High-bandwidth memory-based neural network calculation apparatus and method

Abstract

The invention provides a high-bandwidth memory-based neural network calculation apparatus and method. The neural network calculation apparatus comprises at least one high-bandwidth memory and a neuralnetwork accelerator, wherein each high-bandwidth memory comprises multiple memories accumulated in a stacked manner; and the neural network accelerator is electrically connected with the high-bandwidth memory, performs data exchange with the high-bandwidth memory, and executes neural network calculation. According to the apparatus and the method, the storage bandwidth can be greatly increased; the high-bandwidth memory serves as a memory of the neural network calculation apparatus, so that the data exchange of input data and calculation parameters can be performed between a buffer and the memory more quickly, and the IO time is greatly shortened; the high-bandwidth memory adopts a stacked structure and does not occupy a transverse plane space, so that the area of the neural network calculation apparatus can be greatly reduced and can be reduced to about 5% of the area in the prior art; and the power consumption of the neural network calculation apparatus is reduced.

Description

technical field [0001] The invention relates to the application of high-performance storage in the field of neural network computing, in particular to a high-bandwidth memory-based neural network computing device and method. Background technique [0002] At present, the field of artificial intelligence is developing rapidly, and machine learning is also affecting all aspects of people's lives. As an important part of the field of machine learning, research on neural networks is also a hot spot in both industry and academia. Due to the huge amount of data in neural network calculations, how to accelerate the execution of neural network algorithms has become an important issue that we need to solve. Therefore, a dedicated neural network computing device came into being. [0003] Most of the DRAMs used in the architecture of neural network computing devices at the present stage are GDDR4 or GDDR5. However, in neural network computing devices, due to the consideration of band...

AI Technical Summary

Problems solved by technology

However, in neural network computing devices, due to the need to consider bandwidth, performance, power consumption and area issues, GDDR4 or GDDR5 can no longer fully meet the needs of neural network computing devices, and technology development has entered a bottleneck period.

Adding 1GB per second of bandwidth will result in more power consumption, which is not a wise, efficient or cost-effective choice for designers or consumers

At the same time, GDDR4 or GDDR5 still has a serious problem that it is difficult to reduce the area

Therefore, GDDR4 or GDDR5 will gradually hinder the continuous growth of neural network computing device performance

Images