Back to search results
Qwen3-8B · Q4_K_S
2x AMD Strix HalovLLM v0.23L2 Cross-framework verified
This page aggregates 1 real-world runs of Qwen3-8B (Q4_K_S) on 2x AMD Strix Halo with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.
137.45 tok/s
Decode
Decode speed
721.98 tok/s
Prefill
Prefill speed
0.59 s
TTFT
Time to first token
15.2 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
2 days ago
Last verified
Performance
- vLLM · Q3_K_L Decode 159.39 · Prefill 715.55 ·
- vLLM · Q4_K_S (current)Decode 137.45 · Prefill 721.98 ·
- Strata · IQ1_S Decode 293.93 · Prefill 1381.03 ·
- Strata · Q5_K_M Decode 157.71 · Prefill 787.63 ·
Core figures
- Decode (avg)
- 137.45 tok/s
- Prefill (avg)
- 721.98 tok/s
- TTFT (avg)
- 0.59 s
- VRAM (avg)
- 15.2 GB
- MTP acceptance rate
- —
- TTFB
- 1.1 GB
- Power draw
- 94.3 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Qwen3-8B
- Quantization
- Q4_K_S
- Framework
- vLLM
- Version
- v0.23
- Context length
- 4096 tokens
- Batch size
- 8
- GPU layers
- 999
- 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)
- 15.2 GB
- OS
- Ubuntu 24.04
- Driver
- 565.57
- CUDA
- N/A
- Power draw
- 94.3 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.23 · Ubuntu 24.04 · CUDA N/A · 4096 ctx
137.45 tok/s
Decode
721.98 tok/s
Prefill
0.59 s
TTFT
15.2 GB
VRAM
—
MTP
94.3 W
Power draw
L2 batch seed 2026-10-16: vLLM on hw_id=88 running model_id=53 at Q4_K_S