Nvidia has released a new mobile supercomputer: 1 teraflops on a small board

Nvidia launches the new version of the Jetson TX1 mobile supercomputer. The board replaced the TK1 that appeared last year. The TX1 is designed for use in embedded systems and is engaged in high-speed computing for machine learning. You can use it in areas such as pattern recognition, the creation of autonomous unmanned aerial vehicles, computer vision or analysis of the behavior of the crowd.

A Linux-based computer achieves a 1 teraflop speed and is equipped with Ethernet and Wi-Fi network interfaces. It goes on sale on November 16 at a price of $ 599 ($ ​​299 for educational institutions and students). The power consumption of the device is less than 10 watts.

Onboard the Jetson TX1 is 1 TFLOP / s 256-core Maxwell GPU, 64-bit ARM A57 CPU, 4 GB of DDR4 RAM memory and 16 GB of eMMC flash memory. All this is located on the board the size of a credit card.
')

TX1 complete set with installed board

Speaking at the presentation of the computer, the director of Nvidia, Jen-San Huang [Jen-Hsun Huang], shared his thoughts on how the architecture of computers is changing before our eyes. According to him, produced by the company GPU provide the necessary computer capacity for the implementation of machine learning projects.

According to Huang, so far only Facebook and Google are using these technologies, but after the mass distribution of “mobile supercomputers,” even small companies will be able to create their own devices at no extra cost. Thus, Nvidia is trying to "democratize" the field of machine learning.

The previous version of the computer was released in the spring of 2014 , and could provide 326 gigaflops at a price of $ 192.