NVIDIA Jetson Orin NX 模块规格详解

PDF文件

667KB | 更新于2024-06-16 | 84 浏览量 | 举报 1 收藏

立即下载

"NVIDIA ORIN NX 是一款基于NVIDIA Ampere架构的嵌入式计算模块，结合了高性能的GPU和Arm Cortex-A78AE CPU，配备了LPDDR5内存。此规格书详细介绍了该模块的硬件特性、接口信息、推荐工作条件以及版本更新历史。" NVIDIA Jetson Orin NX系列模块是NVIDIA推出的一款高性能嵌入式计算平台，主要面向自动驾驶、机器人、边缘计算等应用领域。这款模块的核心是Ampere架构的GPU和Arm Cortex-A78AE CPU，提供了强大的处理能力和高效的能源效率。 1. **硬件配置** - **Ampere GPU**: NVIDIA Ampere架构的GPU提供了显著的图形和计算性能提升，支持最新的CUDA、TensorRT、DirectX和OpenGL等技术，为AI推理和计算密集型任务提供强大支持。 - **Arm Cortex-A78AE CPU**: 高能效的多核CPU，专为汽车电子控制单元（ECU）设计，确保在严苛的车载环境中稳定运行。 - **LPDDR5内存**: 提供高速数据传输，加速数据处理和存储，提高系统整体响应速度。 2. **接口与功能** - **PCIe Pin Descriptions**: 提供了PCI Express接口，支持高速数据传输，可用于连接高速存储设备或扩展其他高性能接口。 - **MODULE_ID和SHUTDOWN_REQ**: 这些控制引脚用于模块识别和关闭操作，确保系统管理的灵活性和安全性。 - **Jetson Orin NX Pin List**: 列出了所有外部连接器的引脚定义，方便开发者进行硬件设计和连接。 - **Display Controller**: 支持多种显示接口，如HDMI、LVDS等，可驱动高清显示器。 - **Sensor Processing Engine**: 专门处理传感器数据，优化传感器融合，适用于自动驾驶和机器视觉应用。 - **Security Engine**: 提供安全功能，保护数据和系统的完整性。 3. **工作条件与环境** - **Recommended Operating Conditions**: 规定了模块在正常运行时的温度、电压和电流等参数范围。 - **Temperature Range**: 明确了模块的工作温度范围，确保在各种环境下稳定运行。 4. **版本更新历史** - 自v0.1到v1.0，规格书经历了多次更新，包括对PCIe、CSI、ISP、Display Controller等部分的清晰化描述，增加了Sensor Processing Engine和Security Engine的章节，以及更新了模块尺寸图、绝对最大额定值表和可靠性报告。 5. **设计指南与热设计** - **Product Design Guide and Thermal Design Guide**: 提供了硬件设计和热管理方面的指导，帮助开发者构建高效、可靠的系统。 NVIDIA Jetson Orin NX通过其先进的硬件配置和丰富的功能，为开发者提供了强大的平台，以构建和部署复杂的嵌入式AI应用。无论是自动驾驶车辆的感知系统，还是工业自动化中的智能边缘计算节点，ORIN NX都具备满足这些需求的能力。

NVIDIA Jetson Orin NX Series Modules DS-10712-001_v1.1 | 4

Chapter 2. Functional Description

NVIDIA® Jetson Orin™ NX brings AI supercomputer performance to the edge in a compact

system-on-module (SOM) which is smaller than a credit card. Jetson Orin NX is built around a low-

power version of the NVIDIA Orin SoC, combining the NVIDIA Ampere™ GPU architecture with 64-

bit operating capability, integrated advanced multi-function video and image processing, and

NVIDIA Deep Learning Accelerators.

Compute performance up to 100 (Sparse) INT8 TOPs and 50 (Dense) INT8 TOPs enables the

Jetson Orin NX to run multiple neural networks in parallel and process data from multiple high-

resolution sensors simultaneously. It also offers a unique combination of performance and power

advantages with a rich set of I/Os, from high-speed CSI and PCIe to low-speed I2Cs and GPIOs,

allowing embedded and edge computing devices that demand increased performance but are

constrained by size, weight, and power budgets.

2.1 Ampere GPU

The NVIDIA Ampere GPU introduces a new design for the Streaming Multiprocessor (SM) that

dramatically improves performance per watt and performance per area, along with supporting

Tensor Cores and TensorRT Cores. Ampere GPUs improve on the previous NVIDIA Turing™

generation; and are software compatible so that the same APIs are used.

