What RTX Spark Is and Why Its Smartphone Roots Matter
RTX Spark is NVIDIA’s first Windows SoC chip, a smartphone-style PC processor that combines Arm-based CPU cores, a Blackwell RTX GPU, and unified memory to create an AI-focused Windows platform. Instead of the usual PC layout of separate CPU, GPU, and memory pools, RTX Spark follows a phone-style system-on-chip design, scaled up for creator laptops, workstations, and compact Windows PCs. The heart of the chip is NVIDIA’s N1X GB10 Grace Blackwell Superchip, which brings 20 Armv9 CPU cores and an integrated RTX-class GPU into a single package tuned for AI workloads. This AI-first approach shifts the design goal from peak gaming frame rates to running large local models efficiently, with up to 128GB of shared LPDDR5X memory feeding both CPU and GPU. In short, RTX Spark makes a Windows PC behave more like a powerful AI phone than a traditional desktop computer.
Inside the Smartphone-Style RTX Spark Architecture
At the CPU level, RTX Spark looks closer to a flagship phone chip than a conventional laptop processor. NVIDIA uses ten Arm Cortex-X925 "powerhouse" cores and ten Cortex-A725 performance cores, all based on the modern Armv9 instruction set. MediaTek helped design this CPU block, which explains the similarity to the Dimensity 9400 smartphone platform, though RTX Spark scales clocks to around 4.0GHz on X925 and 2.85GHz on A725 for higher per-core performance. Cache sizes are also phone-inspired: each X925 core can access up to 2MB of L2 cache, A725 cores up to 512KB, backed by 16MB L3 and 16MB of system cache. According to Android Authority, "At its core, RTX Spark is powered by the same Arm CPU technology as flagship smartphones." This smartphone-style core cluster, multiplied and pushed to higher frequencies, aims to deliver strong everyday performance while staying efficient enough for thin-and-light NVIDIA AI laptops.
Unified Memory and NVLink-C2C: Built for Local AI Models
Where RTX Spark breaks sharply from traditional PC design is memory. Instead of separate pools for system RAM and GPU VRAM, the chip uses 128GB of unified LPDDR5X memory positioned around the SoC. CPU and GPU share this single pool through NVIDIA’s NVLink-C2C interconnect, which provides up to 600 GB/s of bidirectional bandwidth and a shared address space with minimal overhead. This mirrors how smartphone SoCs feed CPU, GPU, and AI engines from one LPDDR pool, only scaled up dramatically. For AI-first computing, this matters more than raw CPU scores. Large language models and creative AI tools need fast, contiguous memory more than they need an extra few percent of gaming performance. With RTX Spark, a compact Windows PC can keep far larger models resident in memory, reduce data copying between CPU and GPU, and run on-device AI assistants or creative pipelines that would choke on 16GB, split-memory designs.
Real-World RTX Spark Systems: Laptops and Mini PCs
At Computex, OEMs treated RTX Spark less like a niche experiment and more like a new class of NVIDIA AI laptop and compact Windows PC. Dell’s XPS 16 Creator Edition uses the SoC in a slim 16-inch machine with Tandem OLED and full creator-grade ports. ASUS goes further with ProArt P16 and P14 laptops plus a Mini PC, each wiring the RTX Spark architecture to 128GB unified memory via eight LPDDR5X chips and cooling the SoC with a dual-fan, copper heatpipe design. HP’s OmniBook Ultra 16 and OmniBook X 14 focus on ultra-thin AI PCs with OLED displays and rear-mounted cooling, while Lenovo’s Yoga Pro 9n and MSI’s Prestige N16 Flip AI+ target premium creator and 2‑in‑1 designs. Collectively, these systems show RTX Spark fitting into thin-and-light 14‑inch laptops, 16‑inch workstations, and mini desktops without needing separate, bulky GPUs.
Why Smartphone-Style SoCs Could Redefine Windows PCs
RTX Spark’s smartphone-style PC processor design alters what we expect from a Windows chip. Instead of chasing the highest boost clocks or discrete GPU pairing, NVIDIA’s Windows SoC chip aligns the whole platform around efficient AI throughput. The Blackwell GPU provides up to 1 PFLOPs of AI performance for on-device inference, backed by the massive unified memory pool. That means a future Surface, ThinkPad, or ProArt laptop can act as a standalone AI workstation, running complex local agents or creative tools without constant cloud calls. This shift also enables quieter, smaller systems: mini PCs and thin laptops gain GPU-class AI acceleration without the thermal budget of a separate graphics card. RTX Spark will not replace high-end dedicated gaming rigs, but it creates a new category of compact, AI-capable Windows devices where battery life, thermals, and local model performance matter more than traditional benchmark bragging rights.
