Headless Linux GeForce driver telemetry & unified VRAM pooling across dual GPUs
Enable P2P NVLink/PCIe unified memory pooling on consumer RTX workstation cards for local LLM inference and expose clean Prometheus metrics in nvidia-smi.
Big Tech Product Roadmap · 5 Active Feature Proposals · 23,840 Priority Votes
NVIDIA is a leading Big Tech platform tracked on CrowdFeatures. Engineers, product managers, and power users use this board to propose new NVIDIA capabilities, upvote high-priority roadmap items, and verify features that have already launched in production.
Enable P2P NVLink/PCIe unified memory pooling on consumer RTX workstation cards for local LLM inference and expose clean Prometheus metrics in nvidia-smi.
The one-hour session limit on the free tier is too restrictive for most games. Extend to at least 2 hours and reduce queue wait times to make cloud gaming actually usable.
GPU prices have skyrocketed beyond reason. A flagship GPU should not cost as much as an entire high-end PC. Return to reasonable pricing for consumer gaming cards.
The proprietary NVIDIA Linux drivers cause endless compatibility issues. Fully open-source the drivers to let the Linux community properly integrate and maintain them.
DLSS currently requires per-game developer integration. Develop a driver-level universal DLSS injection that works on any DirectX 11 and 12 game automatically.