The Nvidia RTX Spark SoC is available in two configurations. The first features an 18-core CPU and a GPU with 5120 CUDA cores, while the second offers a 20-core CPU and 6144 CUDA cores. Systems support between 24 GB and 128 GB of RAM, with consistent memory bandwidth across all models.
In standard computing tasks, the RTX Spark performs similarly to the 16-core Apple M6 or the Snapdragon X2E. It is approximately 20% faster than the Core Ultra X7 358H and nearly 50% faster than the Ryzen 9 8945HS, positioning it as a mid-range processor rather than a top-tier mobile solution.
AI performance varies significantly by scenario. In Qwen 3.8 27B tests, the chip outperformed the M5 Max and was more than twice as fast as the Ryzen AI Halo. In AI workloads, the platform is comparable to mobile RTX 5060 or RTX 4070 discrete GPUs, depending on the specific task.
Gaming performance is constrained by the Arm architecture, which lacks native support for all titles. In general, the integrated graphics perform at the level of an RTX 4060 Laptop, occasionally trailing or slightly exceeding the Ryzen AI Max with Radeon 8060S/8065S graphics, depending on the game.






