Builds / RTX 3090 24GB
RTX 3090 at 250 W on Windows 11: Qwen3.8 27B Q4_K_M at 56.4 tok/s with MTP (llama.cpp)
Source: github.com ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
- GPU(s)
- RTX 3090 24GB
Amazon · eBay (used) · Newegg - VRAM / unified memory (GB)
- 24
- CPU
- AMD Ryzen 7 7800X3D
Amazon · eBay (used) · Newegg - Motherboard
- MSI MAG B650 Tomahawk
Amazon · eBay (used) · Newegg - RAM
- 32 GB DDR5
Amazon · eBay (used) · Newegg - PSU
- not stated
- Other
- 3090 headless in a PCIe Gen4 x2 chipset slot, power-limited to 250 W (default 420 W); a separate RTX 5070 drives the displays. Windows 11 Pro 26200, driver 591.86, llama.cpp b11382 (11fe02151) Windows CUDA 13 build
- Reported total cost
- not stated
- Power
- 250 W GPU power limit (as configured)
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Prompt processing (tok/s) | Details |
|---|---|---|---|---|---|
| Qwen3.8 27B | Unsloth UD-Q4_K_M (nextn/MTP head embedded), f16 KV | llama.cpp b11382 (Windows, CUDA 13), MTP n=2 | 56.37 | 1,073 @ ~12K-token prompt est. TTFT @8k ≈ 7.6 s (calculated from measured prefill) | ⚡ spec-decode (MTP)context: 65,536 ctx, thinking off📏 Narrative prompt, mean wall-clock tok/s of 5 measured runs after 3 warmups (CV 2.0%); code prompt 65.38 tok/s. Peak VRAM 20,472 MiB |
| Qwen3.8 27B | Unsloth UD-Q4_K_M, f16 KV | llama.cpp b11382 (Windows, CUDA 13), MTP off, -np 4 | 38.3 | not stated | context: author's own ~500-token narrative prompt📏 Author's MTP A/B at -np 4 on the same rig: off 38.3, n=3 48.4, n=2 52.4 tok/s |
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.
club-3090 GitHub discussion by cburba, 16 Oct 2026. PSU not stated.
Source: https://github.com/noonghunna/club-3090/discussions/1707
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
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 RTX 3090 run Qwen3.8 27B?