Shulou(Shulou.com)12/24 Report--

Meteor Lake processor tray at the Arizona factory: Intel is here to develop packaging technology for laptop electronic brains.

Intel unveiled and detailed the new Meteor Lake processors at the December 14 "AI Everywhere" event, which the company hopes will transform laptops into "AI PC" and Intel uses AI to make these processors.

AI tools can catch errors early in the design process, allowing chips to enter the market faster, and they can oversee the manufacturing process, allowing more tiny wafers made by Intel to end up inside the product. For Intel's first large processor Meteor Lake,AI tool with multiple cores stacked into a package, it can help find the best way to sell each processor, because there are many nuances in the components of each processor.

With these AI tools, "the number of marketable units is increasing significantly," Shlomit Weiss, co-head of Intel Design and Engineering Group, said in an interview.

This is crucial for Intel as it tries to regain the lead in chip manufacturing lost to TSMC. If you are one of the hundreds of millions of users who use Windows PC, this is also important to you, because better manufacturing means better performance and lower cost.

Today's processors consist of tens of billions of tiny electronic switches called transistors. Messages are sent in layers of ones and zeros, using them to spell check your resume, calculate your income tax, and sharpen photos on your phone. Not anyone can know how to best arrange all these transistors and check whether they work on demand, so computers have always been an integral part of the processor manufacturing business.

But using AI is a new attempt. After training, the AI system can recognize patterns in complex real-world data, and can perform tasks such as filtering spam and identifying faces in mobile camera film. Now, they are helping humans better control the production of processors.

With these AI tools, "the number of salable units has increased significantly," Shlomit Weiss, co-head of Intel's design and engineering team, said in an exclusive interview.

An Intel engineer holds a Meteor Lake processor at an Arizona factory where Intel develops product packaging technology.

"with the development of Moore's Law, there are more and more integrated transistors," said David Kanter, an analyst at Real World Insights. "you can't adjust everything manually."

Intel is not alone. Cadence and Synopsys, the two major manufacturers of software tools for designing processors, both provide AI enhancement tools for laying out processor components.

One of the AI used in this technique is called reinforcement learning. In this technology, the AI system explores a range of options and is rewarded for finding results that are closer to the desired goals. "reinforcement learning has the potential to get humans out of this cycle," said Elyse Rosenbaum, an engineering professor at the University of Illinois. "

Weiss and her co-leader, Navid Shahriari, say Intel is combining AI tools from external chip designers with tools developed internally. Some examples of internal tools:

When ● simulates a new design, Intel AI analyzes error reports and classifies errors, which allows human engineers to solve problems faster. "it can save several people hours a day," Weiss said.

● Intel has begun testing generative AI tools that deal with detailed written specifications of chips to create tests that truly ensure that chip designs meet those specifications. Weiss says people do the work today, but they may make mistakes. Intel is using OpenAI's GPT, a large language model that has been modified to meet Intel's special needs to run internally on Intel's own systems.

● Intel uses AI to review chips on the production line, each with slightly different features, such as clock speed, cache size, and power consumption. Shahriari says AI systems can view this complex set of interrelated details and find the best way to install processors into one of the various versions Intel will sell.

The latest Meteor Lake processor consists of multiple cores, which means there are more features to analyze, so finding the best way to sell chips will be more complicated. Fortunately, AI can find a balance to maximize output, efficiency and power.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.