According to overseas media reports, since the game of AlphaGo has outperformed the professional chess player's news, the investment in artificial intelligence (AI) has set off a boom, except for the more attention-taking test robot Dongdajun. Another direction that is less valued is the creation of high-performance computing chips that allow artificial intelligence that now requires advanced servers to be executed on a single chip.
For example, Google designed the TPU (Tensor Processing Unit) to allow AlphaGo, which is running on the server during the show, to be executed on the TPU; other vendors such as Intel, Samsung Electronics, Fujitsu, etc. Similar chips are also being developed.
Allows a single chip to perform advanced artificial intelligence, mainly for the Edge CompuTIng part of Cloud CompuTIng. The bottleneck in big data and cloud computing applications lies in the time when data is transmitted from the sensor to the cloud. For applications that require immediacy, such as the Industrial Internet of Things or autonomous driving, this can be a disaster.
Therefore, for the instant application of cloud computing, the marginal operation is proposed. The data is firstly operated on the terminal or the mid-end platform, and processed for the immediate needs. Only the processed data is returned, and the original data does not need to be included, or When the system is idle, the original data is returned, which will significantly reduce the load on the cloud system and improve the computing performance.
Although the marginal operation can be handled by the low-end system, the focus is on the small size and easy access to the terminal to speed up the response, but the lower the efficiency of the marginal operation, the greater the load on the cloud and the network system, and the complex system like autopilot. At present, even the high-end server's computing performance is still insufficient, and it is necessary to have a high-performance marginal computing system.
In addition to automatic control, FPGA and power semiconductor adjustment with artificial intelligence to enhance power application efficiency is another application suitable for marginal computing.
When Softbank President Sun Zheng bought ARM in 2016, he claimed that the purpose of acquiring ARM was to use the Internet of Things and ARM chip technology to create a higher intelligence, equivalent to or even surpass the human brain computing power, pointing out the high performance margin. The operation and cloud computing architecture make the ability of artificial intelligence to a higher level, and the high-performance marginal computing chip is the key point.
According to the US market research agency TracTIcal, the global market for artificial intelligence CPUs is US$1.1 billion in 2016. In 2025, when autopilot will begin to spread, it can exceed US$57.4 billion and grow 52 times. The market potential is extremely amazing.
At present, these chips researched by major semiconductor factories and IT plants are expected to come out in 2017~2018. After 2020, market competition will become fierce, and subsequent development is worthy of continuous observation.
Disclaimer: The electronic reprinted works of E-Commerce Network are as far as possible to indicate the source, and all rights of the owner of the work are not transferred due to the reprint of this site. If the author does not agree to reprint, please inform the site to delete or correct it. Reprinted works may be subject to change in title or content.
Signal
Response time refers to the response speed of the Liquid Crystal Display to the input signal, that is, the response time of the liquid crystal from dark to bright or from bright to dark (the time for the brightness from 10%-->90% or 90%-->10%) , Usually in milliseconds (ms). To make this clear, we have to start with the human eye's perception of dynamic images. There is a phenomenon of "visual residue" in the human eye, and the high-speed motion picture will form a short-term impression in the human brain. Animations, movies, etc. to the latest games have applied the principle of visual residue, allowing a series of gradual images to be displayed in rapid succession in front of people's eyes to form dynamic images. The acceptable display speed of the picture is generally 24 frames per second, which is the origin of the movie playback speed of 24 frames per second. If the display speed is lower than this standard, people will obviously feel the picture pause and discomfort. Calculated according to this index, the display time of each picture needs to be less than 40ms. In this way, for the liquid crystal display, the response time of 40ms becomes a hurdle, and the display above 40ms will have obvious picture flicker, which makes people feel dizzy. If you want the image screen to reach the level of non-flicker, it is best to achieve a speed of 60 frames per second.
I used a very simple formula to calculate the number of frames per second under the corresponding response time as follows:
Response time 30ms=1/0.030=approximately 33 frames per second
Response time 25ms=1/0.025=approximately 40 frames per second
Response time 16ms=1/0.016=approximately 63 frames per second
Response time 12ms=1/0.012=approximately 83 frames of pictures displayed per second
Response time 8ms=1/0.008=approximately 125 frames per second
Response time 4ms=1/0.004=approximately 250 frames per second
Response time 3ms=1/0.003=approximately display 333 frames per second
Response time 2ms=1/0.002=approximately 500 frames per second
Response time 1ms=1/0.001=approximately 1000 frames per second
Tip: Through the above content, we understand the relationship between response time and the number of frames. From this, the response time is as short as possible. At that time, when the LCD market first started, the lowest acceptable range of response time was 35ms, mainly products represented by EIZO. Later, BenQ's FP series came out to 25ms. From 33 to 40 frames, it was basically undetectable, and it was really quality. The change is 16ms, displaying 63 frames per second to meet the requirements of movies and general games, so 16ms is not obsolete. With the improvement of panel technology, BenQ and ViewSonic started a speed battle. ViewSonic started from 8ms to 4ms. Released to 1ms, it can be said that 1ms is the final controversy of LCD speed. For game enthusiasts, 1ms faster means that CS's marksmanship will be more accurate, at least psychologically, such customers should recommend the VX series of monitors. But everyone should pay attention to the grayscale response when selling. The text difference in full-color response may sometimes mean the same thing as gray-scale 8ms and full-color 5ms. It is the same as when we sold CRTs before, we said that the dot pitch is .28, LG just I have to say that he is .21, but the horizontal dot pitch is ignored. In fact, the two are talking about the same thing. LG has come up with a sharpness of 1600:1. This is also a conceptual hype. Everyone uses the basic screen. There are only a few companies on the list, and how can only the LG family achieve 1600:1, and everyone stays at the level of 450:1? When it comes to consumers, the meaning of sharpness and contrast is obvious, just like AMD's PR value, which has no real meaning.
Active Matrix Lcd,Active Matrix Tft Color Lcd,Tft Lcd Color Monitor,Tft Lcd Monitor
Shenzhen Newvision Technology Co.Ltd , https://www.newvisionlcd.com