For decades, constant innovation in the world of semiconductor chip design has made processors faster, more efficient, and easier to produce. Artificial intelligence (AI) is leading the next wave of innovation, reducing the chip design process from years to months by making it fully autonomous.
Google, Nvidia, and others have showcased specialized AI-designed chips, and electronics design automation (EDA) companies have already taken advantage of AI to accelerate chip design. Software publisher Synopsys takes a bigger view: AI-designed chips from start to finish.
AI has already shown its resilience in the semiconductor world, and in the future it could be involved like never before. Here’s how.
Quite fast and fairly inexpensive
Synopsys announced at Hot Chips, an annual semiconductor conference, an extension to its DSO.ai software that can handle the entire chip design process. DSO.ai is already being used by companies like Samsung to design its Exynos chips for smartphones and other smart devices. However, it currently only handles a small set of design challenges. Now, Synopsys says it can handle the process from start to finish.
Synopsys calls it “software-designed hardware,” which is a break from the “software-defined hardware” that chip designers have used for years. “You have tools where you take a chip from specification to final implementation that you ship to a foundry,” said Stelios Diamantidis, senior director of AI solutions at Synopsys.
This process is not new, just extended. DSO.ai is already working on the layout part of the design process, using AI to determine the optimal layout of components for power and performance. Google recently unveiled something similar, where it used AI to design one of its tensor processing units (TPUs). Basically Google used AI to help design a chip that handles AI Nvidia has also invested in machine learning to help design chip configurations.
This advancement alone speeds up the design process, reducing what would normally take several weeks to days – and usually with a performance advantage. Synopsys took this idea and developed it. “Can we really design chips fast enough and cheap enough to be personalized?” Said Diamantidis.
Time and money are barriers in the semiconductor design world, and they have been barriers for new space-interested companies. This is not only important for the chips inside processors, graphics cards and other PC components, but also for the millions more that are needed for smart devices, medical equipment and cars. , to name a few.
Synopsys uses a model called reinforcement learning, which trains the machine to get the best reward for a task. This model does not identify patterns in datasets or predict outcomes. Rather, it finds the optimal path given large amounts of data through a reward system.
It’s the same model Google used to train AI to win at chess and go, but infinitely more complex. “You are looking at an unfathomable number of states, far more than atoms in the known universe,” Diamantidis said. The DSO.ai solution only addresses one point in the design process, and it’s already “billions of times more complex” than training AI in chess or go.
In terms of what AI can do for chip design, saving time is the biggest. Synopsys says what would normally have taken two years in the past can now be done in as little as three to six months. He claims it will help companies create chips faster, cheaper, and for more specialized purposes.
Synopsys has a portfolio of software that “literally everyone uses to design chips”. While this claim may be hyperbole, it is true that Synopsys touches most areas of the semiconductor industry. Its website says Synospsys software is the source of 90% of FinFET chip designs worldwide, and that’s only a small part of what the company does.
Besides being faster and cheaper, Synopsys claims that the chips designed by AI are more efficient. At a press briefing, the company said it has achieved energy savings of up to 26% using the AI model compared to a human engineer, which is a bigger gain than switching to a new manufacturing process. “When people move from one manufacturing process to the next manufacturing process, let’s say from the 7nm design to the 5nm design, the scale factor they are looking for (is 20% at best),” said Diamantidis.
Concretely, according to Karl Freund, industry analyst and founder of Cambrian AI Research, this will lead to faster innovation. “Over the next few years, therefore, we will benefit from a faster pace of innovation and better products with longer battery life,” Freund wrote in an email to Digital Trends.
The wave of AI
While Synopsys is leading the charge, it isn’t the only company looking to AI for future chip design. Cadence, a competitor to Synopsys, has its Cerebrus tool which helps engineers automate several points of the chip design process. However, Synopsys and industry analysts say this tool is still lagging behind. “Synopsys is at least 18 months ahead in this area, maybe more,” said Freund of Cambrian AI Research.
Larger companies, such as Google and Nvidia, have also invested in this space, although they are unable to provide a tool that manages the design of the chip from start to finish, according to Synopsys. “You can’t really build a self-driving car unless you build a car first,” said Thomas Andersen, vice president of AI and machine learning at Synopsys.
We’re still in the early stages of AI-designed chips. Diamantidis believes adoption will come in several waves, the first of which will help companies optimize their existing design workflows. By the third wave, Synopsys hopes to open “the door to, really, people who aren’t chip design experts.”
Over the past few years, there has been a buzz surrounding the idea of Moore’s Law, where chips become faster and more efficient over a cycle of about two years. AI appears to be leading the charge in a new era of Moore’s Law, helping more companies create chips that are faster and more efficient, and at speeds that would otherwise be impossible.
For fully AI-designed chips, it’s a matter of when, not if. Many of the world’s largest semiconductor companies are already using AI at various stages of the design process, and some of the world’s largest tech companies have invested heavily in the technology. With a tool capable of handling the design from start to finish, chip design could be pushed over the edge in a revolution.