Arao AI
Back to search results

Qwen3.8-9B · IQ3_S

Start a discussion
AMD Radeon RX 9070 XT16GBllama.cpp v0.51L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.8-9B (IQ3_S) on AMD Radeon RX 9070 XT with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

102.09 tok/s

Decode

Decode speed

492.72 tok/s

Prefill

Prefill speed

0.36 s

TTFT

Time to first token

14.57 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

5 days ago

Last verified

Performance

  1. llama.cpp · Q3_K_L Decode 114.6 · Prefill 612.55 ·
  2. Ollama · Q4_K_M Decode 103.73 · Prefill 474.73 ·
  3. llama.cpp · IQ3_S (current)Decode 102.09 · Prefill 492.72 ·
  4. Ollama · IQ3_XS Decode 105.64 · Prefill 538.49 ·
  5. Ollama · Q4_K_S Decode 80.27 · Prefill 324.54 ·

Core figures

Decode (avg)
102.09 tok/s
Prefill (avg)
492.72 tok/s
TTFT (avg)
0.36 s
VRAM (avg)
14.57 GB
MTP acceptance rate
—
TTFB
0.82 GB
Power draw
233.6 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.8-9B
Quantization
IQ3_S
Framework
llama.cpp
Version
v0.51
Context length
4096 tokens
Batch size
512
GPU layers
20
Flash Attention
Off

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
AMD Radeon RX 9070 XT
Nominal VRAM
16 GB
Measured VRAM (avg)
14.57 GB
OS
Ubuntu 22.04
Driver
amdgpu 6.10
CUDA
N/A
Power draw
233.6 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-13

    v0.51 · Ubuntu 22.04 · CUDA N/A · 4096 ctx

    102.09 tok/s

    Decode

    492.72 tok/s

    Prefill

    0.36 s

    TTFT

    14.57 GB

    VRAM

    —

    MTP

    233.6 W

    Power draw

    L2 batch seed 2026-10-13: llama.cpp on hw_id=16 running model_id=52 at IQ3_S