Nvidia Jetson 发展历程

原创已于 2024-08-30 20:58:53 修改 · 1.4k 阅读

17 ·

CC 4.0 BY-SA版权

文章标签：

#Jetson

于 2024-08-29 00:11:01 首次发布

nvidia 专栏收录该内容

5 篇文章

订阅专栏

注：机翻，未校。

The Jetson Evolution

Created on July 26, 2024

2024 · technote · Algoblog

Nvidia Jetson single-board computers (SBCs) provide versatile platforms for developers to implement AI and machine learning applications. Over the years, Nvidia has introduced several key products in the Jetson lineup, each advancing the capabilities and performance of embedded systems.
Nvidia Jetson 单板计算机（SBC）为开发人员提供了实现 AI 和机器学习应用程序的多功能平台。多年来，Nvidia 在 Jetson 系列中推出了几款关键产品，每款产品都提高了嵌入式系统的功能和性能。

example image

Four steps to AI on Jetson, each on two legs.

Jetson TX2 (2017): Embedded AI Jetson TX2 （2017）：嵌入式 AI

The release of the Jetson TX2 in 2017 marked a milestone in the development of embedded AI systems. Building on the success of its predecessors (TK1 and TX1), the TX2 introduced a more powerful and energy-efficient platform, making it suitable for real-time computer vision applications in edge devices.
2017 年发布的 Jetson TX2 标志着嵌入式 AI 系统发展的一个里程碑。在其前身（TK1 和 TX1）的成功基础上，TX2 引入了一个更强大、更节能的平台，使其适用于边缘设备中的实时计算机视觉应用。

The Jetson TX2 featured a Pascal GPU with 256 CUDA cores and a dual-core Denver 2 ARM CPU coupled with a quad-core ARM Cortex-A57. This architecture provided twice the performance of the TX1 while maintaining a similar power envelope. The TX2 found use cases in autonomous drones, robotics, and industrial automation, where real-time image processing and AI inference were essential. Its support for deep learning frameworks allowed developers to deploy AI models directly on the edge, reducing latency and improving time-sensitive performance.
Jetson TX2 采用具有 256 个 CUDA 内核的 Pascal GPU 和一个双核 Denver 2 ARM CPU 以及一个四核 ARM Cortex-A57。这种架构提供了 TX1 的两倍性能，同时保持了相似的功率包络。TX2 在自主无人机、机器人和工业自动化中找到了使用案例，在这些应用中，实时图像处理和 AI 推理是必不可少的。它对深度学习框架的支持使开发人员能够直接在边缘部署 AI 模型，从而减少延迟并提高时间敏感性能。

Jetson Nano (2019): AI at the Edge Jetson Nano （2019）：边缘 AI

The Jetson Nano, introduced in 2019, offered a low-cost platform for embedded computer vision. Designed for hobbyists, researchers, and developers working on budget-constrained projects, the Jetson Nano provided easy access to AI and machine learning tools.
Jetson Nano 于 2019 年推出，为嵌入式计算机视觉提供了一个低成本平台。Jetson Nano 专为从事预算受限项目的业余爱好者、研究人员和开发人员而设计，可轻松访问 AI 和机器学习工具。

Despite its affordable price, the Jetson Nano featured a 128-core Maxwell GPU and a quad-core ARM Cortex-A57 CPU, making it powerful enough to handle real-time image classification, object detection, and segmentation tasks. It supported popular AI frameworks such as TensorFlow and PyTorch, enabling the deployment of deep learning models in various applications. The Jetson Nano became popular in robotics, smart cameras, and educational projects. Its affordability and ease of use made it a platform of choice for developers looking to experiment with AI and computer vision on a smaller scale, without sacrificing performance.
尽管价格实惠，但 Jetson Nano 配备 128 核 Maxwell GPU 和四核 ARM Cortex-A57 CPU，使其功能强大，足以处理实时图像分类、对象检测和分割任务。它支持 TensorFlow 和 PyTorch 等流行的 AI 框架，从而支持在各种应用程序中部署深度学习模型。Jetson Nano 在机器人技术、智能相机和教育项目中广受欢迎。它的经济性和易用性使其成为希望在不牺牲性能的情况下在较小规模上试验 AI 和计算机视觉的开发人员的首选平台。

Jetson Xavier NX (2020): High-Performance AI in a Compact Form Jetson Xavier NX （2020）：紧凑的高性能 AI

