Monday, November 29

The new Nvidia AGX Orin for robots and autonomous cars matches the power of a supercomputer but fits in the palm of your hand

Nvidia is premiering and today it has taken the opportunity to show its new SoC for autonomous cars and robotics. It’s about the new Nvidia Jetson AGX Orin, the new generation regarding Xavier that promises to be up to six times more powerful, with an ultra-small footprint that fits in the palm of your hand.

The company has been trying for years to lead in the field of autonomous cars and from 2022 it will offer an SoC capable of performing up to 200 billion operations per second. That is, the equivalent of a “server with GPU”, as described by Nvidia itself. A small-sized supercomputer that will seek to increase the power that the cars and robots of the future will be able to process.

The brain of future autonomous cars


The new supercomputer comes with an ARM Cortex-A78 “Hercules” CPU and a GPU with 2048 Ampere CUDA core architecture. In total we have up to 17,000 million transistors.

We have 32 GB of LPDDR5 RAM and a 2 GHz clock frequency, slightly lower than what we had in Xavier and predictably focused on achieving better energy efficiency.

The TDP is between 15W and 50W, while Xavier was located in the 30W. But to get an idea, while Orin promises a performance of 200 TOPS (‘trillions operations per second’), the previous generation stayed at 32. A huge jump in Nvidia in a few years.

If you look at the size, Nvidia has kept Xavier’s dimensions and they still fit in the palm of your hand, at 10 centimeters tall. Nevertheless, a leap in power has been made that equates it with servers that have a dedicated GPU and whose size is more appropriate for data centers than for autonomous cars or robots, where the available size for the CPU is much more limited.

More information | Nvidia

Leave a Reply

Your email address will not be published. Required fields are marked *