Daily Tech News Show Everyone Hates NVIDIA’s DLSS 5 - DTNS 5227
18 snips
Mar 17, 2026 Discussion of NVIDIA’s GDC announcements, new inference tools, and the controversy around a lighting tech update. A debate on artistic control versus photorealism in games. Comparison of tri-fold foldables and industry pricing. A startup turning CO2 into textile-grade cellulose. Lab-grown human neurons trained to play Doom and what that means for research.
AI Snips
Chapters
Transcript
Episode notes
NVIDIA Reaffirms Trillion Dollar Data Center Ambitions
- NVIDIA doubled-down on its data center revenue timeline, reaffirming massive growth through 2027 and expanding product lines for training and inference.
- Jensen Huang introduced Rubin for training and Grok 3 LPX inference racks with 256 LPUs and 128 GB SRAM.
Rubin For Training Versus LPUs For Inference
- NVIDIA distinguishes chip lines by use: Rubin chips target model training while LPUs (Grok 3 LPX) serve inference workloads in racks manufactured on 4nm by Samsung.
- Rubin and LPUs will ship in the second half of the year alongside space-optimized variants.
NVIDIA Joins OpenClaw With NemoClaw And OpenShell
- NVIDIA is integrating into the OpenClaw agent ecosystem with NemoClaw, bundling agent tooling and a new open-source security runtime called OpenShell.
- The move aims to add privacy and security features to open agent platforms.