News

In the face of the blockade from the United States, Huawei also chose a positive response, indicating that they have a plan B, have spare chips, and continue to serve customers in an environment of extreme survival!At the beginning, Huawei`s desperation to throw into Haisi was not a fever. Looking at it now, this approach is very far-sighted. In the light of the recent situation, I believe everyone agrees? What does it mean to belong to your own chip? Lower R&D and manufacturing costs, more optimistic bargaining power, and more reliable supply guarantees. Each one has made countless mobile phone manufacturers envy and hate. Huawei started its ASIC design center in 1991 and established HiSilicon in 2004 until it became a representative of China's independent chip design. So what chips does Huawei make? From a broad perspective, Huawei has designed five types of chips: 1, SoC chip 2, AI chip 3, server chip 4, 5G communication chip Other dedicated chips. 1.SoC chip (Kirin series): The mobile phone SoC chip has always been the main research of Huawei. The Kirin 980 processor launched on August 31, 2018 and the Kirin 985 chip is expected to be launched in the second half of this year. Huawei's mobile phone chip has reached the world-class level. Level. 2.AI chip (Shengteng series): On October 10, 2018, at the HC conference of Huawei, two AI chips, the Shengteng 910 and the Shengteng 310, were released, using 7nm process technology and 12nm process technology respectively. The Shengteng series AI chip adopts Huawei's groundbreaking unified and scalable architecture, namely [Da Vinci Architecture", which realizes full coverage from the ultimate low power consumption to the ultimate high power computing scene. 3. Server chip (): Huawei optimized and designed the technology provided by its partner ARM. On January 7, 2019, it released the Kunpeng 920 and the Taishan server and Huawei cloud service based on the Kunpeng 920. 4. 5G communication chip (Barong, Tianzhu series): Huawei's 5G chips are mainly divided into terminal chips (Baron series) and base station chips (Tianlu series). The Baron series is a baseband chip for mobile terminals, and has been a dedicated chip for Huawei mobile phones. On January 24, 2019, Huawei introduced the industry's first 5G base station core chip (Scorpio chip) and 5G multimode terminal chip (Barong 5000). 5.Other dedicated chips: (router chip, NB-IoT chip, IPC video codec and image signal processing chip, etc.): The Lingxiao series is mainly used for home access products; the Lierda NB-IoT module is The world's leading narrowband IoT wireless communication module; IPCSoC chip covers the core technology of video surveillance - ISP technology and video codec technology. 1.1 Kirin chip: the world's leading domestic mobile phone SoC chip At present, the leading manufacturers of mobile phone chips on the market are Apple, Huawei, Qualcomm and Samsung. Apple A series: Apple's A12 processor is the world's first 7-nanometer chip, performance is the first, but because Apple does not have independent baseband technology, GPS / WIFI chips need to be outsourced, so A series chips never contain The baseband part does not need to assume the function of GPS/WIFI; Huawei Kirin series: Huawei's Kirin 980, using the most advanced eight-core design, the highest frequency up to 2.8GHz. It is expected that Kirin 990 will also inherit the 5G baseband to achieve true 5G full Netcom; Qualcomm Xiaolong Series: released the Snapdragon 855 processor, which uses the new Kryo 485 architecture, 7 nanometer process technology, graphics rendering increased by 20%, CPU performance increased by 45%, the highest frequency is 2.84GHz, is also very powerful; Samsung Exynos series: Exynos 9810 processor is Samsung's independent research and development of M3 architecture, with four 2.9GHz M3 big core and four 1.9GHz A55 small core, 10 nanometer production process. 1.2 Newcomers to break the monopoly of the server field In addition to the 920 processor, Huawei also launched three TaiShan series servers, using the 920, including TaiShan 22080, Taishan 5280/5290, and TaiShan X6000, respectively for the balanced server, storage server and high-density server market. 1.3 Shengteng Chip: Full Stack Full Scene AI Solution In 2018, Huawei released two cloud data center AI chips: the single chip to calculate the highest density of the rising 910 and the extremely efficient low-power AI SoC rising 310. The Ascension 910 belongs to the Ascend Max series and is used in data center servers. Its performance exceeds that of NVIDIA's strongest chip AI V100. It is the world's largest single-chip computing density AI chip. Based on the Shengteng 910 Huawei will also launch a large-scale distributed training system to upgrade the Cluster, linking 1024 rising 910 chips to form an AI computing group, providing ultra-high-level AI computing power, computing power up to 256P, enabling AI training speed Achieve new levels. The Shengteng series AI chip adopts Huawei's groundbreaking unified and scalable architecture, namely [Da Vinci Architecture", which realizes the full coverage from the ultimate low power consumption to the ultimate high power computing scene. No other architecture can do it. 1.4 5G chip: base station and terminal all-round layout 5G refers to the fifth-generation mobile communication technology, which is an extension after 4G. Its peak theoretical transmission speed can reach several tens of Gb per second, which is hundreds of times faster than 4G network transmission. At the same time, 5G also has milliseconds. The transmission delay and the connection capacity of hundreds of billions are the communication foundation for opening the Internet of Everything and deep interaction between human and machine. Huawei released the first 5G terminal product based on the Baron 5000 chip: 5G CPE Pro. This is the fastest 5G CPE in the world and supports Wi-Fi6 technology. The main application scenario is smart home. 5G CPE Pro can support 4G and 5G dual mode. It can download 1GB HD video in 3 seconds under 5G network. Even 8K video can be used to open the second time. It sets a new network speed standard for small CPE.

