NVIDIA has finally unveiled Atlan, the fifth-generation SoC built on the next-generation ARM Grace CPU and Ampere Next GPU cores. The Atlan SOC is geared for autonomous vehicles as well as a number of other items that will be revealed in the coming months.
The NVIDIA Atlan SOC has over 20 billion transistors which have Grace ARM Cores and Ampere Next GPU Cores from NVIDIA.
NVIDIA's Atlan SOC, the successor to the Orin SOC, has more advanced and next-gen IPs than ever before. Atlan uses NVIDIA's custom-designed Grace ARM cores and Ampere Next GPU cores, while Orin used the ARM Hercules CPU and Ampere GPU cores. The SOC, like its Orin counterpart, which featured Ampere GPU cores months before the architecture made its formal appearance in the HPC market, is the first official product to use a next-generation graphics architecture.
ARM Cores in the NVIDIA Grace CPU
The Grace CPU is the first, and it is a custom design by NVIDIA built on the ARM architecture. NVIDIA's Grace CPU is based on the next-generation Grace CPU cores, which will outperform the ones announced today. When compared to LPDDR4x memory, the chip would run alongside a fast LPDDR5x memory subsystem that provides more than 2x the bandwidth uplift and 10x the energy consumption. More information on the Grace CPU can be found in our full article here.
Ampere Next GPU Cores from NVIDIA
The Ada Lovelace GPU, which is expected to be the successor to NVIDIA's Ampere GPU, will control the GPU side. In its most recent roadmap, NVIDIA dubbed the next-generation GPU Ampere Next and confirmed that it will be available in 2022, so we can expect Atlan SOCs to follow suit.
The Atlan SOC will have up to 1000 TOPs on a single chip and will also provide 300 Gbps networking through a stable gateway. ASIL-D Safety Island features will be added to the chip on NVIDIA's Drive Atlan boards. By exploiting NVIDIA's current GPU architecture, updated Arm CPU cores, and deep learning and machine vision accelerators, that's a 4x increase over the previous generation.
The platform architecture provides enough computing power for the redundant and complex deep neural networks that will power future AI vehicles, while also allowing developers to incorporate new functionality and enhancements. This stable architecture will ensure the NVIDIA DRIVE platform's protection and stability for future vehicle generations. NVIDIA DRIVE Orin vehicle manufacturing will begin in 2022, and Atlan will follow in 2023 with sampling and production vehicles scheduled for 2025.
| SOC Name | Tegra X1 | Parker | Xavier | Orin | Atlan |
|---|---|---|---|---|---|
| Process Technology | 20nm TSMC | 16nm TSMC | 12nm TSMC | 8nm Samsung | 5nm TSMC? |
| SOC Transistors | 2 Billion (Tegra X1) | N/A | 7 Billion (Xavier) | 17 Billion (Orin) | TBA |
| GPU Architecture | Maxwell (256 Core) | Pascal (256 Core) | Volta (512 Core) | Ampere (2048) | Ada Lovelace? |
| CPU | 16 Core ARM CPU | 12 Core ARM CPU | 8 Core ARM CPU | 12 Core ARM CPU | Grace Next CPU Cores |
| CPU Architecture | 4x Cortex A57 4x Cortex A53 | 4x Denver A53 8x Cortex A57 | Carmel ARM64 8 Core CPU (8 MB L2 + 4 MB L3) | ARM Herclues Cores (A72AE) | Next-Gen Neoverse |
| System Memory | LPDDR4 | 8 GB LPDDR4 (50+ GB/s) | 16 GB 256-bit LPDDR4 | LPDDR4x | LPDDR5x |
Post a Comment