loading
loading
NVIDIA RTX Spark is a new Windows PC platform for AI development, creation, and gaming in slim laptops and compact desktops. Powered by an NVIDIA Blackwell RTX GPU and NVIDIA Grace CPU, it combines up to 128 GB of unified memory with the full NVIDIA CUDA stack and RTX technologies including ray tracing and DLSS 5. This post covers the following technologies for game developers building on RTX Spark: - DLSS suite of technologies that boost FPS and improve image quality - NVIDIA ACE updates for responsive, on-device AI characters - RTX Kit SDKs for neural rendering, geometry, and character rendering - NVIDIA Nsight tools for optimizing and debugging native Arm64 and x64 applications - NVIDIA RTX Branch of Unreal Engine 5.8 for integrating RTX technologies into your Unreal Engine projects - MotionBricks update for human-character motion DLSS on RTX Spark RTX Spark benefits from the full suite of DLSS technologies in supported games, including Super Resolution, Ray Reconstruction, Dynamic Multi Frame Generation, and 3D-Guided Neural Rendering. DLSS Super Resolution reconstructs frames from a lower resolution input to improve performance. DLSS Ray Reconstruction generates additional pixels to enhance ray-traced scenes. DLSS Dynamic Multi Frame Generation can generate up to 5 additional frames per rendered frame, automatically adjusting the multiplier to maximize smoothness. The latest NVIDIA app beta also brings a faster Performance Mode (Preset C) for Frame Generation, with minor trade-offs in image quality. DLSS 5 brings 3D-Guided Neural Rendering to RTX Spark DLSS 5 with 3D-Guided Neural Rendering is supported on NVIDIA RTX Spark, starting with NBA 2K27. An October model update runs 2x faster than the September version and halves GPU overhead while maintaining comparable visual fidelity. It also reduces VRAM use at 4K by 20%. The model will ship by default in future DLSS 5 games. For developers, DLSS 5 adds a final stage to the graphics pipeline that enhances lighting and material detail. The engine-rendered frame, including artist-designed geometry, textures, and lighting buffers, remains the foundation, while developer controls determine what can change. Avalanche Software has taken advantage of DLSS 5’s SDK controls to tune the integration to maintain the artistic intent and style of the game. Read the full DLSS 5 update here. DLSS 4.5 Unreal Engine plugin The updated NVIDIA DLSS 4.5 plugin for Unreal Engine gives developers powerful tools for delivering sharper images and smoother gameplay. Its second-generation transformer model for Super Resolution improves image stability, anti-aliasing, and visual clarity, while Ray Reconstruction enhances ray-traced image quality. On supported GeForce RTX 50 Series and RTX PRO Blackwell GPUs, Dynamic Multi Frame Generation adapts the frame-generation multiplier as workloads change, with 6x Multi Frame Generation producing up to five additional frames per rendered frame. A key addition in plugin version 8.9.0 is official Windows on Arm (WoA) support for Unreal Engine 5.8. Native ARM64 binaries and support for building and packaging UE 5.8 applications give developers a supported path to bring DLSS-enabled projects to RTX Spark systems. Alongside updated Streamline and NGX components and fixes for UE 5.8 Frame Generation issues, this release expands the plugin’s platform reach and helps developers prepare their applications for RTX Spark. NVIDIA ACE introduces Nemotron 3.5 Bolt for on-device AI characters NVIDIA ACE is our technology suite for creating in-game AI agents that deliver next-generation gaming experiences, and we’ve shared two exciting showcases this year: - PUBG rolled out AI teammates in a public beta powered by our first-generation Nemotron small language model (SLM). It helped new players land their first chicken dinners and generated viral clips across social media. - Aniimo debuted our more efficient second-generation Nemotron SLM, turning standard in-game pets into responsive AI companions that react to your voice. Players can check in on their pets’ feelings or teach them new tricks, which will bring a new level of immersion when it releases in early 2027. More intelligence, less memory Nemotron 3.5 Bolt E4B 4GB is now available. This model offers ~40% higher average quality against our prior model on relevant benchmarks and outputs up to ~80% fewer tokens vs peers. It is purpose-built to power richer, faster in-game character interactions, with strong conversational capability, autonomous multi-turn tool use, instruction adherence, and grounding to game state, all running locally on GeForce RTX and RTX Spark. Engineered for maximum intelligence and speed per GB of memory, Nemotron 3.5 Bolt E4B 4GB enables responsive local experiences while preserving headroom for graphics rendering and local speech models. Key features include: - Multi-turn tool use: Call application APIs to carry out tasks across a conversation, with integration guidance for asynchronous tool calls - Conversational mode: Toggle to encourage concise, speech-ready responses for natural back-and-forth interactions. - Configurable reasoning: Enable reasoning for complex requests or disable it for faster responses, with integration options to cap thinking to meet latency targets. - 64k context: Accommodate conversation history, instructions, tool definitions, retrieved information and game state. - Multilingual capability: Support for 35 languages, building on a base model pre-trained on more than 140 languages. Get started with NVIDIA ACE Nemotron 3.5 Bolt E4B 4GB is available to developers via the In-Game Inferencing (IGI) SDK. To