The bipolar tube is a current-controlled device (which controls a large collector current through a small base current), and the MOS transistor is a voltage-controlled device (the on-resistance between the source and the drain is controlled by the gate voltage). MOSFET (field effect transistor) under the conduction voltage drop, the on-resistance is small, the gate drive does not require current, the loss is small, the drive circuit is simple, the self-protection diode, the thermal resistance characteristics are good, suitable for high-power parallel, fault switch The speed is not high and it is expensive. The triode switch speed is high, the Ic of the large triode can be made very large, the disadvantage is large, the base drive current is large, and the drive is complicated. Generally speaking, in low-cost applications, triodes are considered for common applications, and MOS tubes are not considered. In fact, it is wrong to say that current control is slow and voltage control is fast. It is only necessary to understand the way in which bipolar and MOS transistors work. The triode is operated by the movement of carriers. Taking the npn tube emitter follower as an example, when the base is not applied with voltage, the pn junction composed of the base region and the emitter region is blocked by multiple ions (the base region is a hole). The emission region is an electron diffusion motion. At this pn junction, an electrostatic field (ie, a built-in electric field) directed from the emitter region to the base region is induced. When the base plus a positive voltage is directed to the base region, the emitter region is pointed to the emitter region. When the electric field generated by the applied voltage of the base is greater than the built-in electric field, the carrier (electron) in the base region may flow from the base region to the emitter region. The minimum value of this voltage is the forward conduction voltage of the pn junction (it is generally considered in engineering) 0.7v). However, at this time, there is a charge on both sides of each pn junction. At this time, if the collector-emitter is positively charged, the electrons in the emitter region move toward the base region under the action of the electric field (actually, the electrons move in the opposite direction. ), because the width of the base region is small, electrons easily cross the base region to reach the collector region, and recombine with the holes of the PN here (near the collector), in order to maintain the balance, under the action of the positive electric field, the collector region The electron accelerates the movement of the outer collector, while the cavity moves at the pn junction, a process similar to an avalanche process. The collector's electrons return to the emitter through the power supply, which is how the transistor works. When the triode is working, both pn junctions induce charge. When the switch is in the on state, the triode is saturated. If the triode is turned off, the charge induced by the pn junction should be restored to equilibrium. This process takes time. The MOS triode works differently and does not have this recovery time, so it can be used as a high speed switch. (1) The field effect transistor is a voltage control element, and the transistor is a current control element. In the case where only a small current is allowed from the signal source, the FET should be used; and when the signal voltage is low and the current is taken from the signal source, the transistor should be selected. (2) FETs use majority carriers to conduct electricity, so they are called unipolar devices, and transistors have both majority carriers and minority carriers. It is called a bipolar device. (3) The source and drain of some FETs can be used interchangeably. The gate voltage can also be positive or negative, and the flexibility is better than that of the transistor. (4) The FET can work under very low current and very low voltage conditions, and its manufacturing process can easily integrate many FETs on one silicon chip, so the FET is in large scale integrated circuit. It has been widely used. (5) Field effect transistors have the advantages of high input impedance and low noise, and are therefore widely used in various electronic devices. In particular, the field effect tube is used as the input stage of the entire electronic device, and the performance that is difficult to achieve by a general transistor can be obtained. (6) The FET is divided into two types: junction type and insulated gate type, and the control principle is the same. Other comparisons: 1. The triode is a bipolar tube, that is, the inside of the tube is operated by two carriers, a hole and a free electron. The FET is a unipolar tube, that is, the tube works with only holes, or only free electrons participate in conduction, only one type of carrier 2. The triode is a current control device that has an input current to have an output current. FETs are voltage-controlled devices that have an output current without input current. 3. The input impedance of the triode is small, and the input impedance of the FET is large. 4. Some FETs have source and drain interchangeable, and the collector and emitter of the transistor are not interchangeable. 5. The frequency characteristics of the FET are not as good as those of the triode. 6. The FET has a small noise figure and is suitable for the pre-stage of the low noise amplifier. 7. If you want the signal source current to be small, you should use the FET. Otherwise, it is more suitable to use the triode.