DLSS 5 neural rendering: 150 MB model, real-time at 40% compute
DLSS 5's neural rendering model is only 150 MB, uses relatively little VRAM, and runs in real time at about 40% of the compute cost. It runs on FP8 and modders have gotten it working on 40-series cards.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- OpenAI product lead Tara Seshan discusses persistent AI coworkers
- Krea 3 rumored to add editing, may open weights
- 0.8B local fine-tune matches frontier model on dictation cleanup
- How LLMs Actually Work: A Walkthrough of Transformer Architecture
- MiniMax H3 image editor workflow generates character sheets locally