NVIDIA RTX Spark platform gets DLSS 5, Nsight Arm64 tools and NvRTX 5.8

NVIDIA has published a developer guide to RTX Spark, a Windows PC platform for AI development, creation and gaming in slim laptops and compact desktops. It uses an NVIDIA Blackwell RTX GPU and NVIDIA Grace CPU, with up to 128 GB of unified memory, the full CUDA stack and RTX features such as ray tracing and DLSS 5.

On the DLSS side, RTX Spark supports Super Resolution, Ray Reconstruction, Dynamic Multi Frame Generation and 3D-Guided Neural Rendering in supported games. DLSS 5 on RTX Spark starts with NBA 2K27. NVIDIA says an October DLSS 5 model update runs 2x faster than the September version, halves GPU overhead and cuts VRAM use at 4K by 20%, with comparable visual fidelity. It will ship by default in future DLSS 5 games. The latest NVIDIA app beta also adds a faster Performance Mode (Preset C) for Frame Generation, with minor image-quality trade-offs.

For Unreal Engine developers, version 8.9.0 of the DLSS 4.5 plugin adds official Windows on Arm support for Unreal Engine 5.8. That includes native ARM64 binaries and support for building and packaging UE 5.8 applications. The release also updates Streamline and NGX components and fixes UE 5.8 Frame Generation issues. Separately, NVIDIA released NVIDIA RTX Branch of Unreal Engine 5.8 (NvRTX 5.8), which focuses on codebase refactoring, stabilization and compatibility improvements. The source is available on GitHub.

Nsight tools support native Arm64, emulated x64 and Arm64EC applications. Nsight Graphics 2026.3 adds a Windows on Arm installer. From Nsight Systems 2026.5, one installer for all versions of Windows covers native Arm64 and emulated applications. Nsight Aftermath SDK 2026.3 includes a Windows on Arm version for native Arm64 apps, while crash dumps from emulated x64 applications keep working with the x64 SDK. NVIDIA also lists eight RTX Kit technologies with RTX Spark support, including Neural Texture Compression, Neural Shaders, Mega Geometry, RTX Character Rendering and Opacity Micro-Maps.

NVIDIA ACE adds Nemotron 3.5 Bolt E4B 4GB, available through the In-Game Inferencing (IGI) SDK. NVIDIA claims about 40% higher average quality than its prior model on relevant benchmarks and up to about 80% fewer tokens than peers. It runs locally on GeForce RTX and RTX Spark, with multi-turn tool use, configurable reasoning, 64k context and support for 35 languages. NVIDIA points to the SkyrimNET mod for testing it in a game.

MotionBricks, NVIDIA's open-source research on real-time character motion, also has an update. Developers guide motion with target poses, and the system generates the connecting movement at runtime. The update adds SOMA human-motion models alongside the existing G1 robot model, plus downloadable checkpoints, ONNX exports, export tools and a reference implementation built with Newton. It is research code and models, and the Unreal Engine integration from the original demonstration is not included.