Until alternatives to silicon are found, there is no choice but to place the chips in layers

There is a limit to how many things can be put in hardware on one chip - but who said that a chip has to be two-dimensional?

"We all came to the understanding that the technology of silicon production is reaching the end of its journey, it may still take ten years, but it doesn't matter because it takes quite a bit of time to produce alternatives and people already see the end and are working on the next step. This is the reason why important things are happening in the world of hardware development today." This is what Moshe Levinger, director of the hardware field at the IBM research laboratory in Haifa, says in a special interview withChiportal As part of a visit I made to the IBM development center.

Levinger also refers to the end of the conventional CMOS: "There is a limit to how much things can be put in hardware on one chip - but who said that a chip has to be two-dimensional?" Levinger asks. "Why don't we take several chips and place them one above the other and of course connect them. This is done today by both IBM and Intel. The trick is to connect the layers. That the box be transparent to the chips. We have already demonstrated a solution for this in the prototype. Specifically in Haifa we are working on one of the most difficult problems - the vertical connection between neighboring layers. "

"According to Moore's law, the number of transistors was supposed to double within two years, but you can already see that it is difficult to meet it. We are nearing the end and even before we are enough to deliver the next technology, we see that the value that designers have received from silicon technology in the last decades is starting to diminish. According to Levinger, the competition is currently between the planners (designers), the people who write hardware and there are the silicon people themselves.

"Until today, the designers had an easy life, the performance from the hardware was obtained by the fact that the technologists produce the next generation which will be smaller and faster. The designers can plan what to do with these capabilities. Today the creations have turned upside down: the technologists tell the planners, 'For 40 years we have given you another generation of technology and newer processors have come out - it's over, we don't have your turn right now. You need to build more sophisticated and efficient hardware because the base itself - the silicon will not give you what you are used to.' "

"This overturning of creativity caused, for example, the solution of multiple processors as a substitute for increasing the capacity of computers through 'natural' growth. However, if you use more than one processor, you need to know how to utilize them and be more efficient because managing processors involves overhead. This created a burden on the architects and the content."

"An equally important effect on the planners is the cessation of one-size-fits-all planning. The load profile on the servers varies. The transition to the use of virtualization means that servers will work at a very high output, on the other hand there are new types of processing that were not common until today, for example analytics - receiving information from millions of sensors and processing it in real time to solve problems, learn from history and identify patterns. Each such computer will require a different processing capability, so the computers will become more sophisticated and we will have to build computers for specific applications."

According to Levinger, the next steps are the use of the most powerful chips for cloud computers as well as the development of quantum computers.

More of the topic in Hayadan:

• Development of chips that mimic living cells
• A quarter of Germans are willing to have a chip implanted in their skin if it shortens the queue at the supermarket
• IBM scientists have developed biological components for tiny computers
• DNA as a substitute for semiconductors