How will the chips design change

The end of Moore’s law is coming. Engineers and designers can only do so much to miniaturize transistors and pack as many as possible in chips. Then they are resorting to other approaches to the design of chips, incorporating technologies such as AI in the process.

Samsung, for example, is adding to its memory chips to allow memory processing, thus saving energy and accelerating automatic learning. Speaking of speed, the TPU V4 AI chip of Google has doubled its processing power compared to that of its previous version.

But AI is even more promise and potential for the semiconductor industry. To better understand how AI is configured to revolutionize chip design, we talk to Brozo GorrSenior Product Manager for MathWorks‘Matlab Platform.

How is AI currently used to design the next generation of chips?

Heather Gorr: AI is such an important technology because it is involved in most parts of the cycle, including design and manufacturing process. Here are many important applications, even in general process engineering, where we want to optimize things. I think that defect detection is large in all phases of the process, especially in manufacturing. But even thinking about the design process, [AI now plays a significant role] When you are designing light and sensors and all the different components. There is a lot of abnormalities and fault mitigation that you really want to consider.

Brozo GorrMathWorks

Then, thinking about the logistics modeling you see in any industry, there is always planned inactivity time that you want to mitigate; But you also end up having the time of unplanned inactivity. Then, looking back in those historical data when he has had those moments when he took a little longer than expected to manufacture something, he can take a look at all those data and use AI to try to identify the immediate cause or See something that could even jump into the processing and design phases. We often think of AI as a predictive tool, or as a robot that does something, but many times you get a lot of information from the data through AI.

What are the benefits of using AI for chips design?

GORR: Historically, we have seen a lot of model based on physics, which is a very intensive process. We want to do a Reduced order modelWhere instead of solving such an expensive and extensive computationally expensive model, we can do something a bit cheaper. You can create a substitute model, so to speak, of that physics -based model, use the data and then make your Parameter barYour optimizations, you MONTE CARLO SIMULATIONS Using the substitute model. That takes much less computationally than solving the equations based on physics directly. Therefore, we are seeing that benefit in many ways, including efficiency and economy that are the results of rapidly iterating in experiments and simulations that will really help in design.

So is it like having a digital twin in a certain sense?

GORR: Exactly. That is more or less what people are doing, where you have the physical system model and experimental data. Then, together, it has this other model that could adjust and tune in and try different parameters and experiments that allow to sweep all these different situations and in the end the best design.

So will it be more efficient and, as you said, cheaper?

GORR: Yes, definitely. Especially in the phases of experimentation and design, where you are testing different things. Obviously, that will give dramatic cost savings if you are really manufacturing and producing [the chips]. You want to simulate, try, experiment as much as possible without making something use real process engineering.

We have talked about the benefits. What about the inconveniences?

GORR: He [AI-based experimental models] They tend not to be as precise as physics -based models. Of course, that is why you make many simulations and sweeps of parameters. But that is also the benefit of having that digital twin, where it can take it into account: it will not be as precise as that precise model that we have developed over the years.

Both chips design and manufacturing are intensive in the system; You have to consider every small part. And that can be really challenging. It is a case where you can have models to predict something and different parts of it, but you still need to unite everything.

One of the other things to think about is that you need the data to build the models. You must incorporate data from all kinds of different sensors and different types of equipment, and that increases the challenge.

How can engineers use AI to prepare and extract better information from hardware or sensor?

GORR: We always think about using AI to predict something or do a robot homework, but you can use AI to create patterns and choose things that you may not have noticed before your own. People will use when they have high frequency data from many different sensors, and it is often useful to explore the frequency domain and things such as data synchronization or the remake. Those can really be challenging if you are not sure where to start.

One of the things I would say is that he uses the available tools. There is a great community of people working on these things, and you can find many examples. [of applications and techniques] in GITHUB either MATLAB CentralWhere people have shared pleasant examples, even small applications they have created. I think many of us are buried in data and we are not sure what to do with it, so you definitely take advantage of what is already available in the community. You can explore and see what makes sense to you, and bring that balance of knowledge of domain and the vision you get from the tools and AI.

What should engineers and designers consider?In use of AI for chips design?

GORR: Think about what problems you are trying to solve or what ideas you can expect to find, and try to be clear about it. Consider all the different components and document and try each of those different parts. Consider all people involved and explain and deliver in a sensible way for the entire team.

How do you think AI will affect the work of chips designers?

GORR: It will release a lot of human capital for more advanced tasks. We can use AI to reduce waste, to optimize materials, to optimize the design, but then that human involved still has every time it is about decision making. I think it is a great example of people and technology that work in hand. It is also an industry in which all the people involved, even on the manufacturing floor, need to have a certain level of understanding of what is happening, so this is a large industry to advance in AI due to how we prove the Things and how we think about them before we put ourselves. They in the chip.

How do you imagine the future of AI and chips design?

GORR: It depends a lot on that human element: among people in the process and have that interpretable model. We can do many things with the mathematical minute of modeling, but it is about how people use it, how everyone in the process understands and apply. The communication and participation of people of all levels of skill in the process will be really important. We are going to see less of these superprecise predictions and more transparency of information, exchange and that digital twin, not only using AI, but also using our human knowledge and all the work that many people have done over the years.