0% found this document useful (0 votes)

143 views40 pages

Jetson Xavier NX Data Sheet v1.3

Jetson Xavier

Uploaded by

Joko

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

143 views40 pages

Jetson Xavier NX Data Sheet v1.3

Jetson Xavier

Uploaded by

Joko

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 40

DATA SHEET [PRELIMINARY]

NVIDIA Jetson Xavier NX System-on-Module

Volta GPU + Carmel ARM + 8 GB LPDDR4x + 16 GB eMMC 5.1

AI Performance Display Controller

Up to 21 TOPS (INT8) Two multi-mode (eDP/DP/HDMI) Serial Output Resources (SOR)
eDP 1.4a | DP 1.4 | HDMI 2.0a/b
Volta GPU
Maximum Resolution (eDP/DP/HDMI): (up to) 3840x2160 at 60 Hz
384 NVIDIA® CUDA® cores | 48 Tensor cores | End-to-end lossless
(up to 36 bpp)
compression | Tile Caching | OpenGL® 4.6 | OpenGL ES 3.2 |
Vulkan™ 1.1◊ | CUDA 10 | Maximum Operating Frequency: 1100 Multi-Stream HD Video and JPEG
MHz
Video Decode:
Carmel CPU • Standards supported: H.265 (HEVC), H.264, VP9, VP8,
ARMv8.2 (64-bit) heterogeneous multi-processing (HMP) CPU MPEG-4, MPEG-2, VC-1
architecture | 3x dual-core CPU clusters (six NVIDIA Carmel o 2x690 MP/sec (HEVC)
processor cores) connected by a high-performance system o 2x 4K @ 60 (HEVC)
coherency interconnect fabric | L3 Cache: 4 MB (shared across all o 4x 4K @ 30 (HEVC)
clusters) o 12x 1080p @ 60 (HEVC)
NVIDIA Carmel (Dual-Core) Processor: L1 Cache: 128 KB L1 o 32x 1080p @ 30 (HEVC)
instruction cache (I-cache) per core; 64 KB L1 data cache (D- o 16x 1080p @ 30 (H.264)
cache) per core | L2 Unified Cache: 2 MB per cluster | Maximum
Operating Frequency: 1900 MHz Video Encode:
• Standards supported: H.265 (HEVC), H.264, VP9
Audio o 2x464 MP/sec (HEVC)
Dedicated programmable audio processor | ARM Cortex A9 with o 2x 4K @ 30 (HEVC)
NEON | PDM in/out | Industry-standard High Definition Audio (HDA) o 6x 1080p @ 60 (HEVC)
controller provides a multi-channel audio path to the HDMI® o 14x 1080p @ 30 (HEVC)
interface
Peripheral Interfaces
Memory
8 GB 128-bit LPDDR4x DRAM | Secure External Memory Access xHCI host controller with integrated PHY (up to) 1x USB 3.1, 3x
Using TrustZone® Technology | System MMU | Maximum Operating USB 2.0 | PCIe 1x1 (GEN3) + 1x4 (GEN4) | SD/MMC controller
Frequency: 1600 MHz (supporting eMMC 5.1, SD 4.0, SDHOST 4.0 and SDIO 3.0) | 3x
UART | 2x SPI | 4x I2C | 1x CAN | 2x I2S | GPIOs
Storage
Mechanical
16 GB eMMC 5.1 Flash Storage | Bus Width: 8-bit | Maximum Bus
Frequency: 200 MHz (HS400) Module Size: 69.6 mm x 45 mm | 260 pin SO-DIMM Connector

Networking Operating Requirements

10/100/1000 Gigabit Ethernet | Media Access Controller (MAC) Temperature Range (TJ)*: -25°C – 90°C | Supported Power
Modes: 10W – 15W | Power Input: 5V
Imaging
14 lanes (3 x4 or 6 x2) MIPI CSI-2 | D-PHY 1.2 (2.5 Gb/s per pair,
total up to 30 Gbps)

Note: Refer to the Software Features section of the latest L4T Development Guide for a list of supported features; all features may not be
available.
◊
Product is based on a published Khronos Specification and is expected to pass the Khronos Conformance Process. Current conformance status
can be found at www.khronos.org/conformance.

* See the Jetson Xavier NX Thermal Design Guide for details

Revision History
Version Date Description
V1.0 November 6, 2019 Initial release.
V1.1 February 1, 2020 Updated:
• PCIE0_XXX pins from Ctrl #0 to Ctrl #5 under PCIe Pin description in Table
9: PCIe Pin Descriptions
• PCIE1_XXX pins from Ctrl #1 to Ctrl #4 under PCIe Pin description in Table
9: PCIe Pin Descriptions
• Pulse Width Modulator (PWM) to reflect four outputs instead of eight outputs
• SHUTDOWN_REQ* and SYS_RESET* pull up information in Table 21:
Power and System Control Pins
• Table 29: Absolute Maximum Ratings to include the Mounting Force
parameter.
• Mechanical Drawing
V1.2 February 24, 2020 Added:
• Tolerance information for Mechanical Drawing
V1.3 April 21, 2020 Added:
• PMIC_BBAT to reflect RTC accuracy
• SoC height for the Mechanical Drawing
• Table 30: Jetson NX Reliability Report table
• Gen4 information to PCI Express (PCIe) section on page 18.

Table of Contents
1.0 Functional Overview 4
1.1 Volta GPU .................................................................................................................................................................... 4
1.2 Carmel CPU Complex .................................................................................................................................................. 5
1.3 Memory Subsystem ...................................................................................................................................................... 5
1.4 Memory ........................................................................................................................................................................ 6
1.5 Video Input Interfaces .................................................................................................................................................. 6
1.5.1 MIPI Camera Serial Interface (CSI) ............................................................................................................... 6
1.5.2 Video Input (VI) .............................................................................................................................................. 9
1.5.3 Image Signal Processor (ISP) ........................................................................................................................ 9
1.6 Display Controller ......................................................................................................................................................... 9
1.6.1 HDMI and DisplayPort Interfaces ................................................................................................................. 11
1.6.2 Embedded DisplayPort (eDP) ...................................................................................................................... 12
1.7 High-Definition Audio-Video Subsystem ..................................................................................................................... 13
1.7.1 Multi-Standard Video Decoder ..................................................................................................................... 13
1.7.2 Multi-Standard Video Encoder ..................................................................................................................... 14
1.7.3 JPEG Processing Block ............................................................................................................................... 14
1.7.4 Video Image Compositor (VIC) .................................................................................................................... 15
1.7.5 Audio Processing Engine (APE) .................................................................................................................. 15
1.7.6 High Definition Audio (HDA) ........................................................................................................................ 16
1.8 Interface Descriptions ................................................................................................................................................. 16
1.8.1 SD/eMMC .................................................................................................................................................... 16
1.8.2 Universal Serial Bus (USB) .......................................................................................................................... 17
1.8.3 PCI Express (PCIe)...................................................................................................................................... 18
1.8.4 Serial Peripheral Interface (SPI) .................................................................................................................. 20
1.8.5 Universal Asynchronous Receiver/Transmitter (UART) ............................................................................... 21
1.8.6 Controller Area Network (CAN) .................................................................................................................... 22
1.8.7 Inter-Chip Communication (I2C) ................................................................................................................... 23
1.8.8 Inter-IC Sound (I2S) ..................................................................................................................................... 24
1.8.9 Gigabit Ethernet ........................................................................................................................................... 25
1.8.10 Fan............................................................................................................................................................. 26
1.8.11 Pulse Width Modulator (PWM) ................................................................................................................... 26
1.9 Deep Learning Accelerator (DLA) .............................................................................................................................. 26
2.0 Power and System Management 28
2.1 Power Rails ................................................................................................................................................................ 28
2.2 Power Domains/Islands .............................................................................................................................................. 29
2.3 Power Management Controller (PMC) ....................................................................................................................... 29
2.4 Resets ........................................................................................................................................................................ 29
2.5 PMIC_BBATT ............................................................................................................................................................. 29
2.6 Power Sequencing ..................................................................................................................................................... 29
2.6.1 Power Up ..................................................................................................................................................... 30
2.6.2 Power Down ................................................................................................................................................ 30
2.7 Power States .............................................................................................................................................................. 30
2.7.1 ON State ...................................................................................................................................................... 31
2.7.2 OFF State .................................................................................................................................................... 31
2.7.3 SLEEP State ................................................................................................................................................ 31
2.8 Thermal and Power Monitoring .................................................................................................................................. 32
3.0 Pin Definitions 33
3.1 Power-on Reset Behavior........................................................................................................................................... 33
3.2 Sleep Behavior ........................................................................................................................................................... 33
3.3 GPIO .......................................................................................................................................................................... 34
3.4 Pin List........................................................................................................................................................................ 35
4.0 DC Characteristics 36
4.1 Operating and Absolute Maximum Ratings ................................................................................................................ 36
4.2 Digital Logic ................................................................................................................................................................ 37
5.0 Package Drawing and Dimensions 39

1.0 Functional Overview

NVIDIA® Jetson Xavier™ NX brings AI supercomputer performance to the edge in a compact system-on-module (SOM) that’s
smaller than a credit card. Jetson Xavier NX is built around a low-power version of the NVIDIA Xavier SoC, combining the
NVIDIA Volta™ GPU architecture with 64-bit operating capability, integrated advanced multi-function video and image
processing, and NVIDIA Deep Learning Accelerators.
Compute performance up to 14 TOPS (at 10W) or 21 TOPS (at 15W) enables the Jetson Xavier 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.

1.1 Volta GPU

The Graphics Processing Cluster (GPC) is a dedicated hardware block for computing, rasterization, shading, and texturing of
most of the GPU’s core graphics functions. The GPC is comprised of Texture Processing Clusters (TPC), with each TPC
containing two Streaming Multiprocessor (SM) units, and a Raster Engine. The SM unit creates, manages, schedules, and
executes instructions from many threads in parallel. Raster operators (ROPs) continue to be aligned with L2 cache slices and
memory controllers. The SM geometry and pixel processing performance make it highly suitable for rendering advanced user
interfaces, while the efficiency of the Volta GPU enables this performance on devices with power-limited environments.

Each SM is partitioned into four separate processing blocks (referred to as SMPs), each SMP contains its own instruction
buffer, scheduler, CUDA cores, and Tensor cores. Inside each SMP, CUDA cores perform pixel/vertex/geometry shading and
physics/compute calculations, and each Tensor core provides a 4x4x4 matrix processing array to perform mixed-precision
fused multiply-add (FMA) mathematical operations. 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. Finally, the
PolyMorph Engine handles vertex fetch, tessellation, viewport transform, attribute setup, and stream output.

