What RTX Spark Is and Why It Matters
NVIDIA RTX Spark is an Arm-based Windows PC platform that combines a 20-core Grace CPU, a Blackwell RTX GPU, and up to 128GB of unified LPDDR5X memory in a single 3nm package to deliver up to 1 petaFLOP of local AI performance, AAA gaming, and creator workloads in thin laptops and compact desktops. Announced at Computex, RTX Spark is NVIDIA’s first processor built specifically for Windows PCs and its clearest answer yet to Apple Silicon and Qualcomm’s Snapdragon X series. By pairing an Arm CPU, powerful GPU, and shared memory pool, RTX Spark follows a design similar to Apple’s chips but aims to bring that model to the broader Windows ecosystem. Microsoft is a close launch partner, with RTX Spark positioned as the hardware foundation for a new wave of local AI agents and secure, on-device inference on Arm-based Windows PCs.
Inside the 20-Core Grace CPU: Smartphone DNA Scaled Up
At the heart of every RTX Spark processor is a custom 20-core Grace CPU designed with MediaTek and built on Arm cores that echo modern smartphone chips. The configuration combines ten Cortex-X925 performance cores with ten Cortex-A725 efficiency cores, mirroring the big.LITTLE philosophy that powers many flagship phones but scaled for PC-class workloads. This split lets RTX Spark ramp up the X925 cluster for heavy AI inference, AAA gaming, or 12K video editing, while the A725 cores handle background tasks and everyday apps with lower power draw. According to Gizmochina, the Grace CPU design is “developed in collaboration with MediaTek,” which helped tune power efficiency, performance, and connectivity. For Windows on Arm, this hybrid layout is critical: it promises long battery life in 14mm-thick laptops without sacrificing the single-threaded speed needed to keep up with Apple Silicon competitor platforms in everyday use.
Blackwell GPU and Unified Memory: NVIDIA’s Answer to Apple Silicon
RTX Spark’s big swing at Apple Silicon competitor status comes from pairing its Grace CPU with a Blackwell RTX GPU and a large unified memory architecture. The integrated Blackwell GPU offers 6,144 CUDA cores, 48 streaming multiprocessors, fifth-generation Tensor Cores, and around 1 petaFLOP of FP4 AI compute, a figure comparable to NVIDIA’s GeForce RTX 5070 in FP4 throughput. Instead of splitting RAM between CPU and GPU, Spark uses up to 128GB of unified LPDDR5X memory on a single package, with CPU and GPU linked by an NVLink-C2C interconnect. This mirrors Apple’s approach but pushes capacity higher for AI and creative workloads. NVIDIA says RTX Spark systems can run AI models with up to 120 billion parameters and support context windows of up to one million tokens, while also rendering 3D scenes larger than 90GB and editing 12K 4:2:2 video entirely in local memory.
Gaming and AI on Arm-Based Windows PCs
For gamers and AI power users, RTX Spark aims to prove that an Arm-based Windows PC can match or beat traditional x86 laptops. NVIDIA claims that RTX Spark can push modern AAA titles at 1440p above 100 frames per second in slim laptops, helped by DLSS 4.5 and Frame Generation for higher effective frame rates. The Blackwell GPU supports full RTX ray tracing, Reflex, and G-SYNC, putting it closer to desktop RTX cards than past integrated solutions. On the AI side, NVIDIA says RTX Spark can achieve 1 petaFLOP of AI performance and run large local models, enabling on-device copilots and media tools without constant cloud access. Adobe is reworking Photoshop and Premiere for Spark, promising up to 2x gains in AI and graphics tasks, while tools like Blender, Blackmagic Design software, CapCut, ComfyUI, and OTOY are also adopting the platform for Arm-based Windows PC workflows.
Windows on Arm, Ecosystem Support, and the Road Ahead
RTX Spark is NVIDIA’s second serious push into Windows PCs and its first direct step into the Windows on Arm market, arriving later than Apple Silicon but with a mature RTX software stack. CUDA, TensorRT, DLSS, Reflex, ray tracing, and G-SYNC all come to Arm-based Windows PC systems, giving developers a consistent GPU ecosystem across desktops, workstations, and Spark laptops. Microsoft is working with NVIDIA on new Windows security primitives and the OpenShell runtime, which adds policy controls so users can decide what local AI agents can access and how data is routed between local models and cloud services. Hardware partners including Acer, ASUS, Dell, Gigabyte, HP, Lenovo, Microsoft, and MSI plan to ship thin 14–16 inch Spark laptops and small form factor desktops this fall. A lower-tier Spark variant with around 400TFLOPS of FP4 GPU performance is also on the roadmap, pointing to a broader Blackwell GPU laptop lineup ahead.
