Statements (30)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
Tensor Core
|
| gptkbp:architecture |
gptkb:Ampere
|
| gptkbp:developedBy |
gptkb:NVIDIA
|
| gptkbp:feature |
mixed-precision computing
hardware support for structured sparsity improved throughput over Volta sparsity acceleration |
| https://www.w3.org/2000/01/rdf-schema#label |
NVIDIA Ampere Tensor Cores
|
| gptkbp:introducedIn |
gptkb:NVIDIA_Ampere_architecture
|
| gptkbp:marketedAs |
AI research
high performance computing datacenter |
| gptkbp:predecessor |
NVIDIA Volta Tensor Cores
|
| gptkbp:releaseYear |
2020
|
| gptkbp:successor |
NVIDIA Hopper Tensor Cores
|
| gptkbp:supports |
BF16
FP16 INT4 INT8 TF32 FP64 (on A100 only) |
| gptkbp:usedFor |
deep learning
matrix multiplication AI acceleration |
| gptkbp:usedIn |
gptkb:NVIDIA_A100
NVIDIA GA100 NVIDIA GA102 NVIDIA GA104 |
| gptkbp:bfsParent |
gptkb:TensorFloat-32
|
| gptkbp:bfsLayer |
7
|