In addition to Intel (Intel) minority deep-pocketed chip manufacturers, have the ability to continue into miniature chip, the majority of manufacturers have been unable to pay for rising costs. As Moore's law has gradually become a thing of the past, these chip manufacturers are turning to neural networks, and the driving, and the Internet of things needed for the development of special processing technology.
According to a report VentureBeat, chip manufacturing slowdown, let the designers have to seek breakthrough in the field of architecture. Analysts pointed out that the special chip design, performance per watt will improve 10 ~ 100 times. Only before production of general-purpose processor, able to respond to industry chip plant money, and now the importance of specialized processors is becoming more and more attention.
Due to the rise of image recognition, the importance of visual processing technology in recent years. In addition to the driving vision processor by observing all round, the computer needs they recognize faces, game consoles also need them to resolve. At present including Cadence, Ceva, Synopsys, VeriSilicon in visual processor development. Freescale (Freescale) and Intel are acquired respectively Cognivue and Movidius, hoping to gain a foothold in this market.
Artificial intelligence and brain network become a distinguished joint promote the development of the neural network processor. Google push specially developed for TensorFlow customized chips; A Wave of Computing, NVIDIA will launch a new artificial intelligence chips; IBM will be treated as class brain chip used in artificial intelligence; Intel acquired Nervana improve the artificial intelligence Systems.
Gwennap believe that with the popularity of the Internet of things, can chip makers such as satisfaction sensors to the data processing and other aspects of the requirements, was equal to master the key to winning. In addition, many cities have set intelligent street lamp and intelligent parking meters, so the demand for the processor will only grow.
Gwennap pointed out that the development of self-drive will likely need the combination of neural network and multiple processors. He predicted that the first partial automatic vehicle will came out in 2018, and by 2022, the car electronic hardware material listing costs will be likely to rise to $5000.
Synopsys ARC processor Fergus senior r&d manager Casey also said that the Internet of things will create enormous business opportunities, and note that the processor and the sensor information, will become particularly important. In addition to do layers of protection, no one can be a few methods.
At the same time, the server is still the Intel chip market all over the world. Gwennap pointed out that the server vendors are reluctant to over-reliance on a single supplier, under the condition of market competition need to inject new vitality. IBM, ARM with AMD (AMD) are hoping to beat the Intel, continuously introduce more excellent server chip design power efficiency. Qualcomm, Qualcomm and Broadcom (Broadcom) is expected to launch in 2017 the ARM architecture server chips.
