RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2

ACCEL (All-Analogue Chip Combining Electronics and Light)

Product
Developers: SMIC (Semiconductor Manufacturing International Corporation)
Date of the premiere of the system: November 2023
Branches: Electrical and Microelectronics
Technology: Processors

2023: Product Announcement

On November 1, 2023, it became known about the creation of a chip in China that works significantly faster and consumes less power than modern chips for working with AI on tasks such as image recognition and autonomous work.

Although the new chip cannot immediately replace the chips currently used in computers and smartphones, in the near future it can be used in wearable devices, electric vehicles or in smart factories. China hopes the development will help the country spread artificial intelligence.

Chinese scientists produce chip that works significantly faster and uses less power than other modern chips

China is trying to catch up with the rest of the countries in the race for leadership in the AI market, especially after US sanctions in this area, which have limited Chinese companies access to technology.

The new chip is called ACCEL (All-Analysis Chip Combining Electronics and Light). It works on photons. The idea of using light is not new, but photons are harder to control, so the industry mainly uses chips on electricity.

In a laboratory study, ACCEL reached a rate of 4.6 flops. This is three thousand times faster than Nvidia the A100, one of the most popular chips in the world, while the new chip consumes 4 million times less energy. China You do not have access to the A100, since this product is under sanctions.

File:Aquote1.png
The performance of the chip can be further optimized through improving the production process or through the use of more expensive material less than 100 nanometers, commented experts from Tsinghua University, which conducted the testing.
File:Aquote2.png

The new chip was manufactured by SMIC (Semiconductor Manufacturing International Corporation) based on 20-year-old transistor technologies, making it cheap to manufacture.[1]

Notes