Habana Gaudi family

E660955

The Habana Gaudi family is a line of AI training accelerators developed by Habana Labs (an Intel company) designed to deliver high-performance, power-efficient deep learning computation for data centers.

Try in SPARQL Jump to: Statements Referenced by

Statements (43)

Predicate Object
instanceOf AI accelerator family
deep learning training accelerator
architectureType custom AI accelerator
category data center accelerator
compatibleWith PyTorch NERFINISHED
TensorFlow NERFINISHED
deploymentFormFactor PCIe card
integrated server system
mezzanine card
designGoal high performance
power efficiency
scalability
developer Habana Labs NERFINISHED
domain data center computing
focus training rather than inference
includesModel Habana Gaudi NERFINISHED
Habana Gaudi2
Habana Gaudi3
interconnectType RoCE-based Ethernet
introducedBy Habana Labs NERFINISHED
optimizationGoal energy efficiency in AI training
total cost of ownership reduction
parentCompany Intel NERFINISHED
softwareStack SynapseAI NERFINISHED
supports BF16 NERFINISHED
FP16
INT8
convolutional neural networks
data parallelism
large language model training
large-scale distributed training
mixed-precision computation
model parallelism
tensor operations
transformer architectures
targetMarket cloud service providers
enterprise data centers
targetWorkload computer vision models
natural language processing models
recommendation systems
useCase AI model training
deep learning training
vendor Intel NERFINISHED

Referenced by (2)

Full triples — surface form annotated when it differs from this entity's canonical label.

Gaudi belongsToProductFamily Habana Gaudi family
Gaudi2 belongsToProductLine Habana Gaudi family