RTX 4090

Headlines · 2

1. Oct 20
   DLSS 5 foveated-rendering mods claim near-double performance, unverified

   Tom's Hardware reports that two community projects now apply foveated rendering to DLSS 5 Neural Rendering: maximedgoulet's Foveated DLSS5, a ReShade add-on, and ClarkCheekyKent's CheekyFoveatedDLSS, a standalone integration. Rather than changing image resolution, the mods vary the level of detail in DLSS 5 Neural Rendering, keeping the centre of the frame at full quality while reducing detail toward the edges; users can set the centre region's width and height independently and adjust the fall-off rate. The reported speed-up rests on a self-reported Reddit measurement of 48 to 80 FPS in The Witcher 3 at 7680x2160 on an RTX 4090, which Tom's Hardware cites without independent verification.

   Tom’s Hardware · 🔥 0
2. Oct 14
   Intern-Decision paper details 0.8B–4B multimodal decision models

   A new arXiv paper by Tong Zhu and co-authors presents Intern-Decision, a family of 0.8B, 2B and 4B multimodal decision models that return structured choices and candidate probabilities from text and optional images. The method places one decision placeholder per field in a complete assistant JSON skeleton and evaluates all fields in a single causal forward pass, with ground-truth symbols used only in supervision. The authors report that the 4B model reaches 90.02% unweighted average accuracy across seven evaluation suites versus 88.74% for Jev v1.13.0, and a mean latency of 44.16 ms for a fixed 289-token, three-field request on one RTX 4090; they also describe a confidence-based handoff study in which routing 42 of 111 Jevbench-Hard items to Atria-Dawn-Preview raised accuracy from 82/111 to 97/111.

   Hugging Face · Papers · 🔥 0

Experience and discussion from the community

Share my RTX 4090 experienceAsk about RTX 4090