NVIDIA Blackwell
Realizes: Grace Hopper superchips
Upcoming NVIDIA Blackwell architecture is tuned for inference, pairing DPX and Tensor cores in the Grace Hopper superchip line to accelerate dense matrix and sparse transformer workloads.
Examples
🔗
Grace Hopper Superchip Inference
Grace Hopper superchip deployment highlights DPX and Tensor core acceleration for large-scale inference workloads running on NVIDIA Blackwell.
DPX instruction sequences
Tensor Core inference kernels
milliseconds
large
J