In brief
- Resource Management in the NVIDIA App offers controls for allocating GPU resources between apps and background agents.
- Hardware-accelerated GPU scheduling must be enabled in Windows to use the setting.
- Nemotron 3.5 Transcribe supports speech transcription, speaker identification and translation.
- According to NVIDIA’s official news, Phantom Blade Zero launched on 28 October with DLSS 4.5 and ray tracing.
NVIDIA released a software update for RTX Spark computers on 28 October. Its main addition is Resource Management in the NVIDIA App, which lets users allocate GPU resources between foreground apps and background AI agents. The package also includes new AI models for local use and RTX gaming updates.
GPU resource prioritization
Resource Management lets users determine how an app and background agents use GPU resources. Users can allocate more resources to the app they are using in the foreground or prioritize helping agents finish their tasks faster. NVIDIA presents the feature as a way to manage resource allocation when different types of workloads run on the same system; it has not announced a specific performance improvement figure.
The controls are in the NVIDIA App under System, then Advanced. Hardware-accelerated GPU scheduling must also be enabled in Windows Settings to use the feature. This means checking the settings in both the app and the operating system before adding it to a workflow.
NVIDIA describes RTX Spark as a system that combines a Grace CPU and a Blackwell RTX GPU in a single superchip. According to the source, the platform offers up to 128 GB of unified memory and up to 1 petaflop of FP4 AI performance. In scenarios where local model inference and other GPU workloads share the same pool of resources, this design makes resource controls more visible.
Local speech and image-generation models
Nemotron 3.5 Transcribe, included in the update, supports audio transcription, speaker identification and translation. According to NVIDIA, the model covers 40 languages, including seven languages from the Indian subcontinent. It can answer questions about audio content and can be used with natural-language prompts. The Granary v2 pipeline and fine-tuning recipes are also included.
Liquid AI’s LFM3 family includes text, image, audio, multimodal and decision models; the source says the family has open-weight support. Black Forest Labs’ FLUX 3 Image Klein is designed for local image generation and editing. NVIDIA says the optimized 5.6-billion-parameter model can run with 8 GB of VRAM, and that an NVIDIA NeMo AutoModel recipe is available for fine-tuning on a laptop. The source describes Nemotron and LFM3 as open-weight; it does not use the same characterization for FLUX 3 Image Klein.
RTX games and Adobe showcase
The gaming side of the update package also mentions RTX Remix 1.6 and a ray-traced path tracing beta for Minecraft Java Edition 1.21.1. NVIDIA’s official GeForce news says Phantom Blade Zero launched on 28 October with DLSS 4.5 and path tracing support. This date differs from the 29 October date in the update news; the official NVIDIA source is used here for the game’s release date.
NVIDIA and Adobe plan to showcase beta versions of Photoshop and Premiere with RTX Spark support at Adobe MAX, taking place in Miami Beach from 10 to 12 November. The sources do not specify when these beta versions will become generally available.
Impact on architecture and visualization workflows
Architecture and interior design offices could consider using Nemotron 3.5 Transcribe for tasks such as transcribing meeting recordings, identifying speakers or translating speech. Visualization teams could also explore FLUX 3 Image Klein’s local image-generation and editing capabilities for concept work. However, the sources provide no test results on compatibility with specific architectural software or rendering performance.
Resource Management could help prioritize workloads in workflows where an app and background agents share GPU resources. Before adopting it, users should check hardware requirements, compatibility with the software they plan to use and each model’s licensing terms separately. The sources describe the RTX Spark update’s features but do not assess availability or pricing in Turkey.
Sources
1 sourceSource texts are not republished; short quotes are marked, everything else is our own summary and commentary.
Local speech processing and image-generation options offer opportunities for project teams to explore with meeting recordings and concept work. GPU resource allocation controls may also be useful when agents and interactive apps run on the same system. However, the sources include no comparative tests for architectural software or rendering workflows, so a direct productivity gain should not be assumed.
For offices in Turkey, it is important to consider hardware requirements, supply conditions and software compatibility together when making a purchase decision. In addition, open-weight status alone does not explain licensing terms. Testing the models with real project files and the software in use provides a more solid basis for a decision before adoption.
Frequently asked questions
Where can I find the GPU resource settings on RTX Spark?
Resource Management is in the NVIDIA App under System, then Advanced. Hardware-accelerated GPU scheduling must be enabled in Windows to use it.
What can Nemotron 3.5 Transcribe do?
The model supports audio transcription, speaker identification and translation. NVIDIA says it covers 40 languages and can answer questions about audio content.
What is the stated minimum VRAM requirement for FLUX 3 Image Klein?
NVIDIA says the optimized version can run on systems with 8 GB of VRAM.
