IPU
E660958
IPU (Intelligence Processing Unit) is Graphcore's specialized processor architecture designed to accelerate machine learning and artificial intelligence workloads.
All labels observed (1)
| Label | Occurrences |
|---|---|
| IPU canonical | 1 |
Statements (48)
| Predicate | Object |
|---|---|
| instanceOf |
AI accelerator
ⓘ
Intelligence Processing Unit ⓘ processor architecture ⓘ |
| architectureType |
MIMD architecture
ⓘ
manycore processor ⓘ |
| belongsToCategory |
AI hardware
ⓘ
accelerator chips ⓘ parallel processors ⓘ |
| commercialStatus | commercially available ⓘ |
| comparedTo |
CPU
ⓘ
GPU NERFINISHED ⓘ |
| designedFor |
artificial intelligence workloads
ⓘ
deep learning workloads ⓘ machine learning workloads ⓘ |
| developedBy | Graphcore NERFINISHED ⓘ |
| hasFeature |
distributed on-chip memory
ⓘ
fine-grained synchronization ⓘ hardware support for sparsity ⓘ high on-chip memory bandwidth ⓘ high-throughput interconnect ⓘ large number of independent cores ⓘ low-precision arithmetic support ⓘ |
| manufacturer | Graphcore NERFINISHED ⓘ |
| marketedAs | Intelligence Processing Unit NERFINISHED ⓘ |
| optimizedFor |
fine-grained parallelism
ⓘ
model parallelism ⓘ parallel computation ⓘ sparse computation ⓘ |
| partOf | Graphcore IPU systems NERFINISHED ⓘ |
| supports |
inference of neural networks
ⓘ
training of neural networks ⓘ |
| supportsFramework |
ONNX
NERFINISHED
ⓘ
PyTorch NERFINISHED ⓘ TensorFlow NERFINISHED ⓘ |
| targetCustomer |
AI researchers
ⓘ
cloud service providers ⓘ machine learning engineers ⓘ |
| targetWorkload |
computer vision models
ⓘ
graph neural networks ⓘ natural language processing models ⓘ recommendation systems ⓘ |
| usedBy |
enterprises
ⓘ
research institutions ⓘ |
| usedIn |
cloud AI services
ⓘ
data centers ⓘ high-performance computing systems ⓘ |
| usedWith |
Graphcore SDK
NERFINISHED
ⓘ
Poplar software stack ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.