4x NVIDIA RTX 3090
DesktopFour RTX 3090s, 96 GB of VRAM in total. A multi-GPU rig is not a single memory pool, so bandwidth and TDP are left empty.
2
Run records
2
Models covered
2
Frameworks covered
L0 Self-reported × 2
Evidence mix
Typical measured performance
Published benchmark averages per model × framework × quantization on this hardware.
- vLLM · IQ2_XS Decode — · Prefill — ·
- llama.cpp · UD-IQ2_XXS Decode 35.68 · Prefill 420.73 ·
Fit check
Physics-based estimate from VRAM and memory bandwidth: pick a model to see which quantizations fit and the expected speed
Formula: tok/s = (bandwidth ÷ model size) × 0.55, adapted from llmfit; see run records for measured numbers
Key specs
- Vendor
- NVIDIA
- Architecture
- Ampere
- VRAM
- 96 GB
- Memory bandwidth
- 3600 GB/s
Run records
2 configurations
| Model | Framework | Quantization | Decode | Gen time | Samples |
|---|---|---|---|---|---|
| GLM-5.3 | vLLM | IQ2_XS | — | — | 1 |
| DeepSeek-V4-Flash | llama.cpp | UD-IQ2_XXS | 35.7 tok/s | — | 1 |