Dissecting the Graphcore IPU Architecture via Microbenchmarking
Authors:
Preprint
Abstract
This report focuses on the architecture and performance of the Intelligence Processing Unit (IPU), a novel, massively parallel platform recently introduced by Graphcore and aimed at Artificial Intelligence/Machine Learning (AI/ML) workloads. We dissect the IPU's performance behavior using microbenchmarks that we crafted for the purpose. We study the IPU's memory organization and performance. We study the latency and bandwidth that the on-chip and off-chip interconnects offer, both in point-to-point transfers and in a spectrum of collective operations, under diverse loads. We evaluate the IPU's compute power over matrix multiplication, convolution, and AI/ML primitives. We discuss actual performance in comparison with its theoretical limits. Our findings reveal how the IPU's architectural design affects its performance. Moreover, they offer simple mental models to predict an application's performance on the IPU, on the basis of the computation and communication steps it involves. This report is the natural extension to a novel architecture of a continuing effort of ours that focuses on the microbenchmark-based discovery of massively parallel architectures.Keywords
All Related Versions
This publication has 9 references indexed in Scilit:
- Scrambled Linear Pseudorandom Number GeneratorsACM Transactions on Mathematical Software, 2021
- Aggregated Residual Transformations for Deep Neural Networks2017
- SSD: Single Shot MultiBox Detector2016
- Deep Residual Learning for Image Recognition2016
- Rethinking the Inception Architecture for Computer Vision2016
- Cell Multiprocessor Communication Network: Built for SpeedIEEE Micro, 2006
- Microbenchmark performance comparison of high-speed cluster interconnectsIEEE Micro, 2004
- LogP: towards a realistic model of parallel computation1993
- A bridging model for parallel computationCommunications of the ACM, 1990