✅ — Google Cloud A3 (8x H100) Can Run Llama 3.1 405B
The Google Cloud A3 (8x H100) packs 640GB VRAM. Llama 3.1 405B needs ~222.8GB. You're good to go.
✅ Plenty of headroom
Hardware Specs
- GPU
- Google Cloud A3 (8x H100)
- VRAM
- 640 GB
- Bandwidth
- 26400 GB/s
- MSRP
- $0
Model Requirements
- Model
- Llama 3.1 405B
- Parameters
- 405B
- VRAM (Q4)
- ~222.8 GB
- Context
- N/A
Estimated Performance
98 tok/s
Estimated inference speed (Q4 quantization)