In 2020, Nvidia introduced the Jetson Xavier NX, a compact yet powerful SBC designed to bridge the gap between the affordable Jetson Nano and the more robust Jetson AGX Xavier. The Xavier NX brought AI capabilities to edge devices that required a small footprint and low power consumption.
2020 年，Nvidia 推出了 Jetson Xavier NX，这是一款紧凑而强大的 SBC，旨在弥合价格实惠的 Jetson Nano 和更强大的 Jetson AGX Xavier 之间的差距。Xavier NX 为需要小尺寸和低功耗的边缘设备带来了 AI 功能。

The Jetson Xavier NX featured a Volta GPU with 384 CUDA cores and 48 Tensor Cores, providing up to 21 TOPS (trillions of operations per second) of AI performance. This made it suitable for demanding computer vision tasks, such as multi-camera processing, video analytics, and autonomous robotics.
Jetson Xavier NX 采用具有 384 个 CUDA 核心和 48 个 Tensor 核心的 Volta GPU，可提供高达 21 TOPS（每秒数万亿次操作）的 AI 性能。这使其适用于要求苛刻的计算机视觉任务，例如多摄像头处理、视频分析和自主机器人。

The Xavier NX offered a scalable solution for developers, allowing them to build AI-powered devices that could handle complex tasks in real-time. Its ability to process multiple high-resolution video streams simultaneously made it a good choice for applications like smart surveillance and advanced robotics.
Xavier NX 为开发人员提供了可扩展的解决方案，使他们能够构建可以实时处理复杂任务的 AI 驱动设备。它能够同时处理多个高分辨率视频流，使其成为智能监控和高级机器人等应用的不错选择。

Jetson Orin Nano (2023): Supercharged Entry-Level Applications Jetson Orin Nano （2023）：超强入门级应用程序

The Jetson Orin Nano, launched in 2023, represents the next generation of AI computing for entry-level embedded applications. Building on the success of the original Nano, the Orin Nano offers significantly enhanced performance and AI capabilities while maintaining an accessible price point.
Jetson Orin Nano 于 2023 年推出，代表了入门级嵌入式应用的下一代 AI 计算。在原始 Nano 的成功基础上，Orin Nano 提供了显著增强的性能和 AI 功能，同时保持了可承受的价格点。

The Jetson Orin Nano features the Ampere GPU architecture with up to 1024 CUDA cores, delivering up to 40 TOPS of AI performance. This is a substantial upgrade from the original Nano, enabling more complex AI models and real-time computer vision tasks to be executed on edge devices.
Jetson Orin Nano 采用 Ampere GPU 架构，具有多达 1024 个 CUDA 内核，可提供高达 40 TOPS 的 AI 性能。这是对原始 Nano 的重大升级，可以在边缘设备上执行更复杂的 AI 模型和实时计算机视觉任务。

The Orin Nano is designed for applications where both cost and performance are necessary. It is well-suited for video analytics, and entry-level autonomous systems, helping developers to implement AI tasks without the need for higher-end hardware.
Orin Nano 专为需要成本和性能的应用而设计。它非常适合视频分析和入门级自主系统，帮助开发人员在不需要高端硬件的情况下实现 AI 任务。

Jetson AGX Orin (2023): AI Supercomputing at the Edge Jetson AGX Orin （2023）：边缘 AI 超级计算

The Jetson AGX Orin, also released in 2023, stands at the top of Nvidia’s Jetson lineup, offering high performance for the demanding real-time embedded applications. Featuring the Ampere GPU architecture with up to 2048 CUDA cores and 64 Tensor Cores, the Jetson AGX Orin delivers up to 275 TOPS of AI performance. This processing capability enables the execution of the many complex AI models, including those used in autonomous vehicles, robotics, and industrial automation. Support for advanced AI frameworks and libraries makes Jetson AGX Orin an attractive platform for next-generation AI systems.
Jetson AGX Orin 也将于 2023 年发布，是 Nvidia Jetson 系列的顶级产品，可为要求苛刻的实时嵌入式应用提供高性能。Jetson AGX Orin 采用 Ampere GPU 架构，具有高达 2048 个 CUDA 核心和 64 个 Tensor 核心，可提供高达 275 TOPS 的 AI 性能。这种处理能力支持执行许多复杂的 AI 模型，包括用于自动驾驶汽车、机器人和工业自动化的模型。对高级 AI 框架和库的支持使 Jetson AGX Orin 成为下一代 AI 系统的一个有吸引力的平台。

Nvidia Jetson platforms have enabled a wide range of applications, from hobbyist projects to industrial automation and autonomous systems.
Nvidia Jetson 平台支持广泛的应用，从业余项目到工业自动化和自主系统。

