Integrated circuit (IC) design is a complex work that constantly breaks through the boundaries of density and performance. Now, artificial intelligence (AI) plays an increasing role in IC design.
Although AI has recently received more attention due to its amazing abilities such as natural language processors such as ChatGPT, the tool has also been used in various stages of IC design, including design optimization, layout, simulation, and verification. Artificial intelligence algorithms can also help explore design space more effectively, and discover the best design configuration faster than traditional methods.
As design becomes more complicated than ever, artificial intelligence can become a powerful asset for IC designers.
Chip manufacturers use artificial intelligence to meet strict design constraints
Because billions of crystal pipes are limited to a small chip area, the task of IC designers is to optimize the design despite strict restrictions.
First, the area of the chip must be minimized to adapt to the small shape size of today's equipment and reduce the manufacturing cost. The power consumption of the layout is also worthy of attention; power consumption will also affect the deployment cost and the impact of chips on the environment. Some intensive areas and configuration are prone to overheating, and the cooling mechanism or clever layout require. Considering these and many other factors, IC designers spend about eight to nine months to generate chip layouts that meet each strict requirements.
In order to accelerate and optimize the IC design process, many companies (including some of the largest companies in the technology industry) are now investing in artificial intelligence tools to complete some heavy work.
Google artificial intelligence designed chips within a few hours
Google said that Google's deep learning enhanced learning (RL) method can generate effective layout within a time less than humans, and the results are equivalent to quality. In March 2022, Google Research launched Prime, which is a deep learning method that uses existing data such as power and delay to create a faster and smaller accelerator design than a traditional method design.
Google researchers use 10,000 chip layout diagrams to train their models. The design of the chip generated by artificial intelligence is less than six hours. Alphabet said that this method has been used to achieve Google's tensor processing unit (TPU), which is part of Google's cloud -based machine learning application.
EDA company double betting artificial intelligence design system
It's not just Google. Electrical design automation (EDA) such as SynopSys and Cadence also uses artificial intelligence technology in its latest tools. Recently, SynopSys has registered 100 commercial stream films with its SynopSys dso.ai autonomous chip design system.
ST and SynopSys first used DSO.AI to design working chips on the Microsoft Cloud in early February 2022. The tool uses SynopSys's DSO.AI design system, combined with the SynopSys Fusion Compiler and IC Compileer on Microsoft Azure, to increase power consumption, performance and area (PPA) indicators by more than 3 times, reduce the total power consumption by 25%, and the volume of volume Significantly narrowed. size.
In addition to chip design, artificial intelligence is also applied to chip testing and verification -chip manufacturers spend a lot of time in these two fields. To solve the design problem at this stage, Siemens launched Questa Verification IQ, a software platform that helps IC design engineers to speed up the verification process.
NVIDIA is used in GPU design GPU -driven AI
NVIDIA designed another deep learning method for chip design. The company produced a RL model called Prefixrl, which proves that artificial intelligence can learn circuit design from scratch and use the latest EDA tools to manufacture smaller and faster circuits. The architecture of NVIDIA consists of 13,000 circuits for AI technology.
For automated cell migration, the company has developed NVCell, which can migrate 92%of cell libraries without mistakes. Human beings can study the remaining 8% of the cells that cannot be automatically migrated. NVIDIA chief scientist Bill Dally said:
"So this is like an ATARI video game, but it is a video game that repairs the design rules in the standard unit. Through strengthening learning to check and repair these design rules errors, we can basically complete our standard design cells."
The impact of artificial intelligence on the IC design employment market
Although artificial intelligence can automatically perform certain tasks performed by IC designers, such as layout design and optimization, it also reduces the needs of manual manual in the design process. Although this has improved overall efficiency, it may also lead to work positions in some fields in some areas of IC design.
On the other hand, artificial intelligence can also help IC designers to complete the work more efficiently and effectively. For example, artificial intelligence can analyze a large amount of data and provide insights, and provide suggestions for the design plan that engineers may have not considered before. This trend can increase the value of IC designers in the industry, enable them to focus on more complex and creative design, and eventually produce better products.
Artificial intelligence is unlikely to completely replace the demand for skilled IC designers. As artificial intelligence becomes more and more common in the industry, the demand for such engineers may even increase, as individuals who need to accurately verify and use artificial intelligence tools and algorithms in the design process.
Google artificial intelligence designed chips within a few hours
Google said that Google's deep learning enhanced learning (RL) method can generate effective layout within a time less than humans, and the results are equivalent to quality. In March 2022, Google Research launched Prime, which is a deep learning method that uses existing data such as power and delay to create a faster and smaller accelerator design than a traditional method design.
Google researchers use 10,000 chip layout diagrams to train their models. The design of the chip generated by artificial intelligence is less than six hours. Alphabet said that this method has been used to achieve Google's tensor processing unit (TPU), which is part of Google's cloud -based machine learning application.
EDA company double betting artificial intelligence design system
It's not just Google. Electrical design automation (EDA) such as SynopSys and Cadence also uses artificial intelligence technology in its latest tools. Recently, SynopSys has registered 100 commercial stream films with its SynopSys dso.ai autonomous chip design system.
ST and SynopSys first used DSO.AI to design working chips on the Microsoft Cloud in early February 2022. The tool uses SynopSys's DSO.AI design system, combined with the SynopSys Fusion Compiler and IC Compileer on Microsoft Azure, to increase power consumption, performance and area (PPA) indicators by more than 3 times, reduce the total power consumption by 25%, and the volume of volume Significantly narrowed. size.
In addition to chip design, artificial intelligence is also applied to chip testing and verification -chip manufacturers spend a lot of time in these two fields. To solve the design problem at this stage, Siemens launched Questa Verification IQ, a software platform that helps IC design engineers to speed up the verification process.
NVIDIA is used in GPU design GPU -driven AI
NVIDIA designed another deep learning method for chip design. The company produced a RL model called Prefixrl, which proves that artificial intelligence can learn circuit design from scratch and use the latest EDA tools to manufacture smaller and faster circuits. The architecture of NVIDIA consists of 13,000 circuits for AI technology.
For automated cell migration, the company has developed NVCell, which can migrate 92%of cell libraries without mistakes. Human beings can study the remaining 8% of the cells that cannot be automatically migrated. NVIDIA chief scientist Bill Dally said:
"So this is like an ATARI video game, but it is a video game that repairs the design rules in the standard unit. Through strengthening learning to check and repair these design rules errors, we can basically complete our standard design cells."
The impact of artificial intelligence on the IC design employment market
Although artificial intelligence can automatically perform certain tasks performed by IC designers, such as layout design and optimization, it also reduces the needs of manual manual in the design process. Although this has improved overall efficiency, it may also lead to work positions in some fields in some areas of IC design.
On the other hand, artificial intelligence can also help IC designers to complete the work more efficiently and effectively. For example, artificial intelligence can analyze a large amount of data and provide insights, and provide suggestions for the design plan that engineers may have not considered before. This trend can increase the value of IC designers in the industry, enable them to focus on more complex and creative design, and eventually produce better products.
Artificial intelligence is unlikely to completely replace the demand for skilled IC designers. As artificial intelligence becomes more and more common in the industry, the demand for such engineers may even increase, as individuals who need to accurately verify and use artificial intelligence tools and algorithms in the design process.
For more electronic components requirements,pls see:https://www.megasourceel.com/products.html