In-Datacenter Performance Analysis of a Tensor Processing Unit

by   Norman P. Jouppi, et al.

Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU)---deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs (caches, out-of-order execution, multithreading, multiprocessing, prefetching, ...) that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95 NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X - 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X - 80X higher. Moreover, using the GPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU.



There are no comments yet.


page 3

page 4


Improving Multi-Application Concurrency Support Within the GPU Memory System

GPUs exploit a high degree of thread-level parallelism to hide long-late...

Learning to Optimize Tensor Programs

We introduce a learning-based framework to optimize tensor programs for ...

Large-Scale Approximate k-NN Graph Construction on GPU

k-nearest neighbor graph is a key data structure in many disciplines suc...

Power Reduction of Automatic Heterogeneous Device Offloading

In recent years, utilization of heterogeneous hardware other than small ...

Dissecting the NVidia Turing T4 GPU via Microbenchmarking

In 2019, the rapid rate at which GPU manufacturers refresh their designs...

CSM-NN: Current Source Model Based Logic Circuit Simulation – A Neural Network Approach

The miniaturization of transistors down to 5nm and beyond, plus the incr...

Hardware-accelerated Simulation-based Inference of Stochastic Epidemiology Models for COVID-19

Epidemiology models are central in understanding and controlling large s...

Code Repositories


A open source reimplementation of Google's Tensor Processing Unit (TPU).

view repo
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.