Hardware Corner RTX 3090 Ti local LLM benchmark page

Per-card GPU LLM benchmark page. Unlike the sibling RTX 5070 Ti page, this one does NOT print 'Updated Benchmarks: March 2026' — the page title says 2026 and the index that links it is the March 2026 sweep, but the date is not on the page, so published is left null. Same lab method as the already-ingested RTX 5090/PRO 5000 pages: llama.cpp llama-bench, Ubuntu 24.04, CUDA 12.8, Q4_K GGUF (gpt-oss 20B labeled MXFP4), per-request batch 1, context-depth sweep 4k-128k. Full tables transcribed: Qwen3 8B, Qwen3 14B, gpt-oss 20B, Qwen3 30B A3B, Qwen3 32B. NOT the same artifact as hardware-corner-gpu-ranking-local-llm (the Dec 2025 ranking), which prints different numbers for the same GPU and the same model at the same context depth — both are kept, see the contested records.