Builds / RTX 3090
RTX 3090: Qwen3.8 27B Q4_K_M on llama.cpp server at 39 tok/s with an 8K prompt, 34 at 32K (prefill 1,226 / 1,103)
Last updated:
- GPU(s)
- RTX 3090
Amazon · eBay (used) · Newegg - VRAM / unified memory (GB)
- 24
- CPU
- not stated
- Motherboard
- not stated
- RAM
- not stated
- PSU
- not stated
- Other
- x86_64 Linux VM, llama.cpp server
- Reported total cost
- not stated
- Power
- not stated
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Prompt processing (tok/s) | Details |
|---|---|---|---|---|---|
| Qwen3.8 27B | Q4_K_M | llama.cpp server | 39 | 1,226 @ 8,192-token prompt est. TTFT @8k ≈ 6.7 s (calculated from measured prefill) | context: 8,192-token prompt, 128 generated, no-thinking template📏 Server-reported rates, exact uncached prompt, one sample; with a 32,768-token prompt 34 tok/s decode (prefill 1,103); measured 2 Nov 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.
Owner-run post by Joshua Warren (joshuawarren.com, 2 Nov 2026). Each row in the post is a single sample. An earlier 32K run on the same setup decoded at only about 1 tok/s (prefill 1,054); the post does not explain it, so RigBase lists the repeat run.
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Quick answers: Can RTX 3090 run Qwen3.8 27B?