Tegra Parker was used in the Drive PX 2 on-board computer announced earlier this year
At the just concluded Hot Chips conference, NVIDIA released its new generation Tegra processor, Parker, in a high profile. In fact, as early as the beginning of this year, Tegra Parker is used in the Drive PX 2 on-board computer, and is known as the world's first intelligent super computer in autonomous vehicles. When he announced Drive PX 2, Huang said that it used the latest generation of Tegra processor, but he did not release more details about it at that time.
At the Hot Chips conference, NVIDIA officially said that as a new generation of Tegra processor, Tegra Parker is a processor for intelligent driving, which is very suitable for the on-board computer system on smart cars.
In terms of parameter configuration, Parker's CPU part is 2 * Denver 2.0+4 * Cortex-A57. The official description of the CPU part is that Parker's Denver core is version 2.0, 64 bit ARMv8 architecture, 7-Way superscalar architecture, while optimizing the dynamic decoding algorithm, adding additional low power consumption and reserving state for energy conservation. Denver 2.0 and Cortex-A57 cores are interconnected through the same L2 cache.
In terms of GPU, Tegra Parker adopts the Pascal architecture currently used by NVIDIA on the GTX 10 series, with 256 CUDA cores. Although the number of CUDA cores has not changed, the efficiency of CUDA is higher in Pascal architecture, and the frequency of GPU can also be higher. We can understand that the performance of GPU is still greatly improved. NVIDIA said that Parker's GPU can provide strong performance support for complex computing of unmanned driving and deep learning. It can simultaneously drive multiple high-resolution display devices, which can be used for instrument panel, map navigation, on-board central control panel, etc. in smart cars.
Tegra Parker has a strong expansion capability. It supports virtualization hardware, and can support 8 virtual machines. It is used for multiple on-board systems in smart cars, such as on-board entertainment systems, instrument panel systems, automatic driving systems, and so on. Only one Tegra Parker can complete it. In addition, NVIDIA said that if you feel that the performance of a Tegra Parker cannot meet your needs, you can integrate it into more complex designs. For example, NVIDIA Drive Px 2 uses two Tegra Parker and multiple Pascal GPUs.
In other aspects, Tegra Parker supports 12 real-time HDR cameras to work at the same time (which is extremely important for unmanned driving), uses 128bit LPDDR4 memory, and the memory bandwidth is up to 50GB/s (the bandwidth of Snapdragon 820 is 29GB/s). Parker supports 4K 60Hz video decoding/encoding. Compared with the previous generation X1, it has some new features related to intelligent vehicles (driverless, safety protection, etc.).
Although Tegra processors are more and more far away from ordinary consumers, they are still compared with the CPU performance of the latest Samsung Exynos 8890, Qualcomm Snapdragon 820, Apple A9x and other powerful SoCs in PPT. Why not compare GPU? Maybe NVIDIA leads too much, so there is no need to compare~~
Of course, we know that today's Tegra is not a SoC for the mobile market. It is targeted at enterprise level users and companies. NVIDIA hopes that it can become a leader in the market of smart cars and in vehicle intelligent systems. Because of this, the power consumption of Tegra series is certainly not comparable to that of mobile SoC.
It is known that Volvo will be the first automobile manufacturer to deploy NVIDIA Drive PX 2 on-board artificial intelligence supercomputer to support its automatic driving project Drive Me next year. At that time, 100 XC90 SUVs equipped with Drive PX 2 will conduct public automatic driving test on the roads in Gothenburg, Sweden.