Features:
• End-to-end lossless compression
• Tile Caching
• Support for OpenGL 4.6, OpenGL ES 3.2, Vulkan 1.1
• Adaptive Scalable Texture Compression (ASTC) LDR profile supported
• CUDA support
• Iterated blend, ROP OpenGL-ES blend modes
• 2D BLIT from 3D class avoids channel switch
• 2D color compression
• Constant color render SM bypass
• 2x, 4x, 8x MSAA with color and Z compression
• Non-power of 2D and 3D textures, FP16 texture filtering
• FP16 shader support
• Geometry and Vertex attribute instancing
• Parallel pixel processing
• Early-z reject: Fast rejection of occluded pixels acts as multiplier on pixel shader and texture performance while
saving power and bandwidth
• Video protection region
• Power saving: Multiple levels of clock gating for linear scaling of power

Table 1: GPU Operation

Module CUDA Cores Tensor Cores Power Mode Operating Frequency per Core (up to)
Jetson Xavier NX 384 48 10W 800 MHz
15W 1100 MHz

1.2 Carmel CPU Complex

The CPU complex (CCPLEX) is comprised of three Carmel dual-core CPU clusters in a coherent multi-processor
configuration. A high-performance System Coherency Fabric (SCF) connects all CPU clusters enabling simultaneous
operation of all CPU cores (as needed) for a true heterogeneous multi-processing (HMP) environment.
Features include:
• NVIDIA Dynamic Code Optimization
• 10-wide Superscalar architecture
• Dynamic branch prediction with a Branch Target Buffer and Global History Buffer RAMs, a return stack buffer, and an
indirect predictor
• Full implementation of ARMv8.2 ISA compliant architecture including:
o ARMv8 TrustZone
o ARMv8.0 Crypto ISA
o Trusted Memory
o ARMv8.2-FP16 support
• 128 KB 4-way-associative parity protected L1 instruction cache per core
• 64 KB 4-way-associative parity protected L1 data cache per core
• 2 MB 16-way-associative ECC protected L2 cache per CPU cluster
• 4 MB 16-way-associative ECC protected L3 cache (shared across all clusters)
• Performance Monitoring
• Interface to an external Generic Interrupt Controller (vGIC-400)
• Support for power management with multiple power domains

Table 2: CPU Operation

Module Power Mode CPU Cores CPU Maximum Frequency

Jetson Xavier NX 10W 2-core 1.5 GHz
4-core 1.2 GHz
15W 2-core 1.9 GHz
4/6-core 1.4 GHz

1.3 Memory Subsystem

The Memory Subsystem (MSS) provides access to local DRAM, SysRAM, and provides a SyncPoint Interface for inter-
processor signaling. The MSS supports full-speed I/O coherence by routing requests through a scalable coherence fabric. It
also supports a comprehensive set of safety and security mechanisms.

Structurally, the MSS consists of:

1 MSS Data Backbone - routes requests from clients to the MSS Hub and responses from MSS Hub to the clients
2 MSS Hub - receives and arbitrates among client requests, performs SMMU translation, and sends requests to the
MCF
3 Memory Controller Fabric (MCF) - performs security checks, feeds I/O coherent requests to the Scalable Coherence
Fabric (SCF), and directs requests to the multiple memory channels
4 Memory Controller (MC) Channels - row sorter/arbiter and DRAM controllers
5 DRAM I/O - channel-to-pad fabric, DRAM I/O pads, and PLLs

Jetson Xavier NX integrates a 128-bit wide LPDDR4x memory interface implemented as four 32-bit channels with x16 sub-
partitions. The memory controller provides a single read or write command, plus a row address to both sub-partitions in the
channel to transfer 64 bytes. It also provides three independent column address bits to each sub-partition, allowing it access
different 32-byte sectors of a GOB between the sub-partitions. It provides connections between a wide variety of clients,

supporting their bandwidth, latency, quality-of-service needs, and any special ordering requirements that are needed. The
MSS supports a variety of security and safety features and address translation for clients that use virtual addresses.

Features:
• LPDDR4x: x32 DRAM chips
• 128-bit wide data bus
• Low latency path and fast read/response path support for the CPU complex cluster
• Support for low-power modes:
o Software controllable entry/exit from self-refresh, power down, and deep power down
o Hardware dynamic entry/exit from power down, self-refresh
o Pads use DPD mode during idle periods
• High-bandwidth interface to the integrated Volta GPU
• Full-speed I/O coherence with bypass for Isochronous (ISO) traffic
• System Memory-Management Unit (SMMU) for address translation based on the ARM SMMU-500
• High-bandwidth PCIe ordered writes
• AES-XTS encryption with 128-bit key

1.4 Memory
The Jetson Xavier NX integrates 8 GB 128-bit LPDDR4x DRAM. Maximum frequency of 1600 MHz has a theoretical peak
memory bandwidth of 51.2 GB/s.

The Memory Controller (MC) maximizes memory utilization while providing minimum latency access for critical CPU requests.
An arbiter is used to prioritize requests, optimizing memory access efficiency and utilization and minimizing system power
consumption. The MC provides access to main memory for all internal devices. It provides an abstract view of memory to its
clients via standardized interfaces, allowing the clients to ignore details of the memory hierarchy. It optimizes access to shared
memory resources, balancing latency and efficiency to provide best system performance, based on programmable
parameters.

Features:
▪ TrustZone (TZ) Secure and OS-protection regions
▪ System Memory Management Unit
▪ Dual CKE signals for dynamic power down per device
▪ Dynamic Entry/Exit from Self-Refresh and Power Down states

1.5 Video Input Interfaces

1.5.1 MIPI Camera Serial Interface (CSI)
Standard
MIPI CSI 2.0 Receiver specification
MIPI D-PHY® v1.2 Physical Layer specification

The NVIDIA Camera Serial Interface (NVCSI) works with the Video Input (VI) unit to capture an image from a sensor, where
NVCSI is a source of pixel data to VI. NVCSI works in streaming mode while VI captures the required frames using a single-
shot mode of operation. All sync point generation for software is handled at VI; the delay between NVCSI and VI is negligible
in software terms. NVCSI does not have a direct memory port, instead it sends the pixel data to memory through the VI.

Fifth-generation NVIDIA camera solution (NVCSI 2.0, VI 5.0, and ISP 5.0) provides a combination host that supports
enhanced MIPI D-PHY (with lane deskew support) physical layer options in three 4-lane or six 2-lane configurations; or
combinations of these. Each lane can support up to 16 virtual channels (VC) and supports data type interleaving.

• Virtual Channel Interleaving: VCs are defined in the CSI-2 specification and are useful when supporting multiple
camera sensors. With the VC capability, a one-pixel parser (PP) can de-interleave up to 16 image streams.
• Data Type Interleaving: In HDR line-by-line mode, the sensor can output long/short exposure lines using the same
VC and a different programmable data type (DT).
• Frequency Target: The parallel pixel processing rate, measured in pixels-per-clock (PPC), is increased to allow
higher throughput and lower clock speeds. To support higher bandwidth without increasing the operating frequency,
the host processes multiple pixels in one clock. NVCSI is capable of processing four PPCs when bits-per-pixel (BPP)
is greater than 16, and eight PPC when BPP is less than or equal to 16.
• With the new streaming mode in NVCSI, one PP can handle all traffic (embedded data and image data) from one
camera device, including 16 VCs.
Features:
• Supports the MIPI D-PHY v1.2 physical layer option:
o MIPI D-PHY supports up to 2.5 Gbits/sec per pair, for an aggregate bandwidth of 30 Gbps from 12 pairs
• Based on MIPI CSI-2 v2.0 protocol stack
• Includes six-pixel parsers (PP)
• Supports up to 16 virtual channels per active PP
• Supported input data formats:
o RGB: RGB888, RGB666, RGB565, RGB555, RGB444
o YUV: YUV422-8b, YUV420-8b (legacy), YUV420-8b, YUV444-8b
o RAW: RAW6, RAW7, RAW8, RAW10, RAW12, RAW14, RAW16, RAW20
o DPCM (predictor 1): 14-10-14, 14-8-14, 12-8-12, 12-7-12, 12-6-12, 12-10-12, 10-8-10, 10-7-10, 10-6-10
(Predictor 2 not supported)
• Data type interleave support

Table 3: CSI Pin Descriptions

Pin # Signal Name Description Direction Pin Type

10 CSI0_CLK_N Camera, CSI 0 Clock– Input MIPI D-PHY
12 CSI0_CLK_P Camera, CSI 0 Clock+ Input MIPI D-PHY
4 CSI0_D0_N Camera, CSI 0 Data 0– Input MIPI D-PHY
6 CSI0_D0_P Camera, CSI 0 Data 0+ Input MIPI D-PHY
16 CSI0_D1_N Camera, CSI 0 Data 1– Input MIPI D-PHY
18 CSI0_D1_P Camera, CSI 0 Data 1+ Input MIPI D-PHY
9 CSI1_CLK_N Camera, CSI 1 Clock– Input MIPI D-PHY
11 CSI1_CLK_P Camera, CSI 1 Clock+ Input MIPI D-PHY
3 CSI1_D0_N Camera, CSI 1 Data 0– Input MIPI D-PHY
5 CSI1_D0_P Camera, CSI 1 Data 0+ Input MIPI D-PHY
15 CSI1_D1_N Camera, CSI 1 Data 1– Input MIPI D-PHY
17 CSI1_D1_P Camera, CSI 1 Data 1+ Input MIPI D-PHY
28 CSI2_CLK_N Camera, CSI 2 Clock– Input MIPI D-PHY
30 CSI2_CLK_P Camera, CSI 2 Clock+ Input MIPI D-PHY
22 CSI2_D0_N Camera, CSI 2 Data 0– Input MIPI D-PHY
24 CSI2_D0_P Camera, CSI 2 Data 0+ Input MIPI D-PHY
34 CSI2_D1_N Camera, CSI 2 Data 1– Input MIPI D-PHY
36 CSI2_D1_P Camera, CSI 2 Data 1+ Input MIPI D-PHY

Pin # Signal Name Description Direction Pin Type

27 CSI3_CLK_N Camera, CSI 3 Clock– Input MIPI D-PHY
29 CSI3_CLK_P Camera, CSI 3 Clock+ Input MIPI D-PHY
21 CSI3_D0_N Camera, CSI 3 Data 0– Input MIPI D-PHY
23 CSI3_D0_P Camera, CSI 3 Data 0+ Input MIPI D-PHY
33 CSI3_D1_N Camera, CSI 3 Data 1– Input MIPI D-PHY
35 CSI3_D1_P Camera, CSI 3 Data 1+ Input MIPI D-PHY
52 CSI4_CLK_N Camera, CSI 4 Clock– Input MIPI D-PHY
54 CSI4_CLK_P Camera, CSI 4 Clock+ Input MIPI D-PHY
46 CSI4_D0_N Camera, CSI 4 Data 0– Input MIPI D-PHY
48 CSI4_D0_P Camera, CSI 4 Data 0+ Input MIPI D-PHY
58 CSI4_D1_N Camera, CSI 4 Data 1– Input MIPI D-PHY
60 CSI4_D1_P Camera, CSI 4 Data 1+ Input MIPI D-PHY
40 CSI4_D2_N Camera, CSI 4 Data 2– Input MIPI D-PHY
42 CSI4_D2_P Camera, CSI 4 Data 2+ Input MIPI D-PHY
64 CSI4_D3_N Camera, CSI 4 Data 3– Input MIPI D-PHY
66 CSI4_D3_P Camera, CSI 4 Data 3+ Input MIPI D-PHY
76 DSI_CLK_N Camera, CSI 5 Clock– Input MIPI D-PHY
78 DSI_CLK_P Camera, CSI 5 Clock+ Input MIPI D-PHY
70 DSI_D0_N Camera, CSI 5 Data 0– Input MIPI D-PHY
72 DSI_D0_P Camera, CSI 5 Data 0+ Input MIPI D-PHY
82 DSI_D1_N Camera, CSI 5 Data 1– Input MIPI D-PHY
84 DSI_D1_P Camera, CSI 5 Data 1+ Input MIPI D-PHY

Table 4: Camera Pin Descriptions

Pin # Signal Name Description Direction Pin Type

213 CAM_I2C_SCL Camera I2C Clock. 2.2kΩ pull-up to 3.3V Bidir Open Drain – 3.3V
on the module.
215 CAM_I2C_SDA Camera I2C Data. 2.2kΩ pull-up to 3.3V Bidir Open Drain – 3.3V
on the module.
116 CAM0_MCLK Camera 0 Reference Clock Output CMOS – 1.8V
114 CAM0_PWDN Camera 0 Powerdown or GPIO Output CMOS – 1.8V
122 CAM1_MCLK Camera 1 Reference Clock Output CMOS – 1.8V
120 CAM1_PWDN Camera 1 Powerdown or GPIO Output CMOS – 1.8V

1.5.2 Video Input (VI)

The VI block receives data from the CSI receiver and prepares it for presentation to system memory or the dedicated image
signal processor execution resources. The VI block provides formatting for RGB, YCbCr, and raw Bayer data in support of
several camera user models. These models include single and multi-camera systems, which may have up to six active
streams. The input streams are obtained from MIPI compliant CMOS sensor camera modules.

