NVIDIA graphics chips (gpus) companies active in artificial intelligence (AI) chip development, will work with Intel (Intel). 12, Beijing international hotel conference center in mainland China at the "China 2016" GTC conference, NVIDIA latest based on Pascal architecture development in the two processors, respectively, "Tesla P4" and "Tesla P40", mainly for data center market; Other published a self-drive "Drive PX 2" for AI in the field of computers, the future Baidu, Baidu will import own Drive systems.
Two Tesla chip that strengthen the deep learning speed
According to Bloomberg (Bloomberg) and Reuters reported (Reuters), Tesla P4 and Tesla P40 belongs to a kind of deep learning processor, built in 7.2 billion and 12 billion, respectively, transistors, and respectively have 2560 and 3840 unified computing architecture (CUDA) core, two processor performance equal to 40 of the integration of central processing unit (CPU), processing speed of CPU 45 times that of the solution, which combined with Pascal architecture can help the deep learning speed up to 65 times, this let Tesla P4 and Tesla P40 also possess AI processing efficiency.
NVIDIA said, Tesla P4 is aimed at a large number of data center server adopted by the market. Based on Pascal architecture, NVIDIA announced that Tesla P4 efficiency is 40 times quicker than Intel server chips, also from the previous generation of products on the image processing speed more than 3 times quicker. Tesla P40 mainly for super computer such processing performance more powerful single computer market.
NVIDIA claims the company the GPU is the right choice of AI system, and has set up a investment development can let the GPU is more easy to use software. Intel is published on August first an own AI chip. NVIDIA and Intel all want to make sure that Google data center operator will not only by its chip technology, and don't want to try to develop its own customized solutions.
Release Drive 2 PX self-drive AI computer baidu will be adopted
NVIDIA also published 2 self-drive AI computer Drive PX, claimed that only 10 watts of power consumption, compact in January 2016, the company introduced on the prototype style version of the small half, to take the help of AI technology import is suitable for use in the computer can be production models. Drive the PX 2 to automobile sensor to collect the data integration, with deep neural network of vehicle driving around complex objects from images.
Future China baidu own Drive system of the car car computer, will use NVIDIA this Drive PX 2 computers. As the world more and more car manufacturers are actively develop self-driving technology into their own, NVIDIA is trying to launch efficiency through more powerful computers or processor, to help reduce the driving development threshold, let more automakers can enter the market.
NVIDIA and baidu also announced in early September, will jointly develop a whole self-drive architecture, two will be in their respective professional on AI technology, from the cloud to together to build the whole self-drive car end architecture.
Current global AI chip development is still in the initial stage, but in the center of the NVIDIA latest quarterly data double business revenue of $151 million to the growth of the performance, most are from AI related business; Intel's latest quarterly data center business sales of $4 billion, but rarely is related business revenue generated from AI.