The NVIDIA Ampere Architecture GPU has several enhancements for compute and graphics

capability that include:

> Sparsity: fine grained structured sparsity doubles throughput and reduces memory usage.

> 2x CUDA floating-point performance: higher compute math speed.

> SM architecture improves bandwidth to the L1 cache and shared memory and reduces L1

miss latency.

> Improved async compute and post-L2 cache compression compared to NVIDIA Turing.

The GPC is a dedicated hardware block for rasterization, shading, texturing, and compute. The

GPU’s core graphics functions are performed inside the GPC where the SM CUDA cores perform

pixel/vertex/geometry shading and physics/compute calculations. Texture units perform texture

filtering and load/store units fetch and save data to memory. Special Function Units (SFUs)

handle transcendental and graphics interpolation instructions. Tensor cores perform matrix

multiplies to greatly accelerate DL inferencing. The RTcore unit assists Ray Tracing by

accelerating Bounding Volume Hierarchy (BVH) traversal and intersection of scene geometry

during Ray Tracing.

Functional Description

NVIDIA Jetson Orin NX Series Modules DS-10712-001_v1.1 | 5

There are multiple texture processing clusters (TPC) units within a graphics processing cluster

(GPC). Each TPC includes two SMs, a Polymorph Engine, two Texture Units, and a Ray Tracing core

(RTcore). Each GPC includes a Raster Engine (ROP), which can access all of memory. Each SM is

partitioned into four separate processing blocks, each with its own instruction buffer, scheduler,

and 128 CUDA cores.

Finally, the PolyMorph engine handles vertex fetch, tessellation, viewport transform, attribute

setup, and stream output. The SM geometry and pixel processing performance make it highly

suitable for rendering advanced user interfaces and complex gaming applications. The power

efficiency of the Ampere GPU enables this performance on devices with power-limited

environments.

2.1.1 Compute Features

Ampere introduces third-generation NVIDIA Tensor Cores which offer a wider range of precisions

including TensorFloat-32 (TF32), bfloat16, FP16, and INT8, all of which provide unmatched

versatility and performance.

TensorFloat-32 (TF32) is a new format that uses the same 10-bit Mantissa as half-precision

(FP16) math and is shown to have more than sufficient margin for the precision requirements of

AI workloads. In addition, since the TF32 adopts the same 8-bit exponent as FP32 it can support

the same numeric range.

Ampere adds support for structured sparsity. Not all the parameters of modern AI networks are

needed for accurate predictions and inference, some can be converted to zeros to make the

models “sparse” without compromising accuracy. The Tensor Cores in Ampere can provide up to

2x higher performance for inference of sparse models.

Ampere supports Compute Data Compression which can accelerate unstructured sparsity and

other compressible data patterns. Compression in L2 provides up to a 4x improvement in DRAM

read/write bandwidth, up to 4x improvement in L2 read bandwidth, and up to a 2x improvement in

L2 capacity.

Ampere also supports many other enhancements for higher compute throughput.

Table 2-1: GPU Operation

Module CUDA Cores Tensor Cores Operating Frequency per Core

(up to)

Jetson Orin NX 16GB 1024 32 918 MHz

Jetson Orin NX 8GB 1024 32 765 MHz

2.1.2 Graphic Features

Ampere graphics capabilities include:

> End-to-end lossless compression, including Post-L2 compression, enabling compression of M

stores.

> Tiled Caching

> OpenGL 4.6+, Vulkan 1.2+, CUDA 11.4+

剩余53页未读，继续阅读

wms588

粉丝: 20

NVIDIA Jetson Orin NX 模块规格详解

英伟达orin手册和参考

Nvidia_Xavier_SoC_datasheet.pdf

NVIDIA ORIN X/ORIN NX ORIN NANO pin脚配置说明

NVIDIA ORIN NX开发指南：嵌入式开发者的终极路线图

【AI应用性能飞跃】：NVIDIA ORIN NX实战加速技术

NVIDIA ORIN NX功耗优化技巧：系统效率提升的终极指南

NVIDIA ORIN NX的多核性能：多任务处理的极致优化

【GPU内核秘密】：NVIDIA ORIN NX架构优势全面解读

【AI的未来趋势】：NVIDIA ORIN NX如何塑造行业变革

NVIDIA ORIN NX系统集成要点：软硬件协同优化的黄金法则

最新资源