1.5.3 Image Signal Processor (ISP)

The ISP module takes data from the VI/CSI module or memory in raw Bayer format and processes it to YUV output. The
imaging subsystem supports raw (Bayer) image sensors up to 24 million pixels. Advanced image processing is used to convert
input to YUV data and remove artifacts introduced by high-megapixel CMOS sensors and optics with up to 30-degree CRA.
Features:
• Flexible post-processing architecture for supporting custom computer vision and computational imaging operations
• Bayer domain hardware noise reduction
• Per-channel black-level compensation
• High-order lens-shading compensation
• 3x3 color transform
• Bad pixel correction
• Programmable coefficients for de-mosaic with color artifact reduction
Color artifact reduction: a two-level (horizontal and vertical) low-pass filtering scheme that is used to reduce/remove
any color artifacts that may result from Bayer signal processing and the effects of sampling an image.
• Enhanced down scaling quality
• Edge enhancement
• Color and gamma correction
• Programmable transfer function curve
• Color-space conversion (RGB to YUV)
• Image statistics gathering (per-channel)
o Two 256-bin image histograms
o Up to 4,096 local region averages
o AC flicker detection (50 Hz and 60 Hz)
o Focus metric block

1.6 Display Controller

The Jetson Xavier NX integrates a Unified Display Controller (based on the NVIDIA NVDisplay architecture) and two
independent display outputs. The Display Controller includes a Pixel Processing Engine that fetches pixel data to be
processed from DRAM and generates up to six windows of rasterized display-ready pixel data. The instructions for processing
the pixel data are captured by the display controller's Front End (FE) logic, which then generates the individual controls for the
various stages of pixel processing. The pixel data to be processed are fetched in the Isochronous Memory Hub (IsoHub) then
go through the specified pixel processing, including merging the cursor, in four pipe stages: Pre-Composition (Pre-comp);
Composition (Comp); Post-Composition (Post-comp); and Raster Generation (RG). The rasterized display-ready pixel data are
available for the separate panels/devices (referred to as display heads) and are fed through a multi-channel crossbar structure
to the Serial Output Resources (SOR) in the Display Interface for the standard display output format, i.e., DP (Display Port)
and High-Definition Multimedia Interface (HDMI).

Each of the display heads can be run at an independent clock rate and each can drive a different display resolution. Each of
the six display windows (A, B, C, D, E, F) can be arbitrarily assigned to any of the display Heads as required, then connected
to any one of the display heads for the desired output format.

Features:
• Integrated HDCP key storage, no external SecureROM required
• Six windows that can be assigned to any Head

• One special-purpose TrustZone protected window on Head0

• Maximum raster size: 32768 x 32768
• Maximum active region: 8192 x 8192
• Maximum input surface size: 32768 x 32768
• Maximum fetched size: 8192 x 8192
• Input surface color formats:
o 16-bit RGB: R4G4B4A4, R5G6B5, A1R5G5B5, and R5G5B5A1
o 24-bit RGB: A8R8G8B8, X8R8G8B8, A8B8G8R8, and X8B8G8R8
o 32-bit RGB: A2R10G10B10, A2B10G10R10, X2BL10GL10RL10_XRBIAS, and X2BL10GL10RL10_XVYCC
o 64-bit RGB: R16_G16_B16_A16_NVBIAS, and R16_G16_B16_A16
o Packed YUV 422: Y8_U8_Y8_V8_N422, and U8_Y8_V8_Y8_N422
o Semi Planar YUV 422 (8, 10, 12 bpc):
- Y8_V8U8_N422, Y8_V8U8_N422R
- Y10_V10U10_N422, Y10_V10U10_N422R
- Y12_V12U12_N422, Y12_V12U12_N422R
o Semi-planar YUV 420 (8, 10, 12 bpc):
- Y8_V8U8_N420*
- Y10_V10U10_N420*
- Y12_V12U12_N420*
o Semi-planar YUV 444 (8 ,10, 12 bpc):
- Y8_V8U8_N444
- Y10_V10U10_N444
- Y12_V12U12_N444
o Planar YUV 420 (8, 10, 12 bpc):
- Y8_U8_V8_N420
- Y10_U10_V10_N420
- Y12_U12_V12_N420
o Planar YUV 444 (8, 10, 12 bpc):
- Y8_U8_V8_N444
- Y10_U10_V10_N444
- Y12_U12_V12_N444
• Pipeline depth
o 16-bpc, [-1.5, 2.5] range (two range extension bits): De-gamma will clip to 0,1 immediately on the input
• Vsync (VCOUNTER) and immediate (HCOUNTER) flip modes
o Immediate flip supported for RGB only
o Immediate flips occur at the second 8-line boundary after the current line
o
1. Cursor cannot be enabled on a Head unless the Head has at least one window group attached. The window
Notes: group does not need to be enabled.
2. TrustZone cannot be enabled unless Head0 has at least one window group attached. This does not need to
be enabled.
3. Color formats marked with an asterisk (*) are programmed as Y_UV in the display manuals, and then byte-
swapped later to be Y_VU.
4. 10-bpc and 12-bpc YUV color formats are packed into 16-bpc containers. This effectively limits immediate
flips to no faster than one every 16 lines.

1.6.1 HDMI and DisplayPort Interfaces

Standard Notes
High-Definition Multimedia Interface (HDMI) Specification, Scrambling support
version 2.0a/b Clock/4 support (1/40 bit-rate clock)
HDMI 1.4 (up to 340 MHz pixel clock rate)
HDMI 2.0 (up to 594 MHz pixel clock rate)
VESA DisplayPort Standard Version 1.4

A standard DP 1.4 or High-Definition Multimedia Interface (HDMI) 2.0a/b interface is supported. These share the same set of
interface pins, so either DisplayPort (DP) or HDMI can be supported natively. Dual-Mode DisplayPort (DP++) can be
supported, in which the DisplayPort connector logically outputs TMDS signaling to a DP-to-HDMI dongle. Each output collects
the output of a display pipeline from the display controller, formats/encodes that output (to a desired format), and then streams
it to an output device. Each output can provide an interface to an external device; each output can drive only a single output
device at any given time. HDMI support provides a method of transferring both audio and video data; the SOR receives video
from the display controller and audio from a separate high-definition audio (HDA) controller, it combines and transmits them as
appropriate.
Note: A single CEC controller is shared between the two HDMI/DP interfaces. Both DP0 and DP1 support either DP or HDMI.

Features:
• DisplayPort
o Multichannel audio from HDA controller, up to eight channels, 96 kHz, 24-bit
o DP1.4 supports HBR3 at 8.1 Gbps
o (up to) 540 MHz pixel clock rate (i.e., 1.62 GHz for RBR, 2.7 GHz for HBR, 5.4 GHz for HBR2, and 8.1 Gbps for
HBR3.
o 8b/10b encoding support
o External dual-mode standard support
o Audio streaming support
• HDMI
o (up to) 594 MHz pixel clock
- 8/12 bpc RGB and YUV444
- 8/10/12 bpc YUV422
- 8 bpc YUV420 (10/12 bpc YUV frame buffers should be output as YUV422)
o HDMI Vendor-Specific Info frame (VSI) packet transmission
o On HDMI, multichannel audio from HDA controller, up to eight channels, 192 kHz, 24-bit.
o Fuse calibration information for HDMI analog parameter(s)
o 1080i output on HDMI
• DP or HDMI connectors via appropriate external level shifting
• HDCP 2.2 and 1.4 over either DP or HDMI
Note: refer to NVIDIA software release notes for detailed specifications.
• External Dual Mode standard (DP2HDMI passive or active adapters and adapter discovery)
• Generic info frame transmission
• Frame-packed 3D stereo mode

Table 5: HDMI/DisplayPort/eDP Pin Descriptions

Pin # Signal Name Description Direction Pin Type

39 DP0_TXD0_N DisplayPort 0 Lane 0 or HDMI Lane 2– Output AC-Coupled on carrier board
41 DP0_TXD0_P DisplayPort 0 Lane 0 or HDMI Lane 2+ Output AC-Coupled on carrier board
45 DP0_TXD1_N DisplayPort 0 or HDMI Lane 1– Output AC-Coupled on carrier board

Pin # Signal Name Description Direction Pin Type

47 DP0_TXD1_P DisplayPort 0 or HDMI Lane 1+ Output AC-Coupled on carrier board
51 DP0_TXD2_N DisplayPort 0 Lane 2– or HDMI Lane 0– Output AC-Coupled on carrier board
53 DP0_TXD2_P DisplayPort 0 Lane 2+ or HDMI Lane 0+ Output AC-Coupled on carrier board
57 DP0_TXD3_N DisplayPort 0 Lane 3– or HDMI Clk Lane– Output AC-Coupled on carrier board
59 DP0_TXD3_P DisplayPort 0 Lane 3+ or HDMI Clk Lane+ Output AC-Coupled on carrier board
90 DP0_AUX_N Display Port 0 Aux– or HDMI DDC SDA Bidir AC-Coupled on Carrier Board
(eDP/DP)
92 DP0_AUX_P Display Port 0 Aux+ or HDMI DDC SCL Bidir AC-Coupled on Carrier Board
(eDP/DP)
88 DP0_HPD Display Port 0 or HDMI Hot Plug Detect Input CMOS – 1.8V
63 DP1_TXD0_N DisplayPort 1 Lane 0 or HDMI Lane 2– Output AC-Coupled on carrier board
65 DP1_TXD0_P DisplayPort 1 Lane 0 or HDMI Lane 2+ Output AC-Coupled on carrier board
69 DP1_TXD1_N DisplayPort 1 or HDMI Lane 1– Output AC-Coupled on carrier board
71 DP1_TXD1_P DisplayPort 1 or HDMI Lane 1+ Output AC-Coupled on carrier board
75 DP1_TXD2_N DisplayPort 1 Lane 2– or HDMI Lane 0– Output AC-Coupled on carrier board
77 DP1_TXD2_P DisplayPort 1 Lane 2+ or HDMI Lane 0+ Output AC-Coupled on carrier board
81 DP1_TXD3_N DisplayPort 1 Lane 3– or HDMI Clk Lane– Output AC-Coupled on carrier board
83 DP1_TXD3_P DisplayPort 1 Lane 3+ or HDMI Clk Lane+ Output AC-Coupled on carrier board
98 DP1_AUX_N Display Port 1 Aux– or HDMI DDC SDA Bidir AC-Coupled on Carrier Board
(eDP/DP) or Open-Drain, 1.8V
(3.3V tolerant - DDC)
100 DP1_AUX_P Display Port 1 Aux+ or HDMI DDC SCL Bidir AC-Coupled on Carrier Board
(eDP/DP) or Open-Drain, 1.8V
(3.3V tolerant - DDC
96 DP1_HPD Display Port 1 or HDMI Hot Plug Detect Input CMOS – 1.8V
94 HDMI_CEC HDMI CEC Bidir Open Drain, 1.8V

Note: (Resolution + Refresh Rate + Pixel Depth + Format) must be within specification limits to achieve support for desired
pixel depth.

