NVIDIA has announced RTX Spark, a Windows PC platform aimed at slim laptops and compact desktops that combines AI development, content creation and gaming features. RTX Spark systems pair an NVIDIA Blackwell RTX GPU with an NVIDIA Grace CPU and offer up to 128 GB of unified memory. The platform includes the full NVIDIA CUDA stack and RTX technologies such as ray tracing and DLSS 5.
What developers can use
NVIDIA outlined the main building blocks available to game developers targeting RTX Spark:
- DLSS: a suite of features to increase frame rates and improve image quality (Super Resolution, Ray Reconstruction, Dynamic Multi Frame Generation, 3D-Guided Neural Rendering).
- NVIDIA ACE: updates for on-device, responsive AI characters, including Nemotron language models.
- RTX Kit: SDKs for neural rendering, complex geometry and character rendering.
- NVIDIA Nsight: optimization and debugging tools for native Arm64 and x64 (emulated) applications.
- NVIDIA RTX Branch of Unreal Engine 5.8: source branch and integration for bringing RTX tech into UE 5.8 projects.
- MotionBricks: research code for pose-guided, real-time character motion generation.
DLSS on RTX Spark
RTX Spark supports the full DLSS feature set in supported games. Key elements include:
- DLSS Super Resolution, which reconstructs frames from a lower-resolution input to improve performance.
- DLSS Ray Reconstruction, which generates additional pixels to enhance ray-traced scenes.
- DLSS Dynamic Multi Frame Generation, which can create up to five additional frames per rendered frame and automatically adjust the multiplier to maximize smoothness.
A new NVIDIA app beta also introduces a faster Performance Mode (Preset C) for Frame Generation, with modest image-quality trade-offs.
DLSS 5 and 3D-Guided Neural Rendering
DLSS 5 with 3D-Guided Neural Rendering is supported on RTX Spark starting with NBA 2K27. An October model update runs twice as fast as the September version and halves GPU overhead while maintaining comparable visual fidelity; it also reduces 4K VRAM usage by about 20%. NVIDIA says this 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 while leaving the engine-rendered frame—artist-authored geometry, textures and lighting buffers—as the foundation. Developers retain controls over what can change; Avalanche Software used the SDK controls to tune integration so the game’s artistic intent is preserved.
The DLSS 4.5 Unreal Engine plugin’s second-generation transformer improves Super Resolution stability, anti-aliasing and clarity, while Ray Reconstruction raises ray-traced image quality. Plugin version 8.9.0 adds official Windows on Arm (WoA) support for Unreal Engine 5.8, delivering native ARM64 binaries and build/packaging support to help developers bring DLSS-enabled projects to RTX Spark systems.
NVIDIA ACE and Nemotron 3.5 Bolt E4B 4GB
NVIDIA ACE is the company’s technology suite for in-game AI agents. NVIDIA highlighted two recent showcases: PUBG’s AI teammates in public beta powered by the first-generation Nemotron small language model, and Aniimo’s demo of a second-generation Nemotron that turns in-game pets into voice-responsive companions (planned for early 2027).
Now available is Nemotron 3.5 Bolt E4B 4GB. According to NVIDIA’s data, this model delivers roughly 40% higher average quality versus their prior model on relevant benchmarks, and produces up to about 80% fewer tokens compared to peers. Nemotron 3.5 Bolt is designed to run locally on GeForce RTX and RTX Spark systems to power richer, faster in-game character interactions with multi-turn tool use, instruction following and grounding to game state.
Key features:
- Multi-turn tool use: call application APIs across a conversation with guidance for asynchronous tool calls.
- Conversational mode: encourage concise, speech-ready responses for natural back-and-forth.
- Configurable reasoning: enable or limit reasoning to meet latency targets.
- 64k context: hold conversation history, instructions, tool definitions, retrieved info and game state.
- Multilingual capability: support for 35 languages, built on a base pre-trained on more than 140 languages.
Nemotron 3.5 Bolt E4B 4GB is available to developers via the In-Game Inferencing (IGI) SDK. NVIDIA points developers to SkyrimNET, a modding platform by MinLL for The Elder Scrolls V: Skyrim, as an example integration: SkyrimNET converts every NPC into unscripted interactive characters with memory, voice and autonomy, supporting both local and cloud inference across multiple language models.
RTX Kit: neural-rendering SDKs for RTX Spark
NVIDIA RTX Kit provides neural rendering, complex geometry tools and realistic character rendering. Eight RTX Kit technologies published with RTX Spark support are:
- RTX Neural Texture Compression (NTC): AI-driven texture compression with visual fidelity comparable to traditional block compression.
- RTX Texture Filtering: stochastic texture sampling after shading to reduce artifacts and improve quality, including for difficult volumes.
- RTX Neural Shaders: train neural networks and run them inside shaders for new asset-compression and approximation techniques.
- RTX Mega Geometry: speeds BVH construction for cluster-based geometry so developers can ray trace more triangles and improve performance in complex scenes.
- NVIDIA Real-Time Denoisers (NRD): denoisers for diffuse and specular lighting and shadows.
- RTX Character Rendering: tools for strand-based hair and skin rendering with path tracing, including subsurface scattering and Blackwell-accelerated hair primitives.
- RTX Global Illumination (RTXGI): multiple techniques for efficient indirect lighting in ray-traced and path-traced applications.
- Opacity Micro-Maps (OMM): encode opacity states of micro-triangles and micromeshes to accelerate ray tracing of high-detail geometry.
Optimizing and debugging RTX Spark applications
NVIDIA Nsight tools support RTX Spark from day one and cover native Arm64, emulated x64 and Arm64EC application workflows:
- Nsight Graphics: capture and inspect frames, examine draw calls, shaders and resources for rendering bugs and GPU tuning. Release 2026.3 adds a new Windows on Arm installer for RTX Spark systems supporting native Arm64 targets as well as emulated x64 and Arm64EC.
- Nsight Systems: provides a system-wide performance timeline correlating CPU threads, GPU workloads, API calls and memory activity. From Nsight Systems 2026.5, a single installer supports native Arm64 and emulated applications across Windows versions.
- Nsight Aftermath: helps resolve GPU crashes and hangs by producing detailed crash dumps. The Aftermath SDK 2026.3 includes a Windows on Arm version for integrating into native Arm64 applications; crash dumps from emulated x64 applications continue to work with the x64 SDK.
NVIDIA RTX Branch of Unreal Engine 5.8
The NVIDIA RTX Branch of Unreal Engine 5.8 (NvRTX 5.8) gives developers an updated foundation for integrating advanced ray tracing and neural graphics into UE 5.8 projects. The branch focuses on refactoring, stabilization and compatibility improvements, and the source is available on GitHub for developers to access and adapt.
MotionBricks: simplifying runtime character animation
MotionBricks is open research for generating character motion in real time. Instead of hand-authoring every transition in a growing animation graph, developers specify target poses and MotionBricks generates the connecting movements at runtime, adapting to player input and game events.
The update expands the work with:
- New SOMA human-motion models alongside the existing G1 robot model.
- Downloadable checkpoints, ONNX exports and export tools for developers.
- 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 demonstration is not included in this release.
Developer resources and next steps
NVIDIA links a full list of game developer resources and invites developers to subscribe to its newsletter (selecting gaming as the industry) and follow its social channels. The company positions RTX Spark as a combined hardware and software platform to deliver AI-driven rendering and local, responsive in-game characters on compact Windows PCs while providing developers access to NVIDIA’s SDKs and tools.
