More
Best rig by budgetRent before you buySubmit a buildNewsletterFree datasetAdvertise
Submit build

Builds / 2x RTX 3060

2x RTX 3060 12GB (24 GB): Qwen3.8 27B Q4_K_M at 18.1 tok/s, ~950 tok/s prefill on a 9.5K prompt (llama.cpp CUDA)

Last updated:

Source: discourse.homelabbrisbane.com.au ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
GPU(s)
2x RTX 3060
Amazon · eBay (used) · Newegg
VRAM / unified memory (GB)
24
CPU
not stated
Motherboard
not stated
RAM
not stated
PSU
not stated
Other
llama.cpp CUDA build (commit 19ff817 of 27 Oct 2026, then master 8ca4060 of 28 Oct 2026); all layers across both cards (--n-gpu-layers 999); ~20.6 GB VRAM used
Reported total cost
not stated
Power
not stated

Measured speed (generation tokens/sec)

ModelQuantEnginetok/sPrompt processing (tok/s)Details
Qwen3.8 27BQ4_K_M (GGUF)llama.cpp (CUDA, 2 GPUs)18.1949 @ 9.5K-token prompt
est. TTFT @8k ≈ 8.6 s (calculated from measured prefill)		context: short prompt (~20 tokens) for decode; 9,451-token prompt for prefill📏 Forum post by @techman: three decode runs on the new build gave 17.9 / 18.1 / 18.1 tok/s (17.9 on the older build 0005475); cold prefill of a 9,451-token prompt 949-953 tok/s on the new build (759-797 on the old one); a warm prompt-cache hit returned in 177 ms; measured 28 Oct 2026

Prompt processing = prefill speed (how fast the prompt is read) as reported by the source, with the prompt length it was measured at; blank = the source does not report it. est. TTFT @8k is calculated from the measured prefill speed (8,192 ÷ prefill tok/s), shown only when prefill was measured on a prompt of 512+ tokens. It is not a measurement and ignores load time and caching. measured TTFT is the source's own time-to-first-token. Why prefill matters for coding assistants.

Homelab Brisbane community forum thread, posted 28 Oct 2026. The same rig also ran Qwen3.8 Flash-Next (a ~76 GB file, mostly offloaded to system RAM) at 21.0 tok/s with an expert cache and 13.0 without, prefill ~150 tok/s, but the post does not state that quant, so it is not listed. CPU, motherboard, system RAM and PSU are not given in the post.

Current GPU street price (estimate, not the builder's cost)Estimated GPU cost for this build today: ≈€858
New retail prices in Germany incl. VAT; used cards are often much cheaper. Prices change daily.
Source: https://discourse.homelabbrisbane.com.au/t/qwen3-8-27b-q4-k-m-v-s-qwen3-8-flash-next-testing-results-on-a-dual-rtx3060-rig/568

Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.

Embed this result

Copy into a README or blog. Links back to this build. See badge docs (CC BY 4.0 attribution). The SVG does not invent speeds — numbers stay on the build page.

Markdown

[![Measured on RigBase](https://getrigbase.com/badge/2x-rtx-3060-llamacpp-qwen38-27b-homelab-brisbane.svg)](https://getrigbase.com/builds/2x-rtx-3060-llamacpp-qwen38-27b-homelab-brisbane/)

HTML

<a href="https://getrigbase.com/builds/2x-rtx-3060-llamacpp-qwen38-27b-homelab-brisbane/" rel="noopener"><img src="https://getrigbase.com/badge/2x-rtx-3060-llamacpp-qwen38-27b-homelab-brisbane.svg" alt="Measured on RigBase" height="20"/></a>

Preview: Measured on RigBase · How to use

Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)

Quick answers: Can 2x RTX 3060 12GB run Qwen3.8 27B?