1.6.2 Embedded DisplayPort (eDP)

Standard
VESA Embedded DisplayPort Standard Version 1.4a

Embedded DisplayPort (eDP) is a mixed-signal interface consisting of four differential serial output lanes and one PLL. This
PLL is used to generate a high-frequency bit-clock from an input pixel clock enabling the ability to handle 10-bit parallel data
per lane at the pixel rate for the desired mode. eDP modes consist of 1.6 GHz for RBR; 2.16 GHz, 2.43 GHz, and 2.7 GHz for
HBR; 3.24 GHz, 4.32 GHz, 5.4 GHz for HBR2, and 8.1 Gbps for HBR3.

Note: eDP has been tested according to DP1.2b PHY CTS even though eDPv1.4 supports lower swing voltages and
additional intermediate bit rates. This means the following nominal voltage levels (400mV, 600mV, 800mV,
1200mV) and data rates (RBR, HBR, HBR2) are tested. This interface can be tuned to drive lower voltage swings
below 400mV and can be programmed to other intermediate bit rates as per the requirements of the panel and the
system designer.

The eDP block collects pixels from the output of the display pipeline, formats/encodes them to the eDP format, and then
streams them to various output devices. It drives local panels only (does not support an external DP port), and it includes a
small test pattern generator and CRC generator.

Features:
• 1/2/4/ lane, single link
• Additional link rates (2.16, 2.43, 3.24, 4.32 Gbps)
• Enhanced framing
• Power sequencing
• Reduced auxiliary timing
• Reduced main voltage swing
• ASSR (alternate seed scrambler reset) for internal eDP panels
•
Note: For eDP pin information, refer to Table 5 HDMI/DisplayPort/eDP Pin Descriptions.

1.7 High-Definition Audio-Video Subsystem

Standard
High Definition Audio Specification Version 1.0a

The HD Audio-Video Subsystem uses a collection of functional blocks to off-load audio and video processing activities from
the CPU complex, resulting in fast, fully concurrent, and highly efficient operation. This subsystem is comprised of the
following:
• (2x) Multi-standard video decoder
• (2x) Multi-standard video encoder
• JPEG processing block
• Video Image Compositor (VIC)
• Audio Processing Engine (APE)
• High Definition Audio (HDA)

1.7.1 Multi-Standard Video Decoder

The Jetson Xavier NX incorporates two instances of the NVIDIA Multi-Standard Video Decoder (NVDEC). This video decoder
accelerates video decode, supporting low resolution mobile content, Standard Definition (SD), High Definition (HD), and
UltraHD (8K, 4K, etc.) video profiles. The video decoder is designed to be extremely power efficient without sacrificing
performance. The video decoder communicates with the memory controller through the video DMA which supports a variety of
memory format output options. For low power operations, the video decoder can operate at the lowest possible frequency
while maintaining real-time decoding using dynamic frequency scaling techniques.

Table 6: Video Decoder Standards

Standard Profile(s) Resolution (Max Number of Streams) Throughput (up to)

H.265 (HEVC) Main, Main10 2160p60 (2) | 2160p30 (4) | 1080p60 (12) (2x) 690 MP/s
| 1080p30 (32)

Standard Profile(s) Resolution (Max Number of Streams) Throughput (up to)

Supported Video Standards - Decode

• Bitrate (up to): 15 Mbps for 1080p | 60 Mbps for 2160p | 240 Mbps for 4320p
• Maximum throughput is half for YUV444 compared to YUV420
• Note: A/V codec, post-processing, and containers support are subject to software support; refer to NVIDIA software
release notes for detailed specifications. Additional audio codecs may be supported using 3rd parties.

1.7.2 Multi-Standard Video Encoder

The Jetson Xavier NX incorporates two instances of the NVIDIA Multi-Standard Video Encoder (NVENC). This multi-standard
video encoder enables full hardware acceleration of various encoding standards. It performs high-quality video encoding
operations for mobile applications such as video recording and video conferencing. The encode processor is designed to be
extremely power efficient without sacrificing performance.

Table 7: Video Encoder Standards

Standard Profile(s) Resolution (Max Number of Streams) Throughput (up to)

H.265 (HEVC) Main, Main10 2160p30 (2) | 1080p60 (6) | 1080p30 (14) (2x) 464 MP/s
Main 4:4:4, Main 4:4:4 10, MV 1080p60 (2) | 1080p30 (6) (2x) 232 MP/s
(per view)
H.264 Baseline, Main, High 2160p30 (2) | 1080p60 (6) | 1080p30 (14) (2x) 455 MP/s
High 444, High 444 1080p60 (2) | 1080p30 (6) (2x) 227 MP/s
Predictive, MVC (per view)
VP9 profile 0 2160p30 (2) | 1080p60 (4) | 1080p30 (8) (2x) 278 MP/s

Supported Video Standards - Encode

• Bitrate (up to): 15 Mbps for 1080p | 60 Mbps for 2160p
• Maximum throughput is half for YUV444 compared to YUV420
• Note: A/V codec, post-processing, and containers support are subject to software support; refer to NVIDIA software
release notes for detailed specifications. Additional audio codecs may be supported using 3rd parties.

Features:
• Timestamp for Audio/Video Sync
• CBR and VBR rate control (supported in firmware)
• Programmable intra-refresh for error resiliency
• Macro-block based and bit based packetization (multiple slice)
• Motion estimation (ME) only mode

1.7.3 JPEG Processing Block

The JPEG processing block is responsible for JPEG (de)compression calculations (based on JPEG still image standard),
image scaling, decoding (YUV420, YUV422H/V, YUV444, YUV400), and color space conversion (RGB to YUV; decode only).
Input (encode) formats:

• Pixel width: 8 bpc

• Subsample format: YUV420
• Resolution up to 16K x 16K
• Pixel pack format
o Semi-planar/planar for 420
Output (decode) formats:

• Pixel width 8 bpc

• Resolution up to 16K x 16K
• Pixel pack format
o Semi-planar/planar for YUV420
o YUY2/planar for 422H/422V
o Planar for YUV444
o Interleave for RGBA

1.7.4 Video Image Compositor (VIC)

VIC implements various 2D image and video operations in a power-efficient manner. It handles various system UI scaling,
blending, and rotation operations, video post-processing functions needed during video playback, and advanced de-noising
functions used for camera capture.
Features:
• Color Decompression
• High-quality Deinterlacing
• Inverse Teleciné
• Temporal Noise Reduction
o New Bilateral Filter as spatial filter
o Improved TNR3 algorithm
• Scaling
• Color Conversion
• Memory Format Conversion
• Blend/Composite
• 2D Bit BLIT operation
• Rotation
• Geometry transform processing
o Programmable nine-points controlled warp patch for distortion correction
o Real-time on-the-fly position generation from sparse warp map surface
o Pincushion/barrel/moustache distortion correction
o Distortion correction of 180- and 360-degree wide FOV lens
o Scene perspective orientation adjustment with IPT
o Full warp map capability
o Non-fixed Patch size with 4x4 regions
o External Mask bit map surface

1.7.5 Audio Processing Engine (APE)

The Audio Processing Engine (APE) is a self-contained unit with dedicated audio clocking that enables Ultra Low Power (ULP)
audio processing. Software based post processing effects enable the ability to implement custom audio algorithms.

Features:
• 96 KB Audio RAM
• Audio Hub (AHUB) I/O Modules
o 2xI2S/3xDMIC/2xDSPK Audio Hub (AHUB) Internal Modules
• Sample Rate converter

• Mixer
• Audio Multiplexer
• Audio De-multiplexer
• Master Volume Controller
• Multi-Channel IN/OUT
o Digital Audio Mixer: 10-in/5-out
- Up to eight channels per stream
- Simultaneous Multi-streams
- Flexible stream routing
o Parametric equalizer: up to 12 bands
o Low latency sample rate conversion (SRC) and high-quality asynchronous sample rate conversion (ASRC)

1.7.6 High Definition Audio (HDA)

Standard
Intel High Definition Audio Specification Revision 1.0a

The Jetson Xavier NX implements an industry-standard High Definition Audio (HDA) controller. This controller provides a
multi-channel audio path to the HDMI interface. The HDA block also provides an HDA-compliant serial interface to an audio
codec. Multiple input and output streams are supported.

Features:
• Supports HDMI 2.0 and DP1.4
• Support up to two audio streams for use with HDMI/DP
• Supports striping of audio out across 1,2,4[a] SDO lines
• Supports DVFS with maximum latency up to 208 µs for eight channels
• Supports two internal audio codecs
• Audio Format Support
o Uncompressed Audio (LPCM): 16/20/24 bits at 32/44.1/48/88.2/96/176.4/192 [b] kHz
o Compressed Audio format: AC3, DTS5.1, MPEG1, MPEG2, MP3, DD+, MPEG2/4 AAC, TrueHD, DTS-HD
[a] Four SDO lines: cannot support one stream, 48 kHz, 16-bits, two channels; for this case, use a one or two SDO line
configuration.
[b] DP protocol sample frequency limitation: cannot support >96 kHz; i.e., does not support 176.4 kHz and 196 kHz.

1.8 Interface Descriptions

The following sections outline the interfaces available on the Jetson Xavier NX module and details the module pins used to
interact with and control each interface. See the Jetson Xavier NX Product Design Guide for complete functional descriptions,
programming guidelines, and register listings for each of these blocks.

1.8.1 SD/eMMC
Standard Notes
SD Specifications, Part A2, SD Host Controller Standard Specification,
Version 4.1
SD Specifications, Part 1, Physical Layer Specification, Version 4.2
SD Specifications, Part 1, eSD (Embedded SD) Addendum, Version 2.10
SD Specifications, Part E1, SDIO Specification Version, 4.1 Support for SD 4.0 Spec without UHS-II
JEDEC Standard, Embedded Multimedia Card (eMMC) Electrical JESD84-B51
Standard 5.1

The SecureDigital (SD)/Embedded MultiMediaCard (eMMC) controller is capable of interfacing to an external SD card or SDIO
device and provides the interface for the on-module eMMC. It has a direct memory controller interface and is capable of
initiating data transfers between system memory and an external card or device. It also has an AMBA Peripheral Bus (APB)
slave interface to access its configuration registers. To access the on-system RAM for MicroBoot, the SD/MMC controller
relies on the path to System RAM in the memory controller.

