Hardware Corner RTX 5070 local LLM benchmark page (March 2026)

Fields

id

sources/hardware-corner-rtx-5070-llm-benchmark-page-2026-03

URL

https://www.hardware-corner.net/gpu-llm-benchmarks/rtx-5070/

Publisher

Hardware Corner

Kind

benchmark

Published

2026-03

Accessed

2026-10-16

notes

Per-SKU llama.cpp table, page header 'Updated Benchmarks: March 2026'. Single RTX 5070 (12 GB GDDR7, 672 GB/s stated). Tables: Qwen3 8B Q4_K at 4k/16k/32k (PP 3,487.7 / 1,600.8 / 898.8; TG 85.8 / 59.1 / 43.6) and Qwen3 14B Q4_K at 4k/16k (PP 2,144.2 / 1,315.2; TG 54.2 / 40.6). Header '40.6 t/s generation and 1,315 t/s prompt processing, 14B @ 4k' matches the 16k cells, so the header mislabels the context depth; the per-cell table is what was recorded. No build hash.

Notes

Per-SKU llama.cpp table, page header 'Updated Benchmarks: March 2026'. Single RTX 5070 (12 GB GDDR7, 672 GB/s stated). Tables: Qwen3 8B Q4_K at 4k/16k/32k (PP 3,487.7 / 1,600.8 / 898.8; TG 85.8 / 59.1 / 43.6) and Qwen3 14B Q4_K at 4k/16k (PP 2,144.2 / 1,315.2; TG 54.2 / 40.6). Header '40.6 t/s generation and 1,315 t/s prompt processing, 14B @ 4k' matches the 16k cells, so the header mislabels the context depth; the per-cell table is what was recorded. No build hash.

Cross-references

This record references no other records.