gb300 - 技术专题

TokenSpeed is a speed-of-light LLM inference engine.

Receipts-first serving recipes and measurements for two NVIDIA DGX Station GB300s

Emulation handheld GB300 original firmware with original mapped folder structure.

Reproducible MiniMax-H3 benchmarks on NVIDIA GB300: SGLang scaling, host eBPF CUDA tracing, Grace-GPU PMU traffic, energy, and Nsight comparison.

🥧 Open source re-implementation of the GameMaker: Studio runner, ported to no-GPU handhelds like the GB300V2 / SF2000 (Ingenic JZ4775, 320x240). Runs Undertale via a software RGB565 renderer.

A plain-spoken, interactive course on NVIDIA's Blackwell architecture — the chips, engines, and rack-scale systems built for the age of AI reasoning.

Close-to-metal C/CUDA lab for RL inference fast paths: persistent GPU workers, hugepage KV arenas, cacheline-aware command rings, and async reward handoff. Goal: remove page faults, malloc/free, scheduler wakeups, CPU round-trips, and KV migration from the per-token path.

StepMania 3.9 rhythm game engine for the Data Frog GB300 and SF2000 handhelds.

CajuOS GB300 v0.3.3-beta: experimental Portuguese SD-card overlay for the DataFrog GB300

Dell PowerEdge XE9712 · GB300 NVL72 interactive 3D rack explorer with troubleshooting guide