Statements (30)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
gptkb:graphics_card
|
| gptkbp:announced |
March 2022
|
| gptkbp:architecture |
Data center GPU
|
| gptkbp:firstGPU |
gptkb:Nvidia_H100
|
| https://www.w3.org/2000/01/rdf-schema#label |
Nvidia Hopper
|
| gptkbp:manufacturer |
gptkb:Nvidia
|
| gptkbp:memoryBusWidth |
up to 3 TB/s
|
| gptkbp:notableFeature |
gptkb:Confidential_Computing
gptkb:Multi-Instance_GPU_(MIG) gptkb:Transformer_Engine DPX instructions |
| gptkbp:predecessor |
gptkb:Nvidia_Ampere
|
| gptkbp:processNode |
gptkb:TSMC_4N
|
| gptkbp:releaseYear |
2022
|
| gptkbp:successor |
gptkb:Nvidia_Blackwell
|
| gptkbp:supports |
gptkb:HBM3_memory
gptkb:NVLink_4.0 gptkb:PCIe_5.0 gptkb:Tensor_Cores FP16 precision FP8 precision |
| gptkbp:targetMarket |
gptkb:artificial_intelligence
High-performance computing |
| gptkbp:transistorCount |
80 billion (H100)
|
| gptkbp:usedIn |
gptkb:Nvidia_GH200_Grace_Hopper_Superchip
gptkb:Nvidia_H100_NVL gptkb:Nvidia_H100_PCIe gptkb:Nvidia_H100_SXM |
| gptkbp:bfsParent |
gptkb:Nvidia
|
| gptkbp:bfsLayer |
5
|