topic Re:Technical Proposal: Multi-Step Texture-Driven Reconstruction for Low-Latency Upscaling in Graphics

Technical Proposal: Multi-Step Texture-Driven Reconstruction for Low-Latency Upscaling

JAX3KTOJA — Thu, 26 Mar 2026 02:58:49 GMT

To the Graphics Engineering Team,
I propose a concept for Hierarchical Texture-Driven Scaling to improve performance on integrated and entry-level GPUs without the input lag of current Frame Generation.
The Concept:
Instead of a single-pass AI upscale (e.g., 540p to 1080p), the system uses a multi-step recursive process combined with Source Texture Referencing:Adaptive Steps:
Performance (1-step): 540p → 1080p. Near-zero latency for competitive gaming.
Balanced (2-steps): 540p → 720p → 1080p. Eliminates shimmering and artifacts.
Ultra (3+ steps): Gradual 540p → 720p → 1080p → 1440p for AAA titles.Key Advantages:
Zero "Jelly" Feeling: Works within a single frame; no "future frame" analysis required.
Hardware Agnostic: Can run on standard shaders (DP4a) or matrix cores (XMX/Tensor).
Deterministic Detail: Accurate to the developer’s original vision by using real game assets.
I believe this "reconstructive" approach is the future of mobile and budget gaming. I would appreciate your feedback.
Best regards,
JAX3KTOJA

Re:Technical Proposal: Multi-Step Texture-Driven Reconstruction for Low-Latency Upscaling

Nikhil_Intel — Sat, 28 Mar 2026 15:49:23 GMT

Hello JAX3KTOJA,

Thank you for posting on Intel Community Forum.

Thank you for sharing your detailed and thoughtful proposal with the Graphics Engineering team. We appreciate the time and effort you’ve put into outlining this concept.

Your approach to multi-step scaling combined with source texture referencing is interesting, particularly the focus on reducing latency while improving visual stability across different performance tiers. The emphasis on hardware-agnostic implementation and deterministic detail is also well noted.

At the same time, practical considerations such as implementation complexity, memory bandwidth usage, and integration with existing rendering pipelines would need to be carefully evaluated to determine feasibility and real-world performance benefits.

We value your input and have shared your proposal. While we may not be able to provide detailed follow-up on this concept, your feedback is important in helping us explore future improvements.

Thank you again for your contribution.