With super computing power, the computing energy efficiency is 1000 times that of traditional chips,

Mondo Technology Updated on 2024-01-31

Three minutes to talk about popular science

Have you ever thought that if there is such a chip, it can achieve super computing power without taking up too much space and consuming too much power, and can even simulate the human brain for logical judgment and decision-making?In this issue, we will get to know the memristor chip, which is said to be 1000 times more efficient than traditional chips, so is it as magical as the legend?

Memristor chip is a kind of circuit element based on memristor, memristor is the fourth basic circuit element after resistance, capacitance and inductance, it can directly connect the charge and magnetic flux two physical quantities, but also can remember its own resistance value, even if the power is off, it will not be lost. This special function allows the memristor to be used as both a memory and a calculator, and can also use the laws of physics to directly do multiplication and addition operations, which is called storage and computing integration.

What are the advantages of memristor chips?First of all, its small size allows it to be made into a high-density array of intersections to achieve massively parallel in-memory computing, which can save a lot of space and time. Second, it has low power consumption because it doesn't need to move data from memory to combinator and back again, as is the case with traditional computers, which reduces a lot of power consumption and latency for data transfer. Thirdly, it integrates storage and computing, and can calculate while storing data, so that the bottleneck of traditional computer design, that is, the mismatch between the speed of storage and computing, can be avoided.

At present, the main application direction of memristor chips is neural network chips, that is, memristors are used to simulate human neurons and synapses to achieve brain-like computing. Neural network chips are mainly used in the field of artificial intelligence, such as image recognition, speech recognition, natural language processing, etc., they need to perform matrix operations on a large amount of data, which is what memristors are good at. At present, a large number of AI operations are carried out on cloud computers, after all, the performance of mobile devices themselves is very limited, and they cannot undertake huge AI computing tasks, but services such as unmanned driving and voice assistants need to respond quickly, and cloud services always have delays, which are not as fast as local services, and cloud services still have the risk of privacy leakage, so terminal devices are required to be able to complete calculations independently. The memristor chip can greatly increase the computing power without reducing the accuracy of the neural network, and at the same time, it can also significantly reduce the power consumption, which is very useful for scenarios such as edge computing and the Internet of Things.

You might say, if the memristor chip is so powerful, why don't I seem to have heard of this thing before?This is because memristor chips are still in the research and development stage, and there is no large-scale commercial application, but many domestic and foreign scientific research institutions and enterprises have explored and innovated in this regard. For example, Professor Wu Huaqiang's team at Tsinghua University has successfully developed a multi-array memristor storage and computing integrated system that is two orders of magnitude more energy efficient than cutting-edge graphics processor chips when processing convolutional neural networks, a study published in the journal Nature in January this year. In addition, Huawei has already deployed memristor chips and applied for a number of related patents, covering the manufacturing method, structural design, circuit design, and neural network design of memristors, showing that it attaches great importance to memristor chips.

Memristor chip is a revolutionary new type of chip, and its emergence will bring great changes to the computer field and will also bring more possibilities to the field of artificial intelligence. There are many promising directions for the future of memristor chips, such as brain-computer interfaces, quantum computing, biomedicine, etc., all of which require the characteristics of memristor chips with high efficiency, low power consumption, and integrated storage and computing. There are still many challenges in the development of memristor chips, such as the consistency, reliability, and stability of devices, which require continuous research and innovation. However, we believe that with the advancement of science and technology, the potential of memristor chips will be fully realized, bringing more benefits to human life and the development of society.

Related Pages