Features:
• 8-bit data interface to on-module eMMC
• 4-bit data interface for SD cards/SDIO
• Supports card interrupts for SD cards (4-bit SD modes) and SDIO devices
• Supports read wait control and suspend/resume operation for SD cards
• Supports FIFO overrun and underrun condition by stopping SD clock
• Supports addressing larger capacity SD 3.0 or SD-XC cards up to 2 TB

Table 8: SD/SDIO Pin Descriptions

Pin # Signal Name Description Direction Pin Type

229 SDMMC_CLK SD Card or SDIO Clock Output CMOS – 1.8V/3.3V
227 SDMMC_CMD SD Card or SDIO Command Bidir CMOS – 1.8V/3.3V
219 SDMMC_DAT0 SD Card or SDIO Data 0 Bidir CMOS – 1.8V/3.3V
221 SDMMC_DAT1 SD Card or SDIO Data 1 Bidir CMOS – 1.8V/3.3V
223 SDMMC_DAT2 SD Card or SDIO Data 2 Bidir CMOS – 1.8V/3.3V
225 SDMMC_DAT3 SD Card or SDIO Data 3 Bidir CMOS – 1.8V/3.3V

1.8.2 Universal Serial Bus (USB)

Standard Notes
Universal Serial Bus Specification Revision 3.1 Host mode only
Universal Serial Bus Specification Revision 3.0 Device mode only
Universal Serial Bus Specification Revision 2.0 USB Battery Charging Specification, version 1.2;
including Data Contact Detect protocol
Modes: Host and Device
Speeds: Low, Full, and High
Enhanced Host Controller Interface Specification for
Universal Serial Bus revision 1.0

An xHCI/Device controller (named XUSB) supports the xHCI programming model for scheduling transactions and interface
managements as a host that natively supports USB 3.1, USB 2.0, and USB 1.1 transactions with its USB 3.1 and USB 2.0
interfaces. The XUSB controller supports USB 2.0 L1 and L2 (suspend) link power management and USB 3.1 U1, U2, and U3
(suspend) link power managements. The XUSB controller supports remote wakeup, wake on connect, wake on disconnect,
and wake on overcurrent in all power states, including sleep mode.

1.8.2.1 USB 2.0 Operation

Each USB 2.0 port (3x) operates in USB 2.0 high-speed mode when connecting directly to a USB 2.0 peripheral and operates
in USB 1.1 full- and low-speed modes when connecting directly to a USB 1.1 peripheral. When operating in High-Speed mode,
each USB 2.0 port is allocated with one High-Speed unit bandwidth. Approximately a 480 Mb/s bandwidth is allocated to each
USB 2.0 port. All USB 2.0 ports operating in full- or low-speed modes share one full- and low-speed bus instance, which
means 12 Mb/s theoretical bandwidth is distributed across these ports.

Table 9: USB 2.0 Pin Descriptions

Pin # Signal Name Description Direction Pin Type

109 USB0_D_N USB 2.0 Port 0 Data– Bidir USB PHY
111 USB0_D_P USB 2.0 Port 0 Data+ Bidir USB PHY
115 USB1_D_N USB 2.0 Port 1 Data– Bidir USB PHY
117 USB1_D_P USB 2.0 Port 1 Data+ Bidir USB PHY
121 USB2_D_N USB 2.0 Port 2 Data– Bidir USB PHY
123 USB2_D_P USB 2.0 Port 2 Data+ Bidir USB PHY

1.8.2.2 USB 3.1 Operation

The USB 3.1 port supports:
• Generation 1 - SuperSpeed USB (5 Gbps transfer rates)
• Generation 2 - SuperSpeed USB (10 Gbps transfer rates)

Table 10: USB 3.1 Pin Descriptions

Pin # Signal Name Description Direction Pin Type

161 USBSS_RX_N USB SS Receive– (USB 3.1 Ctrl #0) Input USB SS PHY, AC-Coupled (off the
module)
163 USBSS_RX_P USB SS Receive+ (USB 3.1 Ctrl #0) Input USB SS PHY, AC-Coupled (off the
module)
166 USBSS_TX_N USB SS Transmit– (USB 3.1 Ctrl #0) Output USB SS PHY, AC-Coupled on carrier
board
168 USBSS_TX_P USB SS Transmit+ (USB 3.1 Ctrl #0) Output USB SS PHY, AC-Coupled on carrier
board

1.8.3 PCI Express (PCIe)

Standard Notes
PCI Express Base Specification Revision 3.0 Jetson Xavier NX meets the timing requirements for the Gen3 (8.0
GT/s) for x1 and Gen4 (16.0 GT/s) for x4 data rates. Refer to
specification for complete interface timing details.
Although NVIDIA validates that the design complies with the PCIe
specification, PCIe software support may be limited.

The Jetson Xavier NX module integrates two PCIe controllers supporting:

• Connections to two interfaces (1x1 + 1x4).
• x1 (supports Root Port only), x4 (supports Root Port or Endpoint modes) Upstream and downstream AXI interfaces
that serve as the control path from the Jetson Xavier NX to the external PCIe device
• Gen3 (8 GT/s), supported on x1
• Gen4 (16 GT/s) supported on x4
• Two PCIe controllers, five lanes for a total of 144 GT/s. One controller operates in x1 mode only. The second
controller can operate in x1 or x2 or x4 mode.

Table 11: PCIe Pin Descriptions

Pin # Signal Name Description Direction Pin Type

131 PCIE0_RX0_N PCIe #0 Receive 0– (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 0) carrier board if direct connect
133 PCIE0_RX0_P PCIe #0 Receive 0+ (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 0) carrier board if direct connect
137 PCIE0_RX1_N PCIe #0 Receive 1– (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 1) carrier board if direct connect
139 PCIE0_RX1_P PCIe #0 Receive 1+ (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 1) carrier board if direct connect
149 PCIE0_RX2_N PCIe #0 Receive 2– (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 2) carrier board if direct connect
151 PCIE0_RX2_P PCIe #0 Receive 2+ (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 2) carrier board if direct connect
155 PCIE0_RX3_N PCIe #0 Receive 3– (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 3) carrier board if direct connect
157 PCIE0_RX3_P PCIe #0 Receive 3+ (PCIe Ctrl #5 Input PCIe PHY, AC-Coupled for PCIe on
Lane 3) carrier board if direct connect
134 PCIE0_TX0_N PCIe #0 Transmit 0– (PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 0) board
136 PCIE0_TX0_P PCIe #0 Transmit 0+ (PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 0) board
140 PCIE0_TX1_N PCIe #0 Transmit 1– PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 1) board
142 PCIE0_TX1_P PCIe #0 Transmit 1+ (PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 1) board
148 PCIE0_TX2_N PCIe #0 Transmit 2– (PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 2) board
150 PCIE0_TX2_P PCIe #0 Transmit 2+ (PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 2) board
154 PCIE0_TX3_N PCIe #0 Transmit 3– (PCIe Ctrl #5 Output PCIe PHY, AC-Coupled on carrier
Lane 3) board
156 PCIE0_TX3_P PCIe #0 Transmit 3+ (PCIe Ctrl Output PCIe PHY, AC-Coupled on carrier
#5Lane 3) board
181 PCIE0_RST* PCIe #0 Reset (PCIe Ctrl #5). 4.7kΩ Bidir Open Drain 3.3V, Pull-up on the
pull-up to 3.3V on the module. module
180 PCIE0_CLKREQ* PCIe #0 Clock Request (PCIe Ctrl #5). Bidir Open Drain 3.3V, Pull-up on the
47kΩ pull-up to 3.3V on the module. module
160 PCIE0_CLK_N PCIe #0 Reference Clock– Bidir PCIe PHY
Mux controlled by GPIO
162 PCIE0_CLK_P PCIe #0 Reference Clock+ Output PCIe PHY
Mux controlled by GPIO
167 PCIE1_RX0_N PCIe #1 Receive 0– (PCIe Ctrl #4 Input PCIe PHY, AC-Coupled on carrier
Lane 0) board if direct connect.
169 PCIE1_RX0_P PCIe #1 Receive 0+ (PCIe Ctrl #4 Input PCIe PHY, AC-Coupled on carrier
Lane 0) board if direct connect.

Pin # Signal Name Description Direction Pin Type

172 PCIE1_TX0_N PCIe #1 Transmit 0– (PCIe Ctrl #4 Output PCIe PHY, AC-Coupled on carrier
Lane 0) board
174 PCIE1_TX0_P PCIe #1 Transmit 0+ (PCIe Ctrl #4 Output PCIe PHY, AC-Coupled on carrier
Lane 0) board
183 PCIE1_RST* PCIe #1 Reset (PCIe Ctrl #4). 4.7kΩ Output Open Drain 3.3V, Pull-up on the
pull-up to 3.3V on the module. module
182 PCIE1_CLKREQ* PCIe #1 Clock Request (PCIe Ctrl #4). Bidir Open Drain 3.3V, Pull-up on the
47kΩ pull-up to 3.3V on the module. module
173 PCIE1_CLK_N PCIe #1 Reference Clock– (PCIe Ctrl Output PCIe PHY
#4)
175 PCIE1_CLK_P PCIe #1 Reference Clock+ (PCIe Ctrl Output PCIe PHY
#4)
179 PCIE_WAKE* PCIe Wake. 100kΩ pull-up to 3.3V on Input Open Drain 3.3V, Pull-up on the
the module. module

Note: Upstream Type 1 Vendor Defined Messages (VDM) should be sent by the Endpoint Port if the Root Port also belongs
to same vendor/partner; otherwise the VDM is silently discarded.

See the Jetson Xavier NX Product Design Guide for supported USB 3.1/PCIe configuration and connection examples.

1.8.4 Serial Peripheral Interface (SPI)

The Serial Peripheral Interface (SPI) controller allows a duplex, synchronous, serial communication between the controller and
external peripheral devices; it supports both Master and Slave modes of operation on the SPI bus. See the Jetson Xavier NX
Product Design Guide for more information.