test it in a game, check out SkyrimNET, an AI NPC modding platform by MinLL for The Elder Scrolls V: Skyrim. SkyrimNET transforms every NPC into an unscripted, interactive character with personal memory, voice, and autonomy, enabling dynamic conversations and context-aware quests generated on the fly. The platform supports both cloud and local inference across multiple language models. - Download the SkyrimNET Plugin: GitHub Repository - Download the ARM-Supported Loader: Mod Organizer Release NVIDIA RTX Kit brings neural rendering SDKs to RTX Spark NVIDIA RTX Kit brings neural rendering, complex geometry, and realistic character rendering to game development. Eight RTX Kit technologies are published with support for RTX Spark: - RTX Neural Texture Compression (NTC): Uses AI to shrink texture storage while maintaining visual fidelity comparable to traditional block compression. - RTX Texture Filtering: Uses stochastic texture sampling after shading to reduce artifacts and improve image quality, including for difficult volumes. - RTX Neural Shaders: Lets developers train neural networks and run them within shaders for new asset compression and approximation techniques. - RTX Mega Geometry: Speeds up bounding volume hierarchy (BVH) construction for cluster-based geometry, helping developers ray trace more triangles and improve performance in complex scenes. - NVIDIA Real-Time Denoisers (NRD): Provides denoisers for diffuse and specular lighting and shadows, producing cleaner images from low ray-per-pixel inputs. - RTX Character Rendering: Provides tools for rendering strand-based hair and skin with path tracing, including subsurface scattering and Blackwell-accelerated hair primitives. - RTX Global Illumination (RTXGI): Provides several techniques for efficient indirect lighting for ray traced and path traced applications. - Opacity Micro-Maps (OMM): Used to accelerate ray tracing of high detail geometry by efficiently encoding the opacity states of micro-triangles and micro meshes. Optimizing and debugging RTX Spark applications NVIDIA Nsight tools are ready for RTX Spark from day one, supporting graphics developer workflows across native Arm64, emulated x64, and Arm64EC applications. Nsight Graphics lets developers capture, inspect, and profile individual frames, examining every draw call, shader, and resource to diagnose rendering bugs and tune GPU performance. Nsight Graphics release 2026.3 introduces a new Windows on Arm installer for RTX Spark systems, supporting native Arm64 target applications as well as emulated x64 and Arm64EC targets. Nsight Systems delivers a system-wide view of application performance, correlating CPU threads, GPU workloads, API calls, and memory activity on a single timeline so developers can pinpoint bottlenecks and balance work across the platform. Starting with Nsight Systems 2026.5, a single installer for all versions of Windows supports both native Arm64 and emulated applications. Nsight Aftermath helps teams resolve GPU crashes and hangs by generating detailed crash dumps that show what the GPU was doing at the moment of failure, both during development and in the field. Nsight Aftermath SDK 2026.3 includes a Windows on Arm version for integrating into native Arm64 applications. Crash dumps from emulated x64 applications will continue to function as expected with the x64 version of the SDK. - Download Nsight Graphics for RTX Spark here - Download Nsight Systems for RTX Spark here - Download Nsight Aftermath SDK for RTX Spark here NVIDIA RTX Branch of Unreal Engine 5.8 The release of NVIDIA RTX Branch of Unreal Engine 5.8 (NvRTX 5.8) gives developers an updated foundation for bringing advanced ray tracing and neural graphics to UE 5.8 projects. This update focuses on codebase refactoring, stabilization, and compatibility improvements. Developers can access the branch source through GitHub to integrate and customize NVIDIA RTX technologies for their games and applications. Simplify runtime character animation with MotionBricks MotionBricks is open-source research for generating character motion in real time. Rather than hand-authoring every transition and wiring it into an increasingly complex animation graph, developers guide motion through target poses. MotionBricks generates the connecting movement at runtime and adapts as player input or game event changes. Paper GitHub The model supports control over speed, direction, and movement style, along with pose-guided interactions such as sitting and picking up objects. These capabilities let developers build varied character behaviors through a common pose-based interface. The original research demonstration shows these behaviors working together in Unreal Engine. Research demonstrations The update builds on this work with: - New SOMA human-motion models, alongside the existing G1 robot model. - Downloadable checkpoints, ONNX exports, and export tools for developers building their own integrations. - An interactive reference implementation built with Newton, demonstrating pose-guided human and robot motion. MotionBricks is provided as research code and models. The Unreal Engine integration shown in the research demonstration is not included in this release. Resources for game developers Check out the full list of game developer resources and stay up to date with the latest NVIDIA game development news: - Subscribe to our newsletter (select gaming as your industry) - Follow us on social: X, LinkedIn, Facebook, and YouTube - Join our Discord community
document fetched 2026-10-28 02:51 · canonical https://developer.nvidia.com/blog/nvidia-rtx-spark-powers-neural-rendering-and-ai-characters-in-games