What are RTX Spark’s claimed performance specs?

Nvidia claims RTX Spark can blend gaming and AI workloads

Nvidia’s RTX Spark is being marketed as an Arm-based “superchip” for Windows laptops and desktops that combines CPU and GPU capabilities in one platform. Across coverage tied to Computex and partner device plans, Nvidia’s headline claims focus on both traditional consumer graphics performance and the ability to run very large AI models locally.

The most concrete specification figures described in reporting include:

• Up to 20 CPU cores.
• A Blackwell GPU with 6,144 CUDA cores.
• Performance targets ranging from “100 FPS 1440p gaming” to local AI inference claims tied to 120B-parameter models.

One article also characterizes the chip as offering about 1 petaflop of AI computing power, reinforcing Nvidia’s intent to make the platform compelling for AI beyond casual desktop tasks.

These specs matter because they represent a departure from Nvidia’s prior primary role in personal computers: GPU cards attached to CPUs supplied by other vendors. With RTX Spark, Nvidia is trying to make itself a direct architect of the overall PC compute stack.

From a product standpoint, the company is also pairing performance claims with efficiency messaging—RTX Spark is described as “the most efficient PC chip ever built.” That’s particularly relevant for laptops, where power budgets and cooling constraints can determine whether “AI on-device” claims are practical.

The rollout strategy further supports the technical pitch. Nvidia’s CEO has highlighted Microsoft’s and Nvidia’s partnership with multiple PC OEMs, with devices expected in the fall, meaning the company isn’t only advertising silicon—it’s preparing a broader ecosystem to validate its performance claims in shipping hardware.