RTX Spark: A Windows-on-ARM SoC Built Around a Serious GPU
RTX Spark is a consumer-grade system-on-chip that combines a 20-core ARM CPU with a Blackwell-based RTX GPU and unified LPDDR5X memory, designed specifically to run Windows 11 on ARM while delivering desktop-class graphics and AI compute performance in laptops and compact desktops. NVIDIA finally revealed the RTX Spark processor last month after years of rumor, confirming a collaboration with MediaTek and a Grace GB10-class 20-core CPU paired with a Blackwell GPU. The real story, however, is no longer the announcement itself, but what the first Windows 11 drivers expose about the chip’s GPU configuration. Version 616.00 of the GeForce driver for Windows on ARM is labeled as a Developer Preview and targeted at a Microsoft Surface RTX Spark dev box, yet its .inf files give away key RTX Spark GPU specs and core counts. That leak moves RTX Spark from vague promise into concrete product strategy.
Dual RTX Spark GPU Variants: 6,144 vs 5,120 CUDA Cores
The most important takeaway from the 616.00 Windows 11 drivers is that RTX Spark will ship in at least two clear GPU variants. The nv_surface_woa.inf file lists two "NVIDIA RTX Spark N1X" configurations: one with a 6,144-core Blackwell RTX GPU and another with 5,120 CUDA cores. According to the driver data, "Windows Latest were able to analyse this driver and uncover two Nvidia RTX Spark N1X GPU configurations. These include a 6144-core Blackwell RTX GPU and a 5120-core RTX Blackwell GPU." Early reporting suggests the 6,144-core GPU will be tied to a 20-core CPU and 48 Streaming Multiprocessors, while the 5,120-core version pairs with an 18-core CPU and 40 Streaming Multiprocessors. On paper, the flagship RTX Spark matches the CUDA core count of a desktop RTX 5070, while the smaller variant aligns with counts seen in existing laptop GPUs such as RTX 3070 and RTX 3080-class silicon. The numbers alone do not guarantee equal performance, but they confirm NVIDIA is not treating Spark as a token low-end ARM part.
| Spec | Flagship RTX Spark N1X | Smaller RTX Spark N1X |
|---|---|---|
| CUDA core count | 6,144 | 5,120 |
| Likely CPU pairing | 20-core ARM CPU | 18-core ARM CPU |
| Streaming Multiprocessors (reported) | 48 | 40 |
| Memory type | Unified LPDDR5X | Unified LPDDR5X |
Windows 11 Drivers Turn RTX Spark Into a Real ARM GPU Laptop Platform
The first GeForce driver for Windows 11 on ARM is more than a technical milestone; it is NVIDIA’s declaration that RTX Spark is ready for real devices. This 616.00 release is a Developer Preview for the Surface RTX Spark dev box and is not meant for consumers, but it confirms formal GeForce ARM drivers and N1X specifications. It also confirms the SoC’s pairing with ARM-based CPU cores and unified LPDDR5X memory, hitting up to 128GB of supported RAM. For users, that unified memory pool could be more impactful than raw CUDA core count, especially for data-heavy AI workloads that struggle with discrete GPU memory limits. NVIDIA has already shown games such as Alan Wake 2 and Pragmata running smoothly on the platform with path tracing enabled, demonstrating that the RTX Spark architecture can handle demanding real-time graphics despite its mobile power envelope. In practical terms, this marks RTX Spark as a credible ARM GPU laptop option rather than an experimental sideline, and it forces the broader Windows-on-ARM ecosystem to take dedicated drivers and performance seriously.
From Dev Boxes to Desktops: What This Means for Local AI Workloads
While the 616.00 driver currently targets a development Surface box, its contents point directly to a wider RTX Spark rollout across ARM GPU laptops and compact desktops. Partners are already preparing a range of laptops powered by RTX Spark, and desktop products from brands such as Acer, ASUS, Dell, GIGABYTE, Lenovo and MSI are reported for release by the holidays. There are explicit notes for a "Nvidia Desktop Device" inside the driver, confirming that Spark silicon will not be confined to portable machines. For local AI, this matters: NVIDIA advertises Spark desktops as machines "built to run personal AI agents 24/7 right at your desk plus game and create with the full power of RTX graphics." With up to 6,144 CUDA cores, an integrated NPU, and up to 128GB of unified memory, an RTX Spark ARM GPU laptop or ultra-efficient desktop becomes a plausible all-in-one box for fine-tuning models, running AI agents, and still playing modern games on Windows. The constraint will not be hardware ambition but how quickly software and drivers mature around it.
