Racks & Watts
We may earn a commission when you buy through links on this site, at no extra cost to you. Our recommendations are based on genuine research, not commission. Learn more.

Lab notes · September 25, 2026

GPU bench notes, September 2026

The raw numbers our articles cite. Read the method first; repeat it before you trust us.

Bench under test

A desktop board with two NVIDIA cards installed together: a GeForce RTX 3060 Ti 8 GB (PCIe x4 slot) and a GeForce RTX 2060 6 GB (PCIe x16 slot). Both cards run under a software power cap set with nvidia-smi -pl: 150 W on the 3060 Ti and 125 W on the 2060, the 2060's minimum. Roughly 3 GB of the 2060's memory is held by a speech-to-text model that is always resident, so Ollama had about 3 GB of the 2060 and all 8 GB of the 3060 Ti to work with. Ollama 0.34.2 in Docker, models from the Ollama library, Q4_K_M quantisation. Driver telemetry sampled with nvidia-smi --query-gpu during generation. Idle GPU draw is the driver's reported board power with no model loaded, not wall power.

Method

One request to /api/generate with stream:false and num_predict:320, a 40-word prompt, after one warm-up run of the same model. Tokens per second is Ollama's own eval_count / eval_duration. Prompt processing is prompt_eval_count / prompt_eval_duration. Power and temperature were sampled once mid-generation. Single runs, so treat ±5% as noise.

Results

Model (Q4_K_M)Weights on diskWhere it ranGenerationPromptGPU power during runGPU temp
qwen2.5:7b4.4 GBRTX 3060 Ti, fully in VRAM (4.6 GB used)76.3 tok/s3,847 tok/s149.6 W (at the 150 W cap)51 °C
llama3.1:8b4.6 GBRTX 3060 Ti, fully in VRAM (5.0 GB used)73.9 tok/s515 tok/s (cold prompt cache)149.9 W51 °C
qwen2.5:14b8.4 GBSplit: 6.7 GB on the 3060 Ti + 1.7 GB on the 2060; ~0.5 GB spilled to CPU24.9 tok/s1,293 tok/s117.7 W + 39.4 W49 °C / 43 °C

Idle board power with no model loaded: RTX 3060 Ti 16.8 W, RTX 2060 6.6 W (driver-reported). Idle temperatures 40 °C and 37 °C in a case with poor airflow. Earlier in the week, with the 2060 idle and both cards available, the same 14B model ran fully in VRAM across the two cards at about 38 tok/s; the 25 tok/s above is what a partial CPU spill costs.

What these numbers say

Measured September 25, 2026. We re-run these when the hardware or Ollama version changes.