gptkbp:instanceOf
|
hardware accelerator
|
gptkbp:announced
|
2016
|
gptkbp:architecture
|
ASIC
|
gptkbp:availableOn
|
gptkb:Google_Cloud_Platform
|
gptkbp:category
|
machine learning hardware
AI hardware
application-specific integrated circuit
|
gptkbp:cloudBased
|
gptkb:Cloud_TPU
|
gptkbp:competitor
|
gptkb:NVIDIA_GPU
gptkb:AMD_GPU
gptkb:Intel_Habana
|
gptkbp:developedBy
|
gptkb:Google
|
gptkbp:energyEfficiency
|
high
|
gptkbp:formFactor
|
PCIe card
rack-mounted
|
gptkbp:fullName
|
Neural Engine
|
gptkbp:generation
|
v3
v1
v2
v4
|
https://www.w3.org/2000/01/rdf-schema#label
|
TPU
|
gptkbp:location
|
gptkb:cloud_service
|
gptkbp:notableFeature
|
low latency
high throughput
matrix multiplication unit
optimized for TensorFlow
|
gptkbp:platform
|
gptkb:TensorFlow
|
gptkbp:purpose
|
accelerate machine learning workloads
|
gptkbp:RAM
|
gptkb:HBM
|
gptkbp:relatedTo
|
gptkb:Google_Coral
gptkb:Edge_TPU
Neural Engine
|
gptkbp:studiedBy
|
yes
|
gptkbp:supportsPrecision
|
bfloat16
float32
int8
|
gptkbp:usedBy
|
gptkb:Google_Assistant
gptkb:Google_Photos
gptkb:Google_Search
gptkb:Google_Translate
|
gptkbp:usedFor
|
gptkb:artificial_intelligence
deep learning
neural network training
neural network inference
|
gptkbp:website
|
https://cloud.google.com/tpu
|
gptkbp:bfsParent
|
gptkb:graphics_card
gptkb:TensorFlow
gptkb:Compute_Engine
|
gptkbp:bfsLayer
|
5
|