More than 40 years, the iron discipline of Moore's law has been the IT industry, but now after the difficulty of Moore's law is more and more big, recently, the United States team developed a nanoscale transistors for chip makers would be tantamount to "ultimatum". This under a lot of people are not calm, media ji "Moore's law" sing ".

How will they self-salvation?

Developed a nanometer transistor, a researcher at the Ali Javey and Sujay Desai

Recently, the Lawrence Berkeley National Laboratory, Lawrence Berkeley National Laboratory, a team of researchers, has successfully developed the grid (within the transistor current is controlled by the grid) only 1 nm long transistor, claims to be the smallest transistors.

As is known to all, over the years, the development of the computer industry always follows the Moore's law. Moore's law thought, the number of semiconductor circuits of transistors is doubling every 18 to 24 months. And the size of the transistor is very important for the improvement of computer technology. The smaller the transistor, a single chip can accommodate the more the number of transistors; The number of transistors on the chip, the more the faster processor speed, the higher the efficiency. The same price of the electronic product performance, after 18 to 24 months after will be doubled. But conversely, if a IT company today and 18 months ago to sell the same, the same product, its sales will be halved.

That is why, for decades, chip makers have been struggle with Moore's law hard - you can't make a smaller transistors, survival is under threat. But now, it may be to rewrite it all.

In more than 40 years, Moore's law has always been the iron discipline of IT industry. Since the 21st century, however, Moore's law seems to be a "slow" sign.

With the further development of the chip technology, Moore's law gradually meet the laws of physics limit. Industry is generally believed that 7 nanometer silicon transistor is a barrier, once after this node, will encounter problems. Once because of the silicon transistor gate is less than 7 nm, electrons can flow between different transistor, a phenomenon called Quantum Tunneling effect (Quantum Tunneling), it means that the transistor may accidentally opened are supposed to be closed condition.

But even more than 7 nano transistors, still face difficulties from theory to practice. At present the most advanced chip manufacturing process USES is 14 nanoscale transistors, and use of the 10 nanometers transistor processor, such as Intel Cannonlake processor, is expected to launch until 2017 or 2017.

Here are some references: a human hair is about 100000 nanometers thick, a red blood cell diameter of 6000 nanometers, even a strand of DNA is 2.5 nm.

At the same time, the demand for higher-end chip soared. Today, we produce the amount of data is increased exponentially, and emerging technologies such as machine learning, is computing power put forward higher requirements for the chip.

Lawrence Berkeley national laboratory is undoubtedly a great scientific research achievements, because this means that a single chip to accommodate a doubling of the number of transistors can still, at least in theory.

However, the study is still in its early stages, at present is just a concept, to be a real products have a long way to go. In terms of 14 nm transistor, a single chip with the number of transistors more than billion, and Lawrence Berkeley lab team hasn't found can be mass production 1 nm transistor and the new chip method, even if found, its cost is extremely expensive.

Of course, even if only as a concept, to the research results are still important, it proved that we can use new materials to produce smaller size of transistors, to further improve the function and efficiency of the computer.

Chip, in fact, the industry has to accept the reality of the transistor size close to lower limit, and has done for the end of Moore's law. Earlier this year, the U.S. Semiconductor Industry Association (Semiconductor Industry Association, members include Intel, AMD and Global Foundries, published a report. The report claims that by 2021, silicon transistor size reduction will no longer be an economically feasible thing. Instead, the chip will change in another way.

Chip specialization is the comparison of one solution. Traditionally, the industry for all chips are "equally", but now this concept may be behind The Times. Assistant professor at the MIT Sloan school of management, Neil Thompson said: "(such as) we drive the car is not the same, 18 wheel truck, sports car, SUV, the cars have special purpose designed, chip now began to have such a division of labor."

Graphics processor (GPU) is a case in point. GPU was originally designed for rendering graphics to perform complex mathematical and geometrical calculation and design, but now the engineers found that the GPU can also be used for data analysis and other fields.

Chip industry started from high-speed, universal chip to more specialized chips. To this end, Intel acquired visual start-up Movidius chip. Meanwhile, Nvidia is artificial intelligence chip to the global sales.

More efficient chips will also help in reducing the energy consumption on the basis of improve the computing speed. Microsoft and Intel are studying to programming (FPGA) chip, the artificial intelligent algorithm to run more efficiently. Japan's softbank's recent acquisition of the UK ARM chip developers, incredibly to achieve the company's advanced low-power chips, the chip will provide information for rising Internet hardware processing capacity.

And those who do not have a high level of specialization of the processor may change structure, to improve the ability of information processing. For example, using multilayer circuit chips will be more and more, in order to improve the transistor density.