NVIDIA DRIVE DEVELOPER FAQ

NVIDIA DRIVE Hyperion? is an AV development platform and reference architecture for developing Level 2+ and Level 3 highway autonomous solutions. It consists of a complete sensor suite that’s tuned, optimized, and safety-certified, as well as a high-performance AI computing platform, NVIDIA DRIVE AGX?.

The NVIDIA DRIVE AGX? Developer Kit provides the hardware, software, and sample applications needed for the development of production-level AVs. The DRIVE AGX system is built on production, auto-grade silicon and engineered with security in mind, featuring an open software framework.

The NVIDIA DRIVE Software Development Kit (SDK) is a collection of software packages used for developing autonomous vehicles. It comprises the foundational NVIDIA DRIVE? OS and DriveWorks SDK as well as advanced applications such as highly automated supervised driving (DRIVE AV) and AI cockpit (DRIVE IX).

NVIDIA DRIVE OS is a foundational software stack consisting of an embedded Real Time OS (RTOS), hypervisor, NVIDIA DriveWorks, NVIDIA? CUDA? libraries, NVIDIA TensorRT? , and other modules that give you access to the hardware engines. DRIVE OS offers a safe and secure execution environment for applications such as secure boot, security services, firewall, and over-the-air updates. Plus, it offers a real-time environment with an RTOS and a hypervisor for Quality of Service (QoS). The RTOS, AUTOSAR, and hypervisor are ASIL-D components.

NVIDIA DriveWorks is a middleware framework that consists of a library of software modules, sample applications, and tools to enable software development for autonomous vehicles. It leverages the computing power of the DRIVE AGX platforms and is designed to be open, modular, and compliant with automotive industry software standards such as ISO26262 and MISRA-C. DriveWorks is available with DRIVE OS releases.

NVIDIA DRIVE IX is an SDK used for developing in-vehicle applications for safety, comfort, and enhancing the user experience. For more information on how to get started, refer to our DRIVE IX developer page.

The DRIVE AGX Xavier and Pegasus? XT Developer Kits are available directly from NVIDIA and through authorized sublease partners. Software releases and materials are published in the NVIDIA DRIVE AGX SDK Developer Program. Your company or university must have current legal agreements on file. Please contact your Account Representative (or contact us) to ensure necessary agreements have been signed before requesting to join the NVIDIA DRIVE AGX SDK Developer Program. Users may only join with an approved corporate or university email address.

The NVIDIA DRIVE AGX Orin? Developer Kit will be available directly from NVIDIA or through authorized distributors. Users must register their DRIVE AGX Orin Developer Kit to get access to the NVIDIA DRIVE AGX SDK Developer Program where software releases and materials will be published. Users may only join with an approved corporate or university email address.

If you’ve received your NVIDIA DRIVE AGX DevKit, congratulations! The Onboarding Guide provides an overview of all resources relevant to help you start your development.

The DRIVE AGX platforms are compatible with a number of sensors and peripherals from different vendors. A full list is available on the DRIVE Ecosystem Hardware and Software Components.

While both the Jetson AGX and DRIVE AGX kits use the same Orin SoC, DRIVE is built for automotive applications. The following table highlights the differences between the two platforms.

DRIVE AGX
Jetson AGX
Hardware
Segment
Automotive
Embedded & Robotics?
Power
100W (Module power)?
60W (MAXN mode, module power)**?
Input Voltage Range
9V to 16V (Static), 7V to 32V (Transient)?
7V to 20V (High), 5V (Mid)?
DL Inference TOPS (INT8)
254 TOPS
275 TOPS***?
Operating Temperature?
0 to 45C?
0 to 35C?
Safety MCU
Infineon Aurix TC397
No?
Networking
1GbE RJ45?
10GbE RJ45?
100 T1, 1000 T1, 10 GbE T1
? CAN, LIN, Flexray, USS
1GbE RJ45?
10GbE RJ45?
M.2 Key E (WLAN / BT, PCIe, USB 2.0, UART, I2S & I2C)
Storage (default stuffing)
1x UFS (256 GB)?
64GB eMMC 5.1?
MicroSD?
M.2 Key M (NVMe)?
Video In?
16x GMSL-In ?
MIPI CSI-2?
Operating System
Operating System
DRIVE OS Linux with Safety Extensions (Ubuntu) | QNX for Safety?

