The world's first commercial in-memory computing chip has recently been released, and it has the ability to improve the energy efficiency of artificial intelligence (AI) processing by over 10 times. In-memory computing is a technology that embeds computing units directly into the memory of a computer, rather than having them separate like in the traditional von Neumann architecture. This architecture includes both storage units and computing units, with data being migrated between the two. If the memory speed cannot keep up with the CPU calculation speed, it can result in a "memory wall" situation.
In-memory computing aims to overcome this bottleneck by integrating the computing units directly into the memory. This technology was first proposed in the 1960s, but it has taken until now for it to become commercially available. Major manufacturers like Samsung are currently researching this solution as a way to improve the efficiency and speed of AI processing.
In March 2022, Zhicun Technology, a memory-computing integrated chip company, announced the official release of the WTM2101, the world's first in-memory computing system-on-a-chip (SoC) mass-produced for commercial use. This is a major milestone for the company, as the WTM2101 has already been successfully commercialized in terminal applications, providing AI processing solutions such as voice and video recognition.
Zhicun Technology has announced that it will be launching the WTM-8 series of products in 2023 and 2024, which are expected to increase computing power by 500 to 600 times. This is a significant improvement and demonstrates the rapid pace of technological advancement in the field of in-memory computing.
Around 2025, Zhicun Technology will also be launching the WTM-C series of products, which will be designed for use in edge servers. These products will further expand the capabilities of in-memory computing and allow for even more powerful and efficient processing. With advancements in technology and integration scale, it is expected that in-memory computing products will increase their computing power by an average of 5 to 10 times per year in the coming years.
Zhicun Technology is a company that was founded in 2017 and is focused on the development of in-memory computing chips. The company has pioneered the use of Flash memory to complete the storage and computing of neural networks, solving the problem of AI storage walls and improving the efficiency and cost-effectiveness of computing.
The company's research and development (R&D) team is composed of Dr. Wang Shaodi and Dr. Guo Xinjie, as well as a number of scholars and industry practitioners with an average of over 10 years of work experience. This team is known for its international-level expertise in storage and computing integration and chip technology.
The WTM2101 chip from Zhicun Technology is designed for low-power artificial intelligence of things (AIoT) applications. It has the ability to complete large-scale deep learning operations using only microwatts to milliwatts of power, making it particularly suitable for smart voice and health services in wearable devices. The chip has already been mass-produced and is now available on the market.
In the future, Zhicun Technology will be launching the WTM-8 series of chips, which are oriented towards 4-32Tops computing power products. These chips are intended for use in real-time processing of 4K-8K video and other demanding applications.
0 Comments