Back to search results
Llama-3-8B-Instruct · IQ4_XS
NVIDIA RTX 309024GBStrata v0.45L2 Cross-framework verified
This page aggregates 1 real-world runs of Llama-3-8B-Instruct (IQ4_XS) on NVIDIA RTX 3090 with Strata, contributed by 1 independent source platforms; metrics are averages of published measurements.
126.45 tok/s
Decode
Decode speed
522.82 tok/s
Prefill
Prefill speed
2.07 s
TTFT
Time to first token
14.53 GB
VRAM
VRAM usage
L2 Cross-framework verified
1
Measured runs
1
Independent sources
Other
Source platforms
Today
Last verified
Performance
- Ollama · IQ3_S Decode 90.89 · Prefill 298.18 ·
- Ollama · Q4_K_M Decode 84.07 · Prefill 419.77 ·
- Strata · IQ4_XS (current)Decode 126.45 · Prefill 522.82 ·
- Strata · Q3_K_M Decode 136.06 · Prefill 696.33 ·
- llama.cpp · Q8_0 Decode 84.9 · Prefill 636.8 ·
- llama.cpp · Q5_K_M Decode 123.7 · Prefill 927.8 ·
- llama.cpp · Q4_K_M Decode 149 · Prefill 1117.5 ·
Core figures
- Decode (avg)
- 126.45 tok/s
- Prefill (avg)
- 522.82 tok/s
- TTFT (avg)
- 2.07 s
- VRAM (avg)
- 14.53 GB
- MTP acceptance rate
- 0.57
- TTFB
- 1.01 GB
- Power draw
- 304.7 W
Configuration
Member-level fields are taken from the most recent run
- Model
- Llama-3-8B-Instruct
- Quantization
- IQ4_XS
- Framework
- Strata
- Version
- v0.45
- Context length
- 4096 tokens
- Batch size
- 512
- GPU layers
- 15
- Flash Attention
- On
Hardware
Nominal and measured figures are shown side by side; whether it runs is the reader's call
- GPU
- NVIDIA RTX 3090
- Nominal VRAM
- 24 GB
- Measured VRAM (avg)
- 14.53 GB
- OS
- Windows 11 24H2
- Driver
- 565.57
- CUDA
- 13.0
- Power draw
- 304.7 W
Sources & evidence
1 measured records in total, each traceable to its original source
v0.45 · Windows 11 24H2 · CUDA 13.0 · 4096 ctx
126.45 tok/s
Decode
522.82 tok/s
Prefill
2.07 s
TTFT
14.53 GB
VRAM
0.57
MTP
304.7 W
Power draw
L2 batch seed 2026-10-18: Strata on hw_id=2 running model_id=69 at IQ4_XS