{"as_of":"2026-10-12","license":"CC-BY-4.0","license_url":"https://creativecommons.org/licenses/by/4.0/","cite_as":"The Local Bench (thelocalbench.com), data as of 2026-10-12, CC BY 4.0","source_url":"https://thelocalbench.com","hardware":{"id":"rtx-3090-24g-used","name":"GeForce RTX 3090 24 GB (used)","vendor":"NVIDIA","class":"Discrete GPU","kind":"gpu","as_of":"2026-10-12","summary":"The reference card of the hobby: 24 GB and 936 GB/s for about the price of a new mid-range GPU.","memory_gb":{"value":24,"flag":"sourced","unit":"GB","source":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","note":"Vendor spec page, checked 2026-10-12."},"bandwidth_gbs":{"value":936,"flag":"sourced","unit":"GB/s","source":"https://en.wikipedia.org/wiki/GeForce_RTX_30_series","note":"NVIDIA's GeForce spec pages list bus width but not bandwidth; figure from the cited Wikipedia spec table (bus width cross-checked on nvidia.com, 2026-10-12)."},"fp16_tflops":{"value":71,"flag":"sample","unit":"TFLOPS","note":"Sample FP16 dense compute; pending vendor-sheet citation."},"host_ram_gb":{"value":64,"flag":"sample","unit":"GB","note":"Assumed host RAM for offload."},"host_bandwidth_gbs":{"value":64,"flag":"sample","unit":"GB/s"},"load_watts":{"value":420,"flag":"sample","unit":"W","note":"Sample wall power under load."},"tdp_w":{"value":350,"flag":"sourced","unit":"W","source":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","note":"Graphics Card Power (Founders Edition), nvidia.com."},"price_usd":{"value":1200,"flag":"sourced","unit":"USD","source":"https://www.getpcparts.com/market-prices/gpu-graphics-cards/rtx-3090","note":"eBay sold-listing median ≈ US$1,199–1,206, Oct 2026, via the getpcparts aggregator (eBay itself blocks automated access). Checked 2026-10-12."},"price_cad":{"value":1712.52,"flag":"estimate","unit":"CAD","source":"https://www.bankofcanada.ca/valet/observations/FXUSDCAD/json?recent=1","note":"USD price × 1.4271 (Bank of Canada USD/CAD, 2026-10-09); no Canadian retail listing captured."}},"runs":[{"model":"gemma-3-27b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp (CUDA)","test":"tg128 @ 8k depth","tok_s":30.2,"min":29.8,"max":30.6,"repeats":5,"flag":"sample","date":"2026-10-12","note":"Placeholder figure from the approved mockup; not a measurement."},{"model":"gemma-3-27b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp","test":"tg128 @ depth 0","depth":0,"tok_s":30.2,"flag":"sample","date":"2026-10-12","note":"Placeholder depth curve from the mockup."},{"model":"gemma-3-27b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp","test":"tg128 @ depth 4096","depth":4096,"tok_s":29.1,"flag":"sample","date":"2026-10-12","note":"Placeholder depth curve from the mockup."},{"model":"gemma-3-27b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp","test":"tg128 @ depth 16384","depth":16384,"tok_s":26,"flag":"sample","date":"2026-10-12","note":"Placeholder depth curve from the mockup."},{"model":"gemma-3-27b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp","test":"tg128 @ depth 32768","depth":32768,"tok_s":23.1,"flag":"sample","date":"2026-10-12","note":"Placeholder depth curve from the mockup."},{"model":"gpt-oss-20b","hardware":"rtx-3090-24g-used","quant":"MXFP4","runtime":"llama.cpp (CUDA)","test":"tg128 @ 8k depth","tok_s":128.4,"flag":"sample","date":"2026-10-12","note":"Placeholder figure from the approved mockup; not a measurement."},{"model":"qwen3-30b-a3b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp (CUDA)","test":"tg128 @ 8k depth","tok_s":112,"flag":"sample","date":"2026-10-12","note":"Placeholder figure from the approved mockup; not a measurement."},{"model":"qwen3-14b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp (CUDA)","test":"tg128 @ 8k depth","tok_s":52.3,"flag":"sample","date":"2026-10-12","note":"Placeholder figure from the approved mockup; not a measurement."},{"model":"mistral-small-3.2-24b","hardware":"rtx-3090-24g-used","quant":"Q4_K_M","runtime":"llama.cpp (CUDA)","test":"tg128 @ 8k depth","tok_s":34.1,"flag":"sample","date":"2026-10-12","note":"Placeholder figure from the approved mockup; not a measurement."},{"model":"llama-2-7b","hardware":"rtx-3090-24g-used","quant":"Q4_0","runtime":"llama.cpp (CUDA)","test":"tg128","tok_s":158.16,"flag":"sourced","date":"2026-10-12","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","note":"Llama 2 7B Q4_0, llama-bench tg128, full GPU offload, flash attention off, single GPU. Community scoreboard entry; commit and submitter in the source table.","min":157.95,"max":158.37}]}