Features:
• 2x SPI Interface
• Maximum data rate: 65 Mbps in Master Mode, 50 Mbps in Slave Mode
• Master mode operation
o All transfer modes (Mode 0, Mode 1, Mode 2, Mode 3) supported for both transmit and receive transactions
• FIFO Size: 64 x 32 bits
• Programmable packet sizes of 4 to 32 bits
• Programmable clock phase and polarity
• Programmable delay between consecutive transfers
• Chip select controllable by software or generated by hardware on packet boundaries

Table 12: SPI Pin Descriptions

Pin # Signal Name Description Direction Pin Type

91 SPI0_SCK SPI 0 Clock Bidir CMOS – 1.8V
89 SPI0_MOSI SPI 0 Master Out / Slave In Bidir CMOS – 1.8V
93 SPI0_MISO SPI 0 Master In / Slave Out Bidir CMOS – 1.8V
95 SPI0_CS0* SPI 0 Chip Select 0 Bidir CMOS – 1.8V
97 SPI0_CS1* SPI 0 Chip Select 1 Bidir CMOS – 1.8V
106 SPI1_SCK SPI 1 Clock Bidir CMOS – 1.8V
104 SPI1_MOSI SPI 1 Master Out / Slave In Bidir CMOS – 1.8V

Pin # Signal Name Description Direction Pin Type

108 SPI1_MISO SPI 1 Master In / Slave Out Bidir CMOS – 1.8V
110 SPI1_CS0* SPI 1 Chip Select 0 Bidir CMOS – 1.8V
112 SPI1_CS1* SPI 1 Chip Select 1 Bidir CMOS – 1.8V

Figure 1 SPI Master Timing Diagram

tCS
SPIx_CSx_N tCH tCL
tCSS tCSH tCSS tCSH

SPIx_SCK 0 n
tSU tHD
SPIx_MISO MSB IN LSB IN
tDD
SPIx_MOSI MSB OUT LSB OUT

Table 13: SPI Master Timing Parameters

Symbol Parameter Minimum Maximum Unit

Fsck SPIx_SCK clock frequency – 65 MHz
Psck SPIx_SCK period 1/Fsck – ns
tCH SPIx_SCK high time 50%Psck -10% 50%Psck +10% ns
tCL SPIx_SCK low time 50%Psck -10% 50%Psck +10% ns
tCRT SPIx_SCK rise time (slew rate) 0.1 – V/ns
tCFT SPIx_SCK fall time (slew rate) 0.1 – V/ns
tSU SPIx_MISO setup to SPIx_SCK rising edge 2 – ns
tHD SPIx_MISO hold from SPIx_SCK rising edge 3 – ns
tDD Active Clock edge to MOSI data output valid – 6
tCSS SPIx_CSx setup time 2 – ns
tCSH SPIx_CSx hold time 3 – ns
tCS SPIx_CSx high time 10 – ns

1.8.5 Universal Asynchronous Receiver/Transmitter (UART)

The UART controller provides serial data synchronization and data conversion (parallel-to-serial and serial-to-parallel) for both
receiver and transmitter sections. Synchronization for serial data stream is accomplished by adding start and stop bits to the
transmit data to form a data character. Data integrity is accomplished by attaching a parity bit to the data character. The parity
bit can be checked by the receiver for any transmission bit errors.
Note: The UART receiver input has low baud rate tolerance in 1-stop bit mode. External devices must use two stop bits.

In 1-stop bit mode, the UART receiver can lose sync between the receiver and the external transmitter resulting in data
errors/corruption. In 2-stop bit mode, the extra stop bit allows the UART receiver logic to align properly with the UART
transmitter.

Features:
• 3x UART Interface

• Synchronization for the serial data stream with start and stop bits to transmit data and form a data character
• Supports both 16450- and 16550-compatible modes. Default mode is 16450
• Device clock up to 200 MHz, baud rate of 12.5 Mbits/second
• Support for word lengths from five to eight bits, an optional parity bit and one or two stop bits
• Support for modem control inputs
• Auto sense baud detection
• Timeout interrupts to indicate if the incoming stream stopped
• Priority interrupts mechanism
• Flow control support on RTS and CTS
• SIR encoding/decoding (3/16 or 4/16 baud pulse widths to transmit bit zero)

Table 14: UART Pin Descriptions

Pin # Signal Name Description Direction Pin Type

99 UART0_TXD UART #0 Transmit Output CMOS – 1.8V
101 UART0_RXD UART #0 Receive Input CMOS – 1.8V
103 UART0_RTS* UART #0 Request to Send Output CMOS – 1.8V
105 UART0_CTS* UART #0 Clear to Send Input CMOS – 1.8V
203 UART1_TXD UART #1 Transmit Output CMOS – 1.8V
205 UART1_RXD UART #1 Receive Input CMOS – 1.8V
207 UART1_RTS* UART #1 Request to Send Output CMOS – 1.8V
209 UART1_CTS* UART #1 Clear to Send Input CMOS – 1.8V
236 UART2_TXD UART #2 Transmit Output CMOS – 1.8V
238 UART2_RXD UART #2 Receive Input CMOS – 1.8V

1.8.6 Controller Area Network (CAN)

Standard Notes
ISO/DIS 16845-2 CAN conformance test
ISO 11898-1:2015 Data link layer and physical signaling; CAN FD Frame formats
ISO 11898-4:2004 Time-triggered communication

The Jetson Xavier NX integrates the Bosch Time-Triggered Controller Area Network (M_TTCAN) controller version 3.2.0. One
independent CAN port/channel supports connectivity to one CAN network. Each port instantiates the Bosch M_TTCAN
module, a message RAM module, an APB slave interface module, interrupt aggregator, time-triggered control module, and a
wake detect module. All M_TTCAN external modules have direct connections to M_TTCAN except for the wake detect
module.
Features:
• Standard frame and extended frame transmission/reception enable
• Transfer rate: programmable bit rates up to 15 Mbps
• 0 – 8-byte data length, with the ability to receive the first 8 bytes when data length coding is > 8 Bytes
• 32 message buffers per channel
• Prioritization of transmit buffers
• Receive/transmit history list function
• Automatic block transmission function
• Multi-buffer receives block function
• Flexible maskable identifier filter support for two 32-bit, or four 16-bit, or eight 8-bit filters for each channel

• Programmable data bit time, communication baud rate, and sample point.
o As an example, the following sample-point configurations can be configured: 66.7%, 70.0%, 75.0%, 80.0%,
81.3%, 85.0%, and 87.5%
o Baud rates in the range of 10 kbps up to 1000 kbps can be configured
• Enhanced features:
o Each message buffer can be configured to operate as a transmit or a receive message buffer
o Transmission priority is controlled by the identifier or by mailbox number (selectable)
o A transmission request can be aborted by clearing the dedicated Transmit-Request flag of the concerned
message buffer.
o Automatic block transmission (ABT) operation mode
o Time stamp function for CAN channels 0 to n in collaboration with timers
• Release from bus-off state by software
• Wake-up with integrated low-pass filter (debounce) option to prevent short glitches on CAN bus, through CAN receive
signal toggling from CAN transceiver
o For normal operation (after wake) there is a digital filter in the CAN controller
• Listen-only mode to monitor CAN bus
• Wake-up signal to both internal and external (either interrupt signal or GPIO) to initiate power up if needed.
o Ready to receive the first CAN message within 10ms of wake event generated by the CAN master.
o Ready to transmit the first CAN message within 50ms of wake event generated by the CAN master.
• Loop back for self-test

Table 15: CAN Pin Descriptions

Pin # Signal Name Description Direction Pin Type

145 CAN_TX CAN PHY Output CMOS – 3.3V
143 CAN_RX CAN PHY Input CMOS – 3.3V

1.8.7 Inter-Chip Communication (I2C)

Standard Notes
NXP inter-IC-bus (I2C) specification https://i2c.info/i2c-bus-specification

This general purpose I2C controller allows system expansion for I2C-based devices as defined in the NXP inter-IC-bus (I2C)
specification. The I2C bus supports serial device communications to multiple devices. (4x I2C) The I2C controller handles clock
source negotiation, speed negotiation for standard and fast devices, 7-bit slave address support according to the I2C protocol
and supports master and slave modes of operation.

The I2C controller supports the following operating modes:

• Master – Standard-mode (up to 100 Kbit/s), Fast-mode (up to 400 Kbit/s), Fast-mode plus (Fm+, up to 1 Mbit/s);
• Slave – Standard-mode (up to 100 Kbit/s), Fast-mode (up to 400 Kbit/s), Fast-mode plus (Fm+, up to 1 Mbit/s).

Table 16: CAN Pin Descriptions

Pin # Signal Name Description Direction Pin Type

185 I2C0_SCL General I2C 0 Clock. 2.2kΩ pull-up to 3.3V on module. Bidir Open Drain – 3.3V
187 I2C0_SDA General I2C 0 Data. 2.2kΩ pull-up to 3.3V on the module. Bidir Open Drain – 3.3V
189 I2C1_SCL General I2C 1 Clock. 2.2kΩ pull-up to 3.3V on the module. Bidir Open Drain – 3.3V
191 I2C1_SDA General I2C 1 Data. 2.2kΩ pull-up to 3.3V on the module. Bidir Open Drain – 3.3V
232 I2C2_SCL General I2C 2 Clock. 2.2kΩ pull-up to 1.8V on the module. Bidir Open Drain – 1.8V
234 I2C2_SDA General I2C 2 Data. 2.2kΩ pull-up to 1.8V on the module. Bidir Open Drain – 1.8V

1.8.8 Inter-IC Sound (I2S)

Standard
Inter-IC Sound (I2S) specification

The I2S controller transports streaming audio data between system memory and an audio codec. The I2S controller supports
I2S format, left-justified mode format, right-justified mode format, and DSP mode format, as defined in the Philips inter-IC-
sound (I2S) bus specification.

The I2S and PCM (master and slave modes) interfaces support clock rates up to 24.5760 MHz.

The I2S controller supports point-to-point serial interfaces for the I2S digital audio streams. I2S-compatible products, such as
compact disc players, digital audio tape devices, digital sound processors, and those with digital TV sound may be directly
connected to the I2S controller. The controller also supports the PCM and telephony mode of data-transfer. Pulse-Code-
Modulation (PCM) is a standard method used to digitize audio (particularly voice) patterns for transmission over digital
communication channels. The Telephony mode is used to transmit and receive data to and from an external mono CODEC in
a slot-based scheme of time-division multiplexing (TDM). The I2S controller supports Bidirectional audio streams and can
operate in half-duplex or full-duplex mode.
Features:
• Basic I2S modes to be supported (I2S, RJM, LJM, and DSP) in both master and slave modes
• PCM mode with short (one bit-clock wide) and long-fsync (two bit-clock wide) in both master and slave modes
• NW-mode with independent slot-selection for both transmit and receive
• TDM mode with flexibility in number of slots and slot(s) selection
• Capability to drive-out a high-z outside the prescribed slot for transmission
• Flow control for the external input/output stream

Table 17: TDM Timing Parameters (Slave Mode)

Symbol Parameter Minimum Maximum Unit Notes

FSCK Frequency – 24.576 MHz
TCYL I2Sx_SCLK cycle time 1/FSCK – ns
TFDLY I2Sx_LRCK delay 0 4.5 ns
tDDLY I2Sx_SDOUT delay 0 4.5 ns
tDSU I2Sx_SDIN setup time 2 – ns
tDH I2Sx_SDIN hold time 2 – ns
tRT I2Sx_SCLK rise time – 5% * TCYL ns
tFT I2Sx_SCLK fall time – 5% * TCYL ns
tCH I2Sx_SCLK high time 45% * TCYL – ns
tCL I2Sx_SCLK low time 45% * TCYL – ns

Table 18: TDM Timing Parameters (Master Mode)

Symbol Parameter Minimum Maximum Unit Notes

FSCK Frequency – 24.576 MHz
TCYL I2Sx_SCLK cycle time 1/FSCK – ns
tDDLY I2Sx_SDOUT delay 0 4.5 ns
tDSU I2Sx_SDIN setup time 2 – ns

Symbol Parameter Minimum Maximum Unit Notes

tDH I2Sx_SDIN hold time 2 – ns
tFSU I2Sx_LRCK setup 2 45% * TCYL - 2 ns 1

tFSH I2Sx_LRCK hold 55% TCYL + 2 – ns 2

tRT I2Sx_SCLK rise time – 5% * TCYL ns

tFT I2Sx_SCLK fall time – 5% * TCYL ns
tCH I2Sx_SCLK high time 45% * TCYL – ns
tCL I2Sx_SCLK low time 45% * TCYL – ns
1. Maximum tFSU requirement only applies while Fsync Launching on Clock Raising Edge
2. Minimum tFSH (55% TCYL + 2) requirement only applies while Fsync Launching on Clock Raising Edge; in other use cases,
Minimum tFSH is 2ns.

