Microsoft Surface Laptop Ultra powered by NVIDIA RTX Spark. Product image: Microsoft. Used with permission from Microsoft. Original press gallery · Editorial image-use terms.
NVIDIA’s RTX Spark N1X is a Windows laptop and compact-desktop platform, distinct from the DGX Spark AI computer. NVIDIA lists laptop configurations pairing either an 18-core CPU with a 5,120-core Blackwell GPU or a 20-core CPU with a 6,144-core GPU. The larger configuration supports up to 128GB of unified LPDDR5X memory. (NVIDIA, n.d.)
Unified memory gives the CPU and GPU access to a common memory pool, a useful distinction for local AI and large creative workloads. NVIDIA’s developer article advertises one petaflop of FP4 AI compute. That low-precision AI throughput figure is not a CPU benchmark or a claim of equivalent general-purpose floating-point performance. (Chockalingam, 2026)
Windows Central’s Zac Bowden tested the 20-core, 128GB configuration in a Surface Laptop Ultra. His Cinebench 2026 table reports 558 points single-core and 6,885 multi-core when plugged in, versus 554 and 6,844 unplugged. Tests used the Best Performance profile; these are results for the disclosed machine, not every N1X laptop. (Bowden, 2026)
The same review shows why a strong GPU does not mean winning every task. Its Premiere Pro Beta H.264 export took about 42.0 seconds on the Spark system versus 32.1 seconds on a Lenovo Yoga Pro 9 with an RTX 5060. Different laptop cooling and power budgets also affect that comparison; it does not isolate the GPU alone. (Bowden, 2026)
NVIDIA’s comparison with Intel Panther Lake claims faster coding-agent, fine-tuning and compilation workloads, using named tests and models. Those remain vendor-run comparisons, not independent confirmation of a universal speed advantage. Software compatibility and the exact model, precision and memory allocation matter alongside processor specifications. PartialMedia has not benchmarked RTX Spark itself. (Chockalingam, 2026)
Editorial Opinion
RTX Spark is most interesting as a whole-system alternative, not a new CPU score to paste above Intel and AMD. Large shared memory and CUDA support may matter more to an AI developer than a synthetic CPU lead. Source quality still needs attention: Windows Central’s Geekbench narrative gives 2,718 single-core, while its tables give 2,710. We have omitted that disputed result and the review’s inconsistently labeled image-generation speed figures rather than inventing corrections. Its consistent Cinebench and export figures remain narrowly attributed. (Bowden, 2026)
References
Bowden, Z. (2026, October 15). Absolutely everything I learned while testing RTX Spark for a week: It’s a beast, but not at everything. Windows Central. https://www.windowscentral.com/hardware/nvidia/rtx-spark-review-benchmarks-2026
Chockalingam, A. (2026, October 16). Code, test, and deploy on one machine with NVIDIA RTX Spark. NVIDIA Technical Blog. https://developer.nvidia.com/blog/code-test-and-deploy-on-one-machine-with-nvidia-rtx-spark/
NVIDIA. (n.d.). RTX Spark specifications. Retrieved October 25, 2026, from https://www.nvidia.com/en-us/products/rtx-spark/specifications/