Statements (30)
Predicate | Object |
---|---|
gptkbp:instanceOf |
gptkb:graphics_card
Neural Engine |
gptkbp:architecture |
gptkb:Hopper
|
gptkbp:coreCount |
true
|
gptkbp:formFactor |
gptkb:PCIe
|
gptkbp:FP16Performance |
1979 TFLOPS
|
gptkbp:FP8Performance |
3958 TFLOPS
|
https://www.w3.org/2000/01/rdf-schema#label |
Nvidia H100 NVL
|
gptkbp:intendedUse |
AI inference
AI training |
gptkbp:manufacturer |
gptkb:Nvidia
|
gptkbp:market |
gptkb:cloud_service
|
gptkbp:memoryBusWidth |
6.25 TB/s
|
gptkbp:memoryType |
gptkb:HBM3
|
gptkbp:numberOfGPUsPerBoard |
2
|
gptkbp:NVLinkBandwidth |
900 GB/s
|
gptkbp:processNode |
gptkb:TSMC_4N
|
gptkbp:productType |
gptkb:Nvidia_H100
|
gptkbp:RAM |
188 GB
|
gptkbp:releaseYear |
2023
|
gptkbp:successor |
gptkb:Nvidia_A100_NVL
|
gptkbp:supportsFP16 |
true
|
gptkbp:supportsFP8 |
true
|
gptkbp:supportsMultiGPU |
true
|
gptkbp:supportsNVLink |
true
|
gptkbp:supportsPCIeGen5 |
true
|
gptkbp:supportsTransformerEngine |
true
|
gptkbp:TDP |
700 W
|
gptkbp:bfsParent |
gptkb:Nvidia_Hopper
|
gptkbp:bfsLayer |
6
|