Back to search results
Qwen3.8-9B · IQ4_XS
2x AMD Strix Halollama.cpp v0.43L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.8-9B (IQ4_XS) on 2x AMD Strix Halo with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
145.51 tok/s
Decode
Decode speed
650.25 tok/s
Prefill
Prefill speed
1.51 s
TTFT
Time to first token
16.09 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- llama.cpp · IQ1_S Decode 185.41 · Prefill 664.26 ·
- llama.cpp · IQ4_XS (current)Decode 145.51 · Prefill 650.25 ·
- Strata · IQ2_M Decode 243.64 · Prefill 1152.92 ·
- Strata · IQ3_S Decode 220.16 · Prefill 1160.56 ·
Core figures
- Decode (avg)
- 145.51 tok/s
- Prefill (avg)
- 650.25 tok/s
- TTFT (avg)
- 1.51 s
- VRAM (avg)
- 16.09 GB
- MTP acceptance rate
- —
- TTFB
- 1.16 GB
- Power draw
- 49.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.8-9B
- Quantization
- IQ4_XS
- Framework
- llama.cpp
- Version
- v0.43
- Context length
- 32768 tokens
- Batch size
- 256
- GPU layers
- 33
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- 2x AMD Strix Halo
- Measured VRAM (avg)
- 16.09 GB
- OS
- Windows 11 23H2
- Driver
- amdgpu-pro 24.20
- CUDA
- N/A
- Power draw
- 49.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.43 · Windows 11 23H2 · CUDA N/A · 32768 ctx
145.51 tok/s
Decode
650.25 tok/s
Prefill
1.51 s
TTFT
16.09 GB
VRAM
—
MTP
49.3 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=88 running model_id=52 at IQ4_XS