?Comprehensive safety & security services: Functional safety island integration, streamlined safety framework, platform security controller integration, secure boot & PKCS #11 support for crypto acceleration features?
Linux4Tegra (Ubuntu)?

Secureboot, trusted execution environment, disk and dram encryption, crypto acceleration, hardware-based security fuses?
Bootloader
Quickboot (optimized for real-time applications)?
UEFI
Computer Vision
NvMedia (highly optimized APIs to access compute engines and sensors)?
VPI (image-processing algorithms supported on multiple compute engines like CPU, GPU, PVA, VIC)
Multimedia
NvMedia ? (up to 16 cameras for simultaneous processing with hardware acceleration)?
MM API / V4L2 API / Gstreamer?
Imaging
NvMedia SIPL (provides proprietary ISP algorithms with optimization)?
Librargus (Multimedia, ISP access)
SDK
DriveWorks: Sensor abstraction, image/point cloud processing, vehicle I/O, compute graph framework, DNN framework, recorder, calibration, egomotion?

? NvMedia, NvStreams, CUDA, cuDNN, TensorRT?, Triton inference server, NVIDIA container runtime

? Tools: Nsight Systems (performance analysis), Nsight Graphics (debug, profiler and export with Direct3D), Nsight Eclipse Edition (IDE), Nsight Compute (interactive kernel profiler for CUDA applications)?
Compute/Inferencing: CUDA, cuDNN, TensorRT? Inference serving: Triton inference server? Cloud native: NVIDIA container runtime?

? Tools: Nsight Systems (performance analysis), Nsight Graphics (debug, profiler, and export with Direct3D), Nsight Eclipse Edition (IDE), Nsight Compute (interactive kernel profiler for CUDA applications)?
Production
Automotive Grade
Yes
No
ASIL Target*
ASIL-D (QNX)
-
AEC-Q100 qual
Yes
No
NVIDIA Design Support
Available
-

* Intended for production platform only

** Need to choose required compute engines to run application and stay within power budget

*** Bumped performance up on Orin SoC (not auto grade) GPU & DLA

DRIVE AGX Orin
DRIVE AGX Xavier
SoC?
Orin (x1)
Xavier (x2)?
DL Inference TOPS (INT8)?
254 TOPS
60 TOPS
Memory Bandwidth?
Per Orin – 205GB/
Per Xavier – 136GB/s?
System RAM?
32GB LPDDR5 (3200MHz)?
32GB LPDDR4 (2133MHz)?
Networking
100MbE (x6)?
1GbE (x10)?
10GbE (x2)?
100MbE (x5)?
1GbE (x7)?
10GbE (x2)
Automotive Audio Bus (A2B)?
1
No
Wireless Connection?
Wi-Fi/Bluetooth?
No
Operating System?
DRIVE OS 6 Linux | QNX
DRIVE OS 5 Linux | QNX?
Middleware Framework
DriveWorks 5 (compute graph framework)?
DriveWorks 3 & 4?
Developer Tools?
CUDA 11.4?
TensorRT 8.x?
cuDNN 8.x?
CUDA 10.2?
TensorRT 6.4?
cuDNN 7.6?
Support
Yes
EOL

Feature
NVIDIA DRIVE AGX Orin
NVIDIA DRIVE Orin-X SoC
1
NVIDIA dGPU SXM2
N/A
System RAM
32GB LPDDR5
QSPI NOR Flash
64MB
eMMC
64GB
Security Key Storage
8MB
UFS
256GB
Camera
16x GMSL2
MCU
Aurix Safety MCU
Ethernet
10GbE
2
1GbE
10
100MbE
6
CAN
6
FlexRay to MCU
1
USB 2.0
USB3.2
2 Type-A
2 Type-C
Display Output
1 DisplayPort
GPIO
6
PCIe
PCIe x8 Gen4
eSATA
N/A
Automotive Audio Bus (A2B)
1
Debug and Development
JTAG, UART, USB

If you can’t find what you're looking for in the downloads, public documentation, or forum, please contact your NVIDIA representative or reach out directly.

Resources

Peek under the hood to experience NVIDIA’s latest autonomous driving innovations through DRIVE Labs and DRIVE Dispatch.

View DRIVE videos