Back to search results
Qwen3.8-9B · IQ1_S
2x AMD Strix Halollama.cpp v0.26L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3.8-9B (IQ1_S) on 2x AMD Strix Halo with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.
185.41 tok/s
Decode
Decode speed
664.26 tok/s
Prefill
Prefill speed
1.41 s
TTFT
Time to first token
10.04 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 (current)Decode 185.41 · Prefill 664.26 ·
- llama.cpp · IQ4_XS 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)
- 185.41 tok/s
- Prefill (avg)
- 664.26 tok/s
- TTFT (avg)
- 1.41 s
- VRAM (avg)
- 10.04 GB
- MTP acceptance rate
- —
- TTFB
- 1.48 GB
- Power draw
- 93.6 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3.8-9B
- Quantization
- IQ1_S
- Framework
- llama.cpp
- Version
- v0.26
- Context length
- 8192 tokens
- Batch size
- 256
- GPU layers
- 99
- Flash Attention
- Off
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)
- 10.04 GB
- OS
- Ubuntu 22.04
- Driver
- amdgpu 6.11
- CUDA
- N/A
- Power draw
- 93.6 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.26 · Ubuntu 22.04 · CUDA N/A · 8192 ctx
185.41 tok/s
Decode
664.26 tok/s
Prefill
1.41 s
TTFT
10.04 GB
VRAM
—
MTP
93.6 W
Power draw
L2 batch seed 2026-10-16: llama.cpp on hw_id=88 running model_id=52 at IQ1_S