In the early years, NVIDIA used ARM to launch the Tegra chip. It wanted to have a lot of experience in the smart phone and tablet industry. However, under the premise that there are no major advantages in key indicators such as base band, GPU compatibility, and power consumption, Tegra began to transform its Automobile autopilot and other industrial fields.
This is not the case. Two consecutive generations of Tegra (Parker and Xavier) are positioned like this. The advantage is that NV only needs to do graphics and floating-point computing performance, without considering the baseband, GPU compatibility, and the power cap is no longer harsh. .
According to Anandtech, on the GTC 2018, Huang Renxun quickly previewed NV's new ARM product line, the new codename Orin (DC comic hero Orion).
However, with regard to Orin's specific parameters, this time it is really tight-lipped. Currently, it is guessed that it is based on the 7nm process (small chip area in the figure), integrates the new architecture GPU core (Ampere?), and deeply learns floating-point performance. Continue to erupt.
The current Tegra Xavier is based on NVIDIA's self-developed ARM core 'Carmel (U.S. City Name)'. The GPU is based on the Volta architecture and its single-precision floating-point performance is 1.3T.