LLM GPU Requirements

LLM Size Expected GPU Memory GPU Type Minimum Recommended GPUs
1B parameters 4 GB NVIDIA RTX 4060 (8GB)
RTX 3060 (12GB)
RTX 2060 Super (8GB)
GTX 1660 Ti (6GB)
1B parameters 4 GB AMD Radeon RX 6600 XT (8GB)
RX 5700 (8GB)
RX 6700 (10GB)
3B parameters 8 GB NVIDIA RTX 4070 (12GB)
RTX 3070 (8GB)
RTX 2080 Ti (11GB)
RTX 3060 Ti (8GB)
3B parameters 8 GB AMD Radeon RX 6700 XT (12GB)
RX 6800 (16GB)
RX 5700 XT (8GB)
7B parameters 16 GB NVIDIA RTX 4080 (16GB)
RTX 3080 (10GB with optimizations)
RTX 3090 (24GB)
RTX A5000 (24GB)
7B parameters 16 GB AMD Radeon RX 6800 XT (16GB)
RX 6900 XT (16GB)
RX 6950 XT (16GB)
13B parameters 32 GB NVIDIA RTX 4090 (24GB, with optimizations)
RTX A6000 (48GB)
A40 (48GB)
Tesla A100 (40GB/80GB)
13B parameters 32 GB AMD Radeon PRO W6800 (32GB)
Multiple RX 6900 XT (16GB each)
30B parameters 48 GB NVIDIA RTX A6000 (48GB)
A40 (48GB)
Tesla A100 (80GB)
H100 (80GB)
30B parameters 48 GB AMD Multiple Radeon PRO W6800 (32GB each)
MI210 (64GB)
1B parameters 16 GB NVIDIA RTX 4080 (16GB)
RTX 3080 (10GB with optimizations)
RTX 3090 (24GB)
RTX A5000 (24GB)
1B parameters 16 GB AMD Radeon RX 6800 XT (16GB)
RX 6900 XT (16GB)
RX 6950 XT (16GB)
3B parameters 32 GB NVIDIA RTX 4090 (24GB, with optimizations)
RTX A6000 (48GB)
A40 (48GB)
Tesla A100 (40GB/80GB)
3B parameters 32 GB AMD Radeon PRO W6800 (32GB)
Multiple RX 6900 XT (16GB each)
7B parameters 64 GB NVIDIA Tesla A100 (80GB)
H100 (80GB)
Multiple RTX A6000 (48GB each)
7B parameters 64 GB AMD MI210 (64GB)
Multiple Radeon PRO W6800 (32GB each)
MI250 (128GB)
13B parameters 128 GB NVIDIA Multiple Tesla A100 (80GB each)
H100 (80GB)
Multiple A40 (48GB each)
13B parameters 128 GB AMD MI250 (128GB)
MI250X (128GB)
30B parameters 256 GB NVIDIA Multiple H100 (80GB each)
Multiple A100 (80GB each)
DGX H100
DGX A100
30B parameters 256 GB AMD Multiple MI250X (128GB each)
Custom HPC solutions