According to reports, a research team led by Professor Hussam Amrouch of the Technical University of Munich (TUM) has developed an architecture that can be used for artificial intelligence and is twice as powerful as similar in-memory computing methods.
The latest research results have been recently published in the magazine "Nature". It is said that the innovative new chip technology integrates data storage and processing functions, greatly improving efficiency and performance. The chips, inspired by the human brain, are expected to be commercially available within three to five years and will require interdisciplinary collaboration to meet industry safety standards.
It is reported that the Amrouch team applied a new computing model using a special circuit called a ferroelectric field-effect transistor (FeFET). Within a few years, this may prove applicable to generative artificial intelligence, deep learning algorithms, and robotics applications.
Their basic idea is simple: In the past, chips were only used for calculations on transistors, but now they are also used for data storage. This saves both time and effort. Amrouch said: "As a result, the performance of the chip has also improved."
As human needs continue to improve, future chips must be faster and more efficient than previous ones. Therefore, they cannot heat up quickly. This is essential if they are to support applications such as real-time computing while drones are flying.
Researchers say that such tasks are very complex and energy-consuming for computers
These key requirements for the chip can be summarized by the mathematical parameter TOPS/W: "terahertz operations per second per watt". This can be seen as an important technical indicator of future chips: when one watt (W) of power is provided, how many teraflops of operations (TOP) can the processor perform per second (S)
This new artificial intelligence chip can provide 885 TOPS/W. That makes it twice as powerful as similar AI chips, including Samsung's MRAM chips. The operating speed of currently commonly used CMOS (complementary metal oxide semiconductor) chips is between 10-20 TOPS/W.
Specifically, the researchers borrowed the principles of modern chip architecture from humans. "In the brain, neurons process signals and synapses remember this information," said Amrouch, describing how humans are able to learn and recall complex relationships."
To achieve this, the chip uses "ferroelectric" (FeFET) transistors. This electronic switch has the unique additional property of reversing polarity when voltage is applied, allowing it to store information even in the event of a power outage. Additionally, they are able to store and process data simultaneously
Amrouch believes: "We can now build efficient chipsets for applications such as deep learning, generative artificial intelligence or robotics, for example, where data must be processed where it is generated."
However, professors from the Munich Institute for Integrated Robotics and Machine Intelligence (MIRMI) at the Technical University of Munich believe that it will take several years to achieve this goal. He believes that the first memory chip suitable for practical applications will not be available until three to five years at the earliest
Rewritten content: Quoted from Financial Associated Press
The above is the detailed content of Next-generation AI chip performance doubled? New technology could mimic the human brain to save energy…. For more information, please follow other related articles on the PHP Chinese website!