, with the latest Asus ProArt P16 laptop featuring the RTX Spark. Creative apps such as Blender can take full advantage of the chip's power, making it a versatile tool for artists and designers.