Why Intel $400 million acquisition of the company's 2 years of machine learning companies?
zhidongxi· 2016-08-15 17:21:54
Zhi Bian Yi Yuan, "
8 10 day, Intel spent $400 million to buy a set up only 2 years Engineering Intelligence depth study of the entrepreneurial company Systems Nervana, for its chip to join the depth of learning, this is probably a position in the case of shrinking PC market, while seeking new opportunities to move.
1.6; "Intel as the world's largest chip maker has been the business focus in the field of PC, in the mobile terminal chip war has already lost to ARM. In recent years, the development of PC market has brought great obstacles to the development of Intel, looking for a way to become an inevitable choice. A
at the beginning of August Intel released on the future development strategy of the file can be seen in the future, the focus is not on the PC computer chip business, but on the data center and network equipment development chip. And the acquisition of Systems Nervana may indicate that Intel has found a way to enhance the PC chip.
Nirvana CEO Naveen Rao said in a blog, the company will provide the deep learning framework, as well as the hardware platform for Intel, its 48 staff all joined Intel data center work.
Nervana Systems was founded in 2014, headquartered in California, San Diego. It has attracted including DFJ, Data Collective, Fuel Capital, Lux Capital and Allen & Co. investment company, $25 million, in the hardware solutions and training of neural networks has made significant achievements. />
Arjun, the company's three founders Bansal Naveen, Rao Amir and Khosrowshahi
Arjun in an interview, Bansal
is what Intel needs another architecture?
GPU is very suitable for deep learning, because it has tens of thousands of floating point units can be used in parallel matrix operation, which constitutes a large number of the depth of the neural network. But most of the GPU also has many other features, such as clip output images. In addition, GPU also provides a higher degree of accuracy computing capabilities, such as financial aspects of computing, simulation and modeling, etc., these do not require deep learning algorithms. All of these features are occupied by the GPU chip valuable space and operational capabilities. In theory, the way Nervana can have a higher performance and lower prices, but also to reduce the burden on the computer. Nirvana can provide deep learning capabilities for Xeon and Phi Xeon and other processors to reduce development costs.
Nervana not too much disclosure on their progress, is still mainly concentrated in the NEON accelerated GPU to ensure the completion of the new products in 2017. Prior to the news that the Nervana system includes a steel mold structure, can be connected to these devices in the form of 3D topology. This feature allows the system to measure a large number of cooperative accelerator, can better train the depth of the neural network. To develop this feature requires additional vendor or Intel. Maybe next year we will know how AI works. />
Intel needs a bigger GPU? />
Intel next plan is what? />
apple, Google have AI in the entire layout, the future will be based artificial intelligence chip calculation way deep into the vast field.
creator and a BAT here. Make friends！
intellectual things (zhidxcom)
Car owners please pay attention! BAT has begun to invade your car
50 days of bankruptcy, 3! Shared bicycle horror collective killed
Naen Bo President Wang Ye: buy the balance of the car originator after 2...
Shen Hui: an average of 21 days there is a custom National Smart car production
In the era of intelligent smart 2016 + future summit will be opened in...
Dialogue millet Wang Chuan: only do have the ability to do the product
"59 seconds" Shenzhen watch copycat tide pushed the four-wheel drive car...
Since the end of Lenovo AR glasses low-key two years of development of AR...