Nvidia Jetson single-board computers have shown how quickly real-time embedded computer vision and AI have advanced. Starting with the Jetson TX2, which offered a big jump in performance and efficiency, Nvidia made AI more accessible with the Jetson Nano. Now, with the powerful Jetson AGX Orin, they have brought supercomputing power to edge AI and embedded vision systems, constantly pushing the limits of what’s possible.
Nvidia Jetson 单板计算机展示了实时嵌入式计算机视觉和 AI 的发展速度。从性能和效率大幅提升的 Jetson TX2 开始，Nvidia 通过 Jetson Nano 使 AI 更易于访问。现在，借助强大的 Jetson AGX Orin，他们为边缘 AI 和嵌入式视觉系统带来了超级计算能力，不断突破可能的极限。

via：

The Jetson Evolution | Algotechniq

https://www.algotechniq.com/blog/2024/jetson-evolution/

Nvidia Jetson

Nvidia Jetson is a series of embedded computing boards from Nvidia. The Jetson TK1, TX1 and TX2 models all carry a Tegra processor (or SoC) from Nvidia that integrates an ARM architecture central processing unit (CPU). Jetson is a low-power system and is designed for accelerating machine learning applications.
Nvidia Jetson 是 Nvidia 的一系列嵌入式计算板。Jetson TK1、TX1 和 TX2 型号都配备了 Nvidia 的 Tegra 处理器（或 SoC），该处理器集成了 ARM 架构中央处理器（CPU）。Jetson 是一种低功耗系统，专为加速机器学习应用程序而设计。

Nvidia Jetson TK1

The Jetson family includes the following boards:
Jetson 系列包括以下板：

Nvidia Jetson TK1

In late April 2014, Nvidia shipped the Nvidia Jetson TK1 development board containing a Tegra K1 SoC in the T124 variant and running Ubuntu Linux.[1]
2014 年 4 月下旬，Nvidia 发布了 Nvidia Jetson TK1 开发板，其中包含 T124 变体的 Tegra K1 SoC 并运行 Ubuntu Linux。[1]

Nvidia Jetson TX1

The Nvidia Jetson TX1 development board bears a Tegra X1 of model T210.[2]
Nvidia Jetson TX1 开发板搭载了 T210 型号的 Tegra X1。[2]

Nvidia Jetson TX2

The Nvidia Jetson TX2 board bears a Tegra X2 of microarchitecture GP10B[3] (SoC type T186 or very similar). This board and the associated development platform was announced in March 2017 as a compact card design for low power scenarios, e.g. for the use in smaller camera drones. A matrix describing a set of performance modes was provided by the media along with that.[4] Further a TX2i variant, said to be rugged and suitable for industrial use cases, is mentioned.[5]
Nvidia Jetson TX2 主板搭载微架构 GP10B[3] 的 Tegra X2（SoC 类型 T186 或类似产品）。该板和相关开发平台于 2017 年 3 月发布，作为适用于低功耗场景的紧凑型卡设计，例如用于较小的相机无人机。媒体提供了一个描述一组性能模式的矩阵。[4]此外，还提到了 TX2i 变体，据说坚固耐用，适用于工业用例。[5]

Nvidia Jetson Xavier NX

Xavier was announced as a development kit in end of August 2018. Indications were given that a 20x acceleration for certain application cases compared to predecessor devices should be expected, and that the application power efficiency is 10x improved. Nvidia Jetson Xavier NX has a 6-core Nvidia Carmel ARMv8.2.
Xavier 于2018年8月底宣布作为一款开发套件。指出与前代设备相比，在某些应用案例中应期待20倍的加速，并且应用的功率效率提高了10倍。Nvidia Jetson Xavier NX 配备有6核心的Nvidia Carmel ARMv8.2。

Nvidia Jetson AGX Xavier

The Nvidia Jetson AGX Xavier is the 8-core version on the same core architecture (Carmel Armv8.2).[6]
Nvidia Jetson AGX Xavier 是同一核心架构（Carmel Armv8.2）上的8核心版本。[6]
An Nvidia Jetson Nano developer kit

Nvidia Jetson Nano

The Nvidia Jetson Nano was announced as a development system in mid-March 2019[7] .
Nvidia Jetson Nano 于2019年3月中旬宣布作为一款开发系统[7]。

The intended market is for hobbyist robotics due to the low price point. The final specs expose the board being sort of a power-optimized, stripped-down version of what a full Tegra X1 system would mean.
其目标市场是面向业余机器人爱好者，因其低价格而备受欢迎。最终规格表明，这块电路板实际上是一种经过功耗优化的、简化版的 Tegra X1 系统。

