NVIDIA released the AI big move, they introduced the smallest and most powerful edge AI supercomputer Jetson Xavier NX, a hardware with only one bank card size, can provide up to 21 TOPS calculations at 15W power consumption. And for just $399, you can take it home.
NVIDIA released the AI big move, they introduced the smallest and most powerful edge AI supercomputer Jetson Xavier NX, a hardware with only one bank card size, can provide up to 21 TOPS calculations at 15W power consumption. And for just $399, you can take it home.
This is also a new breakthrough made by NVIDIA in the calculation of AI reasoning. Jetson Xavier NX has expanded their Jetson family product line, and NVIDIA has become more and more versatile in making AI money with GPU.
The Sparrow May be Small but It Has All the Vital Organs.
Judging from the officially released parameters, this small-sized supercomputer can be said to be "small but complete":
GPU: 384-core NVIDIA Volta GPU and 48 Tensor cores;
CPU: 6-core Carmel ARMv8.2 64-bit CPU (6MB L2 + 4MB L3);
Accelerator: Dual NVIDIA Deep Learning Accelerator (NVDLA) engine;
RAM: 8GB 128-bit LPDDR4x; 51.2GB/sec;
Computing power: 14 TOPS (when power consumption is 10W) or 21 TOPS (when power consumption is 15W);
Module size: 70x45mm
More parameters are as follows:
It is understood that the Jetson Xavier NX module will be available in March 2020 at a price of $399. As can be seen from the size of this supercomputer, it is very suitable for embedding into small hardware products, dealing with data reasoning at the edge. The official also said that Jetson Xavier NX is aimed at embedded edge computing devices that have high performance requirements but are limited in size, weight, power consumption and budget. For example, the small commercial robots (14.030, -0.12, -0.85%), drones, intelligent high-resolution sensors (for factory logistics and production lines), optical inspection, network video recorders, portable medical devices and other industrial Internet of Things (IoT) system are all included.
Currently, NVIDIA's Jetson platform has released the small AI computer Jetson Nano, the Jetson AGX Xavier series for fully autonomous machines and the Jetson TX2 series for edge artificial intelligence. Together with the just-released Jetson Xavier NX, the four products are suitable for different performance and budget needs.
The specific comparison parameters are as follows:
In addition, NVIDIA also benchmarked the hardware performance evaluation for the Jetson series, based on the MLPerf Inference 0.5 benchmark, comparing the real-time performance of several products in image classification, object detection, pose estimation, segmentation, etc.
Throughout the entire Jetson series, Jetson Xavier NX plays the role of “linking the preceding and the following”. It is smaller and cheaper than the “big brother” Jetson AGX Xavier, and it is much faster than the other two “little brothers”. Jetson Xavier NX can run on the same CUDA-X AI software architecture as all Jetson products, ensuring that products can be quickly brought to market at lower development costs.
Rob Csongor, vice president and general manager of NVIDIA, said that the Jetson series has 400,000 developers and more than 3,000 customers worldwide.
Work Harder to Return to the Glory with Richer AI Products
Looking back at NVIDIA's AI layout, the first in 2007, they introduced the parallel computing platform and programming model CUDA, through the GPU to obtain the first profit of AI rising. At that time, many research institutions or developers chose NVIDIA's GPU solutions for various artificial intelligence training and application scenarios.
In 2015, they released DRIVE PX products for autonomous driving. In the following years, NVIDIA released a new graphics card architecture, new accelerators, new supercomputers, and announced that they are an AI company.
In the two years of its prosperity, NVIDIA swept the artificial intelligence and encrypted digital currency industries. The peak annual revenue growth rate remained above 50% for a long time, and the stock price has increased sevenfold in the past two years. However, as the price of encrypted digital currencies fell, and more and more technology giants and AI startups developed dedicated AI chips, NVIDIA ’s share price plummeted and its performance declined.
At the same time, high-power, high-priced GPU product solutions have discouraged many companies and turned to more cost-effective solutions. On the other hand, architectural innovation has become the decisive factor to shark the chip market. The fierce FPGA architecture and the open source of RISC-V all indicate that the choice of technology has begun to increase. In order to reduce costs, Tesla chose to part with NVIDIA, while Google, Huawei, Amazon, etc. have already researched server chips and got rid of their dependence on NVIDIA.
Under extremely difficult circumstances, NVIDIA, which has the advantage of AI first move, is surrounded by more and more competitors.
In the second quarter of the 2020 fiscal year released by Nvidia in August, the game business is still a large revenue, accounting for more than 50%. The data center business achieved revenue of 655 million US dollars this quarter, down 14% year-on-year.
Therefore, Nvidia, which once dominated the cloud AI chip market, gradually began to look at the edge and terminal after facing the breakout of competitors. Leverage cloud- and edge-integrated solutions use to attract more customers, and the Jetson line of products for the terminal and the edge is the new weapon of NVIDIA. In terms of size, computing power, power consumption, price and other different dimensions, NVIDIA has launched matching products to meet the needs of different users.
With the gradual improvement of the AI product line, coupled with their strong hardware and software comprehensive strength and early industrial accumulation, Nvidia wants to use AI to return to the glorious moment again, there still have chance.
At last
This time NVIDIA launched a cheap and beautiful edge AI supercomputer, it will hit some domestic AI chip companies. Because most of the current AI chip manufacturers will cut into the market from edge-end chips, and NVIDIA has mature software and hardware platforms, its compatibility and synergy are higher. The Jetson series products may steal some customers of AI chip manufacturers. The market for AI chips that are surging, with the influx of more and more giants, the opportunities for new entrants to survive are getting lower and lower.
