32GB
HBM2e
820
GB/s bandwidth
// about AWS Inferentia2
AWS's second-generation inference chip
// similar hardware in Datacenter
NVIDIA B200
NVIDIA Blackwell flagship datacenter GPU for frontier model training
NVIDIA · Datacenter · N/A
NVIDIA GB200 NVL72
Rack-scale Grace Blackwell superchip for trillion-parameter models
NVIDIA · Datacenter · N/A
NVIDIA GB300 NVL72
Rack-scale Blackwell Ultra superchip with expanded memory
NVIDIA · Datacenter · N/A
NVIDIA H100 PCIe
PCIe-form-factor Hopper datacenter GPU
NVIDIA · Datacenter · ~$25000
NVIDIA H100 SXM
Hopper-architecture datacenter GPU for large-scale AI training
NVIDIA · Datacenter · ~$30000
NVIDIA H200
High-memory Hopper refresh for large-model inference
NVIDIA · Datacenter · ~$32000