AIU

NVIDIA DRIVE AGX Thor

Edit@3 months ago

Intergrated Memory
No available data
Vector Compute
FP64
X
FP32
X
FP16
X
BF16
X
INT32
X
INT8
X

NVIDIA DRIVE AGX Thor General-Purpose Float-Point performance (Vector Performance / Scalar Performance)

Matirx Compute
FP64
X
FP32
X
FP16
250 T
500 T
FP8
500 T
1000 T
TF32
X
BF16
X
INT16
X
INT8
500 T
1000 T
INT4
X

NVIDIA DRIVE AGX Thor AI performance (Tensor Performance / Matrix Performance)

FP16: 250 TFLOPS, with sparsity: 500 TFLOPS

FP8: 500 TFLOPS, with sparsity: 1000 TFLOPS

INT8: 500 TOPS, with sparsity: 1000 TOPS

Hardware Specs
Process Node
nm
Launch Year