Table 19: I2S Pin Descriptions

Pin # Signal Name Description Direction Pin Type

199 I2S0_SCLK I2S Audio Port 0 Clock Bidir CMOS – 1.8V
197 I2S0_FS I2S Audio Port 0 Left/Right Clock Bidir CMOS – 1.8V
193 I2S0_DOUT I2S Audio Port 0 Data Out Output CMOS – 1.8V
195 I2S0_DIN I2S Audio Port 0 Data In Input CMOS – 1.8V
226 I2S1_SCLK I2S Audio Port 1 Clock Bidir CMOS – 1.8V
224 I2S1_FS I2S Audio Port 1 Left/Right Clock Bidir CMOS – 1.8V
220 I2S1_DOUT I2S Audio Port 1 Data Out Output CMOS – 1.8V
222 I2S1_DIN I2S Audio Port 1 Data In Input CMOS – 1.8V

1.8.9 Gigabit Ethernet

Standard Notes
Gigabit Ethernet (GbE) IEEE 802.3ab

The Jetson Xavier NX integrates a Realtek RTL8211FDI Gigabit Ethernet controller. The on-module Ethernet controller
supports:
• 10/100/1000 Gigabit Ethernet
• IEEE 802.3u Media Access Controller (MAC)

Table 20: Gigabit Ethernet Pin Descriptions

Pin # Signal Name Description Direction Pin Type

184 GBE_MDI0_N GbE Transformer Data 0– Bidir MDI
186 GBE_MDI0_P GbE Transformer Data 0+ Bidir MDI
190 GBE_MDI1_N GbE Transformer Data 1– Bidir MDI
192 GBE_MDI1_P GbE Transformer Data 1+ Bidir MDI
196 GBE_MDI2_N GbE Transformer Data 2– Bidir MDI
198 GBE_MDI2_P GbE Transformer Data 2+ Bidir MDI

Pin # Signal Name Description Direction Pin Type

202 GBE_MDI3_N GbE Transformer Data 3– Bidir MDI
204 GBE_MDI3_P GbE Transformer Data 3+ Bidir MDI
188 GBE_LED_LINK Ethernet Link LED (Green) Output
194 GBE_LED_ACT Ethernet Activity LED (Yellow) Output

1.8.10 Fan
The Jetson Xavier NX includes a Pulse Width Modulator (PWM) and Tachometer functionality to enable fan control as part of a
thermal solution. The PWM controller is a frequency divider with a varying pulse width. The PWM runs off a device clock
programmed in the Clock and Reset controller and can be any frequency up to the device clock maximum speed of 48 MHz.
The PWM gets divided by 256 before being subdivided based on a programmable value.

1.8.11 Pulse Width Modulator (PWM)

Jetson Xavier NX has four Pulse Width Modulator (PWM) outputs. Each PWM output is based on a frequency divider whose
pulse width varies. Each has a programmable frequency divider and a programmable pulse width generator. The PWM
controller supports one PWM output for each of its four instances. Each instance is allocated a 64 KB independent address
space.

Frequency division is a 13-bit programmable value, and pulse division is an 8-bit value. The PWM can run at a maximum
frequency of up to 408 MHz. The PWM controller can source its clock from either CLK_M or PLLP. CLK_M (19.2 MHz) is
derived from the OSC clock (38.4 MHz). PLLP operates at 408 MHz.

The PWM clock frequency is divided by 256 before subdividing it based on the programmable frequency division value to
generate the required frequency for the PWM output. The maximum output frequency that can be achieved from this
configuration is 408 MHz/256 = 1.6 MHz. This 1.6 MHz frequency can be further divided using the frequency divisor in PWM.

The OSC clock is the primary/default source for the PWM IP clock. For higher PWM output frequency requirements, PLLP is
the clock source (up to 408 MHz).

Table 21: PWM Pin Descriptions

Pin # Signal Name Description Direction Pin Type

206 GPIO07 Pulse Width Modulator or GPIO #7 Bidir CMOS – 1.8V
218 GPIO12 Pulse Width Modulator or GPIO #12 Bidir CMOS – 1.8V
228 GPIO13 Pulse Width Modulator or GPIO #13 Bidir CMOS – 1.8V
230 GPIO14 Pulse Width Modulator or GPIO #14 Bidir CMOS – 1.8V

1.9 Deep Learning Accelerator (DLA)

The Deep Learning Accelerator (DLA) is a new fixed function engine used to accelerate inference operations on convolution
neural networks (CNNs). The DLA supports accelerating some or all desired CNN layers such as convolution, activation,
pooling, local response normalization, and full-connected layers.

Table 22: DLA Clock

Module Power Mode Max Frequency

Jetson Xavier NX 10W 900 MHz
15W 1100 MHz

DLA hardware is comprised of the following components:

1. Convolution Core – optimized high-performance convolution engine.
Convolution operations work on two sets of data: one set of offline-trained “weights” (which remain constant between
each run of inference), and one set of input “feature” data (which varies with the network’s input). The convolutional
engine exposes parameters to map many different sizes of convolutions onto the hardware with high efficiency.
2. Single Data Point Processor – single-point lookup engine for activation functions.
The Single Data Point Processor (SDP) allows for the application of both linear and non-linear functions onto
individual data points. This is commonly used immediately after convolution in CNN systems. The SDP has a lookup
table to implement non-linear functions, or for linear functions it supports simple bias and scaling. This combination
can support most common activation functions, as well as other element-wise operations, including ReLU, PReLU,
precision scaling, batch normalization, bias addition, or other complex non-linear functions, such as a sigmoid or a
hyperbolic tangent.
3. Planar Data Processor – planar averaging engine for pooling.
The Planar Data Processor (PDP) supports specific spatial operations that are common in CNN applications. It is
configurable at runtime to support different pool group sizes, and supports three pooling functions: maximum-pooling,
minimum-pooling, and average-pooling.
4. Cross-Channel Data Processor – multi-channel averaging engine for advanced normalization functions.
The Cross-channel Data Processor (CDP) is a specialized unit built to apply the local response normalization (LRN)
function – a special normalization function that operates on channel dimensions, as opposed to the spatial
dimensions.
5. Data Reshape Engines – memory-to-memory transformation acceleration for tensor reshape and copy operations.
The data reshape engine performs data format transformations (e.g., splitting or slicing, merging, contraction,
reshape-transpose). Data in memory often needs to be reconfigured or reshaped in the process of performing
inferencing on a convolutional network. For example, slice operations may be used to separate out different features
or spatial regions of an image, while reshape-transpose operations (common in deconvolutional networks) create
output data with larger dimensions than the input dataset.
6. Bridge DMA – accelerated path to move data between two non-connected memory systems.
The bridge DMA (BDMA) module provides a data copy engine to move data between the system DRAM and the
dedicated memory interface.

2.0 Power and System Management

VIN must be supplied by the carrier board that the module is designed to connect to. All interfaces are referenced to on-
module voltage rails, additional I/O voltage is not required to be supplied to the module. See the Jetson Xavier NX Product
Design Guide for details on connecting to each of the interfaces.

Table 23: Power and System Control Pin Descriptions

Pin # Signal Name Description Direction Pin Type

251 VDD_IN Main power – Supplies PMIC and other registers Input 5.0V
252
253
254
255
256
257
258
259
260
235 PMIC_BBAT PMIC Battery Back-up. Optionally used to provide back-up Bidir 1.65V-5.5V
power for the Real-Time Clock (RTC). Connects to Lithium
Cell or super capacitor on Carrier Board. PMIC is supply
when charging cap or coin cell. Super cap or coin cell is
source when system is disconnected from power.
233 SHUTDOWN_REQ* Used by the module to request a shutdown from the carrier Input CMOS –
board. Pull up to VDD_IN with ~5kΩ (4.02k + 1k) on the 5.0V
module.
237 POWER_EN Signal for module on/off: high level on, low level off. Input CMOS –
Connects to module PMIC EN0 through converter logic. 5.0V
POWER_EN is routed to a Schmitt trigger buffer on the
module. A 100kΩ pulldown is also on the module.
239 SYS_RESET* Module Reset. Reset to the module when driven low by the Bidir Open
carrier board. Used as carrier board supply enable when Drain, 1.8V
driven high by the module when module power sequence is
complete. Used to ensure proper power on/off sequencing
between module and carrier board supplies. Pull up to 1.8V
with 1kΩ resistor on the module.
178 MOD_SLEEP* Module Sleep. When active (low), indicates module has Output CMOS –
gone to Sleep (SC7) mode. 1.8V
210 CLK_32K_OUT Sleep/Suspend clock Output CMOS –
1.8V
214 FORCE_RECOVERY* Force Recovery strap pin Input CMOS –
1.8V
240 SLEEP/WAKE* Configured as GPIO for optional use to indicate the system Input CMOS –
should enter or exit sleep mode. 5.0V

2.1 Power Rails

VDD_IN must be supplied by the carrier board that the Jetson Xavier NX is designed to connect to. All Jetson Xavier NX
interfaces are referenced to on-module voltage rails and no I/O voltage is required to be supplied to the module. See the
Jetson Xavier NX Product Design Guide for details of connecting to each of the interfaces.

2.2 Power Domains/Islands

Jetson Xavier NX has a single three-channel INA that can measure power of CPU_GPU_CV combined rail, Core, and module
input power.

2.3 Power Management Controller (PMC)

The PMC power management features enable both high speed operation and very low-power standby states. The PMC
primarily controls voltage transitions for the SoC as it transitions to/from different low power modes; it also acts as a slave
receiving dedicated power/clock request signals as well as wake events from various sources (e.g., SPI, I2C, RTC, USB)
which can wake the system from deep sleep state. The PMC enables aggressive power-gating capabilities on idle modules
and integrates specific logic to maintain defined states and control power domains during sleep and deep sleep modes.

2.4 Resets
If you assert reset, then the Jetson Xavier NX and onboard storage will be reset. This signal is also used for baseboard power
sequencing.

2.5 PMIC_BBATT
An optional back up battery can be attached to the VCC_RTC module input to maintain the module real-time clock (RTC)
when VIN is not present. This pin is connected directly to the onboard PMIC. Details of the types of backup cells that optionally
can be connected are found in the PMIC manufacturer's data sheet. When a backup cell is connected to the PMIC, the RTC
retains its contents and can be configured to charge the backup cell as well. RTC accuracy is 2 seconds/day.
The following backup cells may be attached to this pin:
• Super capacitor (gold cap, double layer electrolytic)
• Standard capacitors (tantalum)
• Rechargeable Lithium Manganese cells

The backup cells must provide a voltage in the range 2.5V to 3.5V. These are charged with a constant current, and a constant
voltage charger that can be configured between 2.5V and 3.5V (constant voltage) output and 50 uA to 800 uA (constant
current).

Table 24: PMIC_BBAT Pin Descriptions

Pin # Signal Name Description Direction Pin Type

235 PMIC_BBAT PMIC Battery Back-up. Optionally used to provide back-up power Bidir 1.65V-5.5V
for the Real-Time Clock (RTC). Connects to Lithium Cell or super
capacitor on Carrier Board. PMIC is supply when charging cap or
coin cell. Super cap or coin cell is source when system is
disconnected from power.

2.6 Power Sequencing

The Jetson Xavier NX is required to be powered on and off in a known sequence. Sequencing is determined through a set of
control signals; the SYS_RESET* signal (when deasserted) is used to indicate when the carrier board can power on. The
following sections provide an overview of the power sequencing steps between the carrier board and Jetson Xavier NX. Refer
to the Jetson Xavier NX Product Design Guide for system level details on the application of power, power sequencing, and
monitoring. The Jetson Xavier NX and the product carrier board must be power sequenced properly to avoid potential damage
to components on either the module or the carrier board system.