Only half of the CPU (only 4x A57 @ 1.43 GHz) and GPU (128 cores of Maxwell generation @ 921 MHz) cores are present and only half of the maximum possible RAM is attached (4 GB LPDDR4 @ 64 bit + 1.6 GHz = 25.6 GB/s) whilst the available or usable interfacing is determined by the baseboard design and is further subject of implementation decisions and specifics in an end user specific design for an application case.[10]
仅有一半的 CPU（4 个 A57 核心，主频为 1.43 GHz）和 GPU（128 个 Maxwell 架构核心，主频为 921 MHz）核心，同时只配备了最大可用 RAM 的一半（4 GB LPDDR4，64 位 + 1.6 GHz = 25.6 GB/s），而可用的接口则由基板设计决定，并且还受到特定应用案例中的实施决策和细节的进一步影响。[10]
The Nvidia Jetson Nano Developer Kit is an AI computer for makers, learners, and developers that brings the power of modern artificial intelligence to a low-power, easy-to-use platform, to start quickly with out-of-the-box support for many popular peripherals, add-ons, and ready-to-use projects.[11]
Nvidia Jetson Nano 开发者套件是一款面向创客、学习者和开发者的 AI 计算机，它将现代人工智能的强大功能引入一个低功耗、易于使用的平台，通过对许多流行的外围设备、附加组件和即用型项目的开箱即用支持快速入门。[11]

在这里插入图片描述
An Nvidia Jetson Orin developer kit

Nvidia Jetson Orin Nano

In September 2022 Nvidia announced the Jetson Orin Nano.13 The modules have the same 260-pin SO-DIMM connector and 69.6 mm x 45 mm dimensions, and come in two variants. The 4 GB variant provides 20 Sparse or 10 Dense TOPs, using a 512-core Ampere GPU with 16 Tensor cores, while the 8 GB variant doubles those numbers to 40/20 TOPs, a 1024-core GPU and 16 Tensor cores. Both have 6 Arm Cortex-A78AE cores. The 4 GB module starts at 199 and the 8 GB variant for 299, when purchasing 1000 units.
2022 年 9 月，Nvidia 宣布推出 Jetson Orin Nano。13这些模块具有相同的 260 针 SO-DIMM 连接器和 69.6 mm x 45 mm 尺寸，并有两种型号。4 GB 变体提供 20 个稀疏或 10 个密集 TOP，使用具有 16 个 Tensor 核心的 512 核 Ampere GPU，而 8 GB 变体将这些数字翻倍，达到 40/20 TOP、一个 1024 核 GPU 和 16 个 Tensor 核心。两者都有 6 个 Arm Cortex-A78AE 内核。购买 4 台时，8 GB 模块的起价为 199 美元，8 GB 模块的起价为 299 美元，购买 1000 台。

Performance

The published performance modes of the Nvidia Jetson TX2 are as follows.
Nvidia Jetson TX2 发布的性能模式如下。

Mode	Max Clocks (Denver 2 + A57)	Max-P (Denver 2 + A57)	Max-P (only Denver 2)	Max-P (only A57)	Max-Q (only A57)
GPU Clock (MHz)	1302	1122	854
Denver 2 Clock (MHz)	2000	1400	2000	stopped	stopped
Cortex-A57 (MHz)	2000+	1400	stopped	2000	1200
TDP / W	might vary	15	15	15	7.5

Jetson TX2 also has 5 power modes, numbered 0 through 4 as published by NVIDIA.[15] The default mode is mode 3 (MAX-P).
Jetson TX2 具有 5 种电源模式，由 NVIDIA 发布，编号为 0 到 4。[15]默认模式为模式 3 （MAX-P）。

Property	MAX-N (Mode 0) （模式 0）	MAX-Q (Mode 1) （模式 1）	MAX-P (Mode 2) （模式 2）	MAX-P* (Mode 3) （模式 3）	MAX-P (Mode 4) （模式 4）
Power budget	N/A	7.5W	15W	15W	15W
Online A57 CPU	4	4	4	4	1
Online D15 CPU	2	0	2	0	1
A57 CPU max freq (MHz)	2000	1200	1400	2000	345
D15 CPU max freq (MHz)	2000	N/A	1400	N/A	2000
GPU max freq (MHz)	1300	850	1122	1122	1122
Memory max freq (MHz)	1866	1331	1600	1600	1600

