It is reported that the big coffee in the AI ​​field is concerned about the research of brain-like chips. Compared with traditional chips, brain-like chips do have absolute advantages in power consumption. Researchers try to make the machine surpass the human brain by simulating the human brain operation mechanism. It is reported that with the advancement of technology, brain-like chips that subvert the traditional architecture have opened a new door for the chip industry.
Recently, in the field of AI, whether it is a big business in the academic world or an amnesty in the industry, the research on brain-like chips is in full swing, and of course, many achievements have been made. Recently, Professor Jeehwan Kim of the Electronic and Microsystems Technology Laboratory of Stanford University Research Institute published a paper in Nature, which attracted the attention of the industry, academia and research community. The reason is that Professor Jeehwan Kim and researchers have developed an artificial synapse chip using a material called silicon germanium to support machine learning algorithms that recognize handwritten fonts. Coincidentally, the brain information processing (BRAVE) research group of the Brain Intelligence Research Center of the Institute of Automation of the Chinese Academy of Sciences has also made breakthrough research in the neural network modeling and human-like learning research of biological neural structures.
Since the birth of the computer, people have been demanding more computing power. As the chip integration becomes higher and higher, the performance gap between CPU and memory is getting larger and larger. The shortcomings of the computer structure based on the von Neumann structure are also becoming more and more obvious. Some people call this a memory wall, which means that the CPU is faster and has to wait for memory. In contrast, the human brain does not have such problems. According to research, the human brain can perform an operation of 100 billion operations per second on average, and the energy required is only 10 to 25 watts. Therefore, researchers are turning to simulating human brain research, trying to simulate the human brain to make the computer perform calculations with low energy consumption and high efficiency, and even make the computer superior to human intelligence.
Many companies and organizations at home and abroad are investing a lot of energy in the research and development of brain-like chips. The United States started earlier in this research. In 2014, IBM launched the industry's first brain-like chip, TrueNorth. In recent years, domestic research and development of chips has not been shown to be weak. There are also startups such as Xijing Technology who have invested in the research and development of brain-like chips. Tsinghua and other well-known universities have also established brain-like research centers.
Compared with traditional chips, brain-like chips do have absolute advantages in power consumption. For Loihi, the self-learning chip that Intel exhibited at this CES, not only is its learning efficiency 1 million times higher than other smart chips, but also The energy consumed to complete the same task is nearly 1000 times less than the traditional chip. The integration of brain-like chips is also very high. With the "Darwin" chip introduced by Zhejiang University, its area is 25 square millimeters, which means that the side length is only 0.5 cm, but the internal can contain 5 million transistors. As the industry demands more and more computing power, Feng's bottleneck will become more and more obvious. The brain-like chip that subverts the traditional architecture has opened a new door for the chip industry.
First, the traditional chip encounters the von Neumann bottleneck to simulate the neuron into a new ideaModern computers are basically based on the von Neumann structure, which stores the program and the data that processes the program in two ways in the same way, one called the instruction set and one called the data set. Each time the computer performs an operation, it needs to be called back and forth in the two areas of CPU and memory, thus generating data traffic between the two parties. With the advent of deep learning algorithms, the computational power requirements of the chip continue to increase, and the von Neumann bottleneck is obvious: when the CPU needs to execute some simple instructions on huge data, the data traffic will seriously reduce the overall efficiency, CPU Will be idle when data is input or output.
Not only that, but the big problem with traditional chips is that they are inefficient. When the chip is in operation, most of the power is converted into heat. A computer without a heat sink can melt the heat generated by the CPU in a short time. Other intelligent devices are also due to the high energy consumption of the chip, resulting in poor endurance. No matter how to improve the process, high temperature and leakage are difficult to avoid.
In order to solve the problem of high efficiency and low energy consumption of CPU in large amount of data, there are currently two development routes: First, the traditional von Neumann architecture is used, mainly represented by 3 types of chips: GPU, FPGA, ASIC; Human brain neuron structure design chip to improve computing power, has been completely anthropomorphized as a goal, and pursues the human brain in the chip architecture, such chips are called brain-like chips.
After receiving stimulation, human brain neurons will change the distribution of charged ions inside and outside the cell membrane, thus forming a potential difference. The potential difference will be bidirectionally transmitted along the axons and dendrites of the nerve cells to form a pulse current. When the electrical signal is transmitted to the synapse, the presynaptic neurons release neurotransmitters (such as dopamine, adrenaline) from the postsynaptic neurons to receive neurotransmitters to generate excitement (the process is one-way), and The lower transfer acts with the human reactor and reacts.
The brain-like chip architecture is a synaptic transmission structure that mimics the human brain. Many processors are similar to neurons, and the communication system is similar to nerve fibers. The calculation of each neuron is performed locally. From the whole, the neurons work in a distributed manner, that is, the overall task is divided. Each neuron is only responsible for a portion of the calculation. This approach has significant advantages in processing massive amounts of data and consumes less power than traditional chips. For example, IBM's TrueNorth chip consumes only 20 milliwatts per square centimeter.
Laptop power adapter charger for Macbook, Macbook Air and Macbook Pro.
45w, 60w, 80w charger with Magsafe 1 or Magsafe 2 connector.
Four replacement ac plugs for your choices, US / EU / UK / AU plug as your requirement.
Stable output and high charging efficiency.
Elegant outlook design as original one, touch smoothly and comfortable.
Easy carry if you want to take your computer outside for business or travel.
Original charger is good, but as a replacement, our product has more reasonable price when your original charger is broken.
And, the market of the replacement adapters becomes bigger and bigger. People would rather buy a copy one then the original because of the price.
But at the same time, people worry about that they will buy something defective. So the problem comes, how to buy a good quality one with a good price?
As a professional power adapter manufacturer, we have excellent R&D team, skilled staffs and responsible after-sale service. All your benefits can be under protected after you buy products for our company.
Our certificates :ISO9001:2008 & ISO14001:2004 , CCC , CE , FCC , ROHS.
All our products has 1 year warranty. In other words, if you get the dad products which are not damaged physically from us in one year, we will replace you the new one or the whole bulk order.
Macbook Adapter,45W Adapter For Macbook,45W Charger For Macbook,60W Charger For Macbook,Macbook Charger
Shenzhen Waweis Technology Co., Ltd. , https://www.waweispowerasdapter.com