✅ — NVIDIA GB200 NVL72 Can Run Llama 3.1 8B
The NVIDIA GB200 NVL72 packs 14131.2GB VRAM. Llama 3.1 8B needs ~4.4GB. You're good to go.
✅ Plenty of headroom
Hardware Specs
- GPU
- NVIDIA GB200 NVL72
- VRAM
- 14131.2 GB
- Bandwidth
- 576000 GB/s
- MSRP
- $3,000,000
Model Requirements
- Model
- Llama 3.1 8B
- Parameters
- 8B
- VRAM (Q4)
- ~4.4 GB
- Context
- N/A
Estimated Performance
108000 tok/s
Estimated inference speed (Q4 quantization)
Buy vs Rent — Do the Math
Rent breaks even after 10,344,827hrs • Used pays off faster