The RTX 4090D 24GB (China) is a NVIDIA accelerator with 24 GB GDDR6X and about 1008 GB/s peak memory bandwidth. AI Hashrate estimates local LLM decode (batch 1) at 8K context: fit counts weights + KV cache + 1 GB runtime overhead — not weights alone. At Q4, about 33 curated models fit fully on this card; at FP16, about 17 fit. Top Q4 speeds: Qwen3-0.6B ≈ 801.8 tok/s (estimated, fits); gemma-3-1b-it ≈ 641.5 tok/s (estimated, fits); Llama-3.2-1B-Instruct ≈ 534.5 tok/s (estimated, fits); Qwen3-1.7B ≈ 320.7 tok/s (estimated, fits); Qwen3-30B-A3B ≈ 213.8 tok/s (estimated, fits). Relative ranking is more reliable than absolute tok/s. See methodology for the bandwidth formula and measured-anchor policy. Methodology.