Graphcore IPU
Vendor: Graphcore
Category: Innovative Architecture / Machine Intelligence
Architecture: IPU (Intelligence Processing Unit)
Introduction
A UK AI chip pioneer. The IPU uses a massively parallel architecture, optimized specifically for sparse computation and graph neural networks, delivering outstanding performance on specific AI workloads.
Specifications
| Model | Compute | Memory | Interface | TDP | Process |
|---|---|---|---|---|---|
| Bow IPU | 350 TFLOPS (FP16) | 900MB on-chip SRAM | IPU-Fabric Interconnect | 350W | 7nm |
| Mk2 IPU | 250 TFLOPS (FP16) | 900MB on-chip SRAM | IPU-Fabric Interconnect | 275W | 7nm |
Official Website
Driver Downloads
Linux
Related Documentation
- Graphcore Documentation Center
- Poplar Programming Model
- PyTorch Adaptation (PopTorch)
- TensorFlow Adaptation (PopART)
OS Support
| Windows | Linux | macOS | Android |
|---|---|---|---|
| ❌ | ✅ | ❌ | ❌ |
Version History
| Version | Release Date | Description |
|---|---|---|
| PopTorch 2.0 | 2024 | PyTorch 2.x compatibility |
Performance Benchmarks
| Model | Task | Performance Metric |
|---|---|---|
| Bow IPU | BERT-Large Training | ~85% H100 efficiency |
| Bow IPU | PopART Graph Training | Supports 1B+ parameter models |
| Mk2 IPU | General AI | Previous generation product |
Pricing Information
| Model | Reference Price | Notes |
|---|---|---|
| Bow IPU | Contact sales | Sold by RapidAI (China subsidiary) |
| Mk2 IPU | Contact sales | Gradually replaced by Bow |
Quick Setup
Linux
# 1. Install Poplar SDK (Graphcore programming framework)
pip install poplar
pip install popart
# 2. Verify
gc-info
Code Examples
Python (PopART)
import popart
import popart.ir as pir
import torch
# Use PopART IPU backend
ir = pir.Ir()
with ir.main_graph:
x = pir.ops.host_store(pir.constant([1.0, 2.0]))
Architecture Highlights
- IPU (Intelligence Processing Unit): A processor designed for large model training, with intensive on-chip communication
- Bow Architecture: Second-generation IPU, introducing memory optimizations and computational efficiency improvements
- Poplar + PopART: Graphcore's unified compiler and runtime framework
Model Compatibility
| Model/Framework | Support Status | Notes |
|---|---|---|
| PopART (PyTorch) | ✅ Native | Best support |
| TensorFlow | ✅ | PopART backend |
| PaddlePaddle | ⚠️ | Adaptation in progress |
| Large Model Training | ✅ | Distributed training support |
Related Products
If you are evaluating alternatives, the following products may also fit your scenario:
- Google Cloud TPU — Google (TPU Tensor Processing Unit)
- Cerebras WSE-3 — Cerebras (ASIC dedicated accelerator)
- SambaNova SN40L — SambaNova (ASIC dedicated accelerator)
- NVIDIA GPU / CUDA — NVIDIA (GPU Graphics Processor)
- AMD ROCm / GPU — AMD (GPU Graphics Processor)
- Intel Gaudi — Intel (ASIC dedicated accelerator)