The published operation modes of the Nvidia Jetson Nano are:
Nvidia Jetson Nano 发布的操作模式为：

Mode	0	1
GPU Clock (MHz)	921	640
Cortex-A57 (MHz)	4x 1428	2x 918 2x stopped
TDP / W	10	5

Versions

There are various versions of the Jetson board available. Some of them are:

Year	Version	Performance	GPU	CPU	Memory	Power~~~
2017	Jetson TX2 [16]	1.33 TFLOPS	256-core Nvidia Pascal architecture GPU	Dual-core Nvidia Denver 2 64-bit CPU and quad-core ARM Cortex-A57 MPCore processor	8 GiB	7.5–15 W
2020	Jetson Xavier NX	21 TOPS	384-core Nvidia Volta architecture GPU with 48 Tensor cores	6-core Nvidia Carmel ARMv8.2 64-bit CPU 6MB L2 + 4MB L3	8 GiB	10–20W
2018	Jetson AGX Xavier [17]	32 TOPS	512-core Nvidia Volta architecture GPU with 64 Tensor cores	8-core NVIDIA Carmel ARMv8.2 64-bit CPU 8MB L2 + 4MB L3	32-64 GiB	10W - 30W
2019	Jetson Nano	472 GFLOPS	128-core Nvidia Maxwell architecture GPU	Quad-core ARM Cortex-A57 MPCore processor	4 GiB	5–10 W
2023	Jetson Orin Nano [18]	20–40 TOPS	from 512-core Nvidia Ampere architecture GPU with 16 Tensor cores	6-core ARM Cortex-A78AE v8.2 64-bit CPU 1.5MB L2 + 4MB L3	4-8 GiB	7–10 W
2023	Jetson Orin NX	70–100 TOPS	1024-core Nvidia Ampere architecture GPU with 32 Tensor cores	up to 8-core ARM Cortex-A78AE v8.2 64-bit CPU 2MB L2 + 4MB L3	8–16 GiB	10–25 W
2023	Jetson AGX Orin	200-275 TOPS	up to 2048-core Nvidia Ampere architecture GPU with 64 Tensor cores	up to 12-core ARM Cortex-A78AE v8.2 64-bit CPU 3MB L2 + 6MB L3	32–64 GiB	15–60 W

Software

Various operating systems and software might be able to run on the Jetson board series.
各种操作系统和软件可能能够在 Jetson 板系列上运行。

Linux

JetPack is a software development kit (SDK) from Nvidia for their Jetson board series. It includes the Linux for Tegra (L4T) operating system and other tools. The official Nvidia download page bears an entry for JetPack 3.2 (uploaded there on 2018-03-08) that states:
JetPack 是 Nvidia 为其 Jetson 板系列提供的软件开发套件（SDK）。它包括 Linux for Tegra （L4T）操作系统和其他工具。Nvidia 官方下载页面有一个 JetPack 3.2 的条目（于 2018 年 3 月 8 日上传），其中指出：

JetPack 3.2 adds support for the Linux for Tegra r28.2 image for the Jetson OS. It is packaged with newer versions of Tegra System Profiler, TensorRT, and cuDNN from the last release.[19]
JetPack 3.2 增加了对 Jetson OS 的 Linux for Tegra r28.2 映像的支持。它与上一版本中较新版本的 Tegra System Profiler、TensorRT 和 cuDNN 打包在一起。[19]

RedHawk Linux is a high-performance RTOS available for the Jetson platform, along with associated NightStar real-time development tools, CUDA/GPU enhancements, and a framework for hardware-in-the-loop and man-in-the-loop simulations.[20]
RedHawk Linux 是一款适用于 Jetson 平台的高性能 RTOS，以及相关的 NightStar 实时开发工具、CUDA/GPU 增强功能以及用于硬件在环和人在环仿真的框架。[20]

QNX

The QNX operating system also available for the Jetson platform, though it is not widely announced. There are success reports of installing and running specific QNX packages on certain Nvidia Jetson board variants. Namely the package qnx-V3Q-23.16.01 that is seemingly in parts based on Nvidia’s Vibrante Linux distribution is reported to run on the Jetson TK1 Pro board.[21]
QNX 操作系统也可用于 Jetson 平台，但尚未广泛发布。有关于在某些 Nvidia Jetson 板变体上安装和运行特定 QNX 软件包的成功报告。即据报道，似乎部分基于 Nvidia 的 Vibrante Linux 发行版的软件包 qnx-V3Q-23.16.01 在 Jetson TK1 Pro 板上运行。[21]