2.6.1 Power Up
During power up, the carrier board must wait until the signal SYS_RESET* is deasserted from the Jetson module before
enabling its power; the Jetson module will deassert the SYS_RESET* signal to enable the complete system to boot.

Note: I/O pins cannot be high (>0.5V) before SYS_RESET* goes high. When SYS_RESET* is low, the maximum voltage
applied to any I/O pin is 0.5V.
Figure 2 Power-up Sequence

VDD_IN

POWER_EN

Module Power

SYS_RESET*

Carrier Board Supplies

2.6.2 Power Down

Shutdown events can be triggered by either the module or the baseboard, but the shutdown event will always be serviced by
the baseboard. To do so, the baseboard deasserts POWER_EN, which begins the shutdown power sequence on the module.
If the module needs to request a shutdown event in the case of thermal, software, or under-voltage events, it will assert
SHUTDOWN_REQ*. When the baseboard sees low SHUTDOWN_REQ*, it should deassert POWER_EN as soon as
possible.

Once POWER_EN is deasserted, the module will assert SYS_RESET*, and the baseboard may shut down. SoC 3.3V I/O
must reach 0.5V or lower at most 1.5ms after SYS_RESET* is asserted. SoC 1.8V I/O must reach 0.5V or lower at most 4ms
after SYS_RESET* is asserted.

Figure 3 Power Down Sequence

VDD_IN

SHUTDOWN_REQ*

POWER_EN T < 10 uS

2.7 Power States

The Jetson Xavier NX operates in three main power modes: OFF, ON, and SLEEP. The module transitions between these
states are based on various events from hardware or software. Figure 4 shows the transitions between these three states.

Figure 4 Power State Transition Diagram

ON SLEEP
EVENT EVENT

OFF ON SLEEP

OFF WAKE
EVENT EVENT

2.7.1 ON State
The ON power state is entered from either OFF or SLEEP states. In this state, the Jetson Xavier NX module is fully functional
and operates normally. An ON event must occur for a transition between OFF and ON states. The only ON EVENT currently
used is a low to high transition on the POWER_EN pin. This must occur with VDD_IN connected to a power rail and
POWER_EN is asserted (at a logic1). The POWER_EN control is the carrier board indication to the Jetson module that the
VDD_IN power is good. The carrier board should assert this high only when VDD_IN has reached its required voltage level
and is stable. This prevents the Jetson Xavier NX Module from powering up until the VDD_IN power is stable.

2.7.2 OFF State

The OFF state is the default state when the system is not powered. It can only be entered from the ON state, through an OFF
event. OFF events are listed in the table below.

Table 25: OFF State Events

Event Details Preconditions

HW Shutdown Set POWER_EN pin to zero for at least 100 µs, the internal In ON State
PMIC starts the shutdown sequence.
SW Shutdown Software initiated shutdown ON state, software operational
Thermal Shutdown If the internal temperature of the Jetson Xavier NX module Any power state
reaches an unsafe temperature, the hardware is designed to
initiate a shutdown.

Note: HW shutdown, SW shutdown, and Thermal shutdown will all assert SHUTDOWN_REQ* low. System on Module will not
initiate power supply shutdown sequence until POWER_EN is deasserted. POWER_EN debounce is 1ms on Jetson Xavier
NX.

2.7.3 SLEEP State

The SLEEP state can only be entered from the ON state. This state allows the module to quickly resume to an operational
state without performing a full boot sequence. The SLEEP state also includes a low power mode SC7 (deep sleep) where the
module operates only with enough circuitry powered to allow the device to resume and re-enter the ON state. During this state
the output signals from the module are maintained at their logic level prior to entering the state (i.e., they do not change to a
0V level). To exit the SLEEP state a WAKE event must occur; WAKE events can occur from within the module or from external
devices through various pins on the module connector.

Table 26: SLEEP and WAKE Events

Event Details
RTC WAKE up Timers within the module can be programmed, on SLEEP entry. When these expire, they
create a WAKE event to exit the SLEEP state.
Thermal Condition If the module internal temperature exceeds programmed hot and cold limits the system is
forced to wake up, so it can report and take appropriate action (shut down for example).
USB VBUS detection If VBUS is applied to the system (USB cable attached) then the device can be configured to
Wake and enumerate.
SD Card detect The card detect pin may be configured to enable the system to wake.
Module connector Programmable signals on the module connector.
Interface WAKE signal

2.8 Thermal and Power Monitoring

The Jetson Xavier NX is designed to operate under various workloads and environmental conditions. It has been designed so
that an active or passive heat sink solution can be attached. The module contains various methods through hardware and
software to limit the internal temperature to within operating limits. See the Jetson Xavier NX Thermal Design Guide for more
details.

3.0 Pin Definitions

The function(s) for each pin on the module is fixed to a single Special-Function I/O (SFIO) or software-controlled General
Purpose I/O (GPIO). The Jetson Xavier NX has multiple dedicated GPIOs and each GPIO is individually configurable as
Output/Input/Interrupt sources with level/edge controls. SFIO and GPIO functionality is configured using Multi-Purpose I/O
(MPIO) pads with each MPIO pad consisting of:
• An output driver with tristate capability, drive strength controls and push-pull mode, open-drain mode, or both
• An input receiver with either Schmitt mode, CMOS mode, or both
• A weak pull-up and a weak pull-down

MPIO pads are partitioned into multiple pad control groups with controls being configured for the group. During normal
operation, these per-pad controls are driven by the pinmux controller registers. During deep sleep, the PMC bypasses and
then resets the pinmux controller registers. Software reprograms these registers as necessary after returning from deep sleep.

Refer to the Jetson Xavier NX Product Design Guide for more information.

3.1 Power-on Reset Behavior

Each MPIO pad has a deterministic power-on reset (PoR) state. The reset state for each pad is chosen to minimize the need
of additional on-board components; for example, on-chip weak pull-ups are enabled during PoR for pads which are usually
used to drive active-low chip selects eliminating the need for additional pull-up resistors.

The following list is a simplified description of the Jetson Xavier NX boot process focusing on those aspects which relate to the
MPIO pins:
• System-level hardware executes the power-up sequence. This sequence ends when system-level hardware releases
SYS_RESET_N.
• The boot ROM begins executing and programs the on-chip I/O controllers to access the secondary boot device
(QSPI).
• The boot ROM fetches the Boot Configuration Table (BCT) and boot loader from the secondary boot device.
• If the BCT and boot loader are fetched successfully, the boot ROM transfers control to the boot loader.
• Otherwise, the boot ROM enters USB recovery mode.

3.2 Sleep Behavior

Sleep is an ultra-low-power standby state in which the module maintains much of its I/O state while most of the chip is
powered off. During deep sleep most of the pads are put in a state called Deep Power Down (DPD). The sequence for
entering DPD is same across pads.

MPIO pads can vary during deep sleep. They differ regarding:
• Input buffer behavior during deep sleep
o Forcibly disabled OR
o Enabled for use as a GPIO wake event, OR
o Enabled for some other purpose (e.g., a clock request pin)
• Output buffer behavior during deep sleep
o Maintain a static programmable (0, 1, or tristate) constant value OR
o Capable of changing state (i.e., dynamic while the chip is still in deep sleep)
• Weak pull-up/pull-down behavior during deep sleep
o Forcibly disabled OR
o Can be configured
• Pads that do not enter deep sleep
o Some of the pads whose outputs are dynamic during deep sleep are of special type and they do not enter deep
sleep (e.g., pads that are associated with PMC logic do not enter deep sleep, pads that are associated with
JTAG do not enter into deep sleep any time.

3.3 GPIO
The Jetson Xavier NX has multiple dedicated GPIOs. Each GPIO can be individually configurable as an Output, Input, or
Interrupt source with level/edge controls. The pins listed in the following table are dedicated GPIOs; some with alternate SFIO
functionality. Many other pins not included in this list are capable of being configured as GPIOs instead of the SFIO
functionality the pin name suggests (e.g., UART, SPI, I2S, etc.). All pins that can support GPIO functionality have this exposed
in the Pinmux.

Table 27: GPIO Pin Descriptions

Pin # Signal Name Description Direction Pin Type

87 GPIO00 GPIO #0 or USB 0 VBUS Enable #0 Bidir CMOS – 1.8V
118 GPIO01 GPIO #1 or Generic Clocks Bidir CMOS – 1.8V
124 GPIO02 GPIO #2 Bidir CMOS – 1.8V
126 GPIO03 GPIO #3 Bidir CMOS – 1.8V
127 GPIO04 GPIO #4 Bidir CMOS – 1.8V
128 GPIO05 GPIO #5 Bidir CMOS – 1.8V
130 GPIO06 GPIO #6 Bidir CMOS – 1.8V
206 GPIO07 GPIO #7 or Pulse Width Modulator Bidir CMOS – 1.8V
208 GPIO08 GPIO #8 or SD Card Detect Bidir CMOS – 1.8V
211 GPIO09 GPIO #9 or Audio Codec Master Clock Bidir CMOS – 1.8V
212 GPIO10 GPIO #10 Bidir CMOS – 1.8V
216 GPIO11 GPIO #11 or Generic Clocks Bidir CMOS – 1.8V
218 GPIO12 GPIO #12 or Pulse Width Modulator Bidir CMOS – 1.8V
228 GPIO13 GPIO #13 or Pulse Width Modulator Bidir CMOS – 1.8V
230 GPIO14 GPIO #14 or Pulse Width Modulator Bidir CMOS – 1.8V

3.4 Pin List

Signal Name Pin # Pin # Signal Name Signal Name Pin # Pin # Signal Name
Top Bottom Top Bottom
Odd Even Odd Even
GND 1 2 GND PCIE0_RX0_P 133 134 PCIE0_TX0_N
CSI1_D0_N 3 4 CSI0_D0_N GND 135 136 PCIE0_TX0_P
CSI1_D0_P 5 6 CSI0_D0_P PCIE0_RX1_N 137 138 GND
GND 7 8 GND PCIE0_RX1_P 139 140 PCIE0_TX1_N
CSI1_CLK_N 9 10 CSI0_CLK_N GND 141 142 PCIE0_TX1_P
CSI1_CLK_P 11 12 CSI0_CLK_P CAN_RX 143 144 GND
GND 13 14 GND CAN_TX 145 146 GND
CSI1_D1_N 15 16 CSI0_D1_N GND 147 148 PCIE0_TX2_N
CSI1_D1_P 17 18 CSI0_D1_P PCIE0_RX2_N 149 150 PCIE0_TX2_P
GND 19 20 GND PCIE0_RX2_P 151 152 GND
CSI3_D0_N 21 22 CSI2_D0_N GND 153 154 PCIE0_TX3_N
CSI3_D0_P 23 24 CSI2_D0_P PCIE0_RX3_N 155 156 PCIE0_TX3_P
GND 25 26 GND PCIE0_RX3_P 157 158 GND
CSI3_CLK_N 27 28 CSI2_CLK_N GND 159 160 PCIE0_CLK_N
CSI3_CLK_P 29 30 CSI2_CLK_P USBSS_RX_N 161 162 PCIE0_CLK_P
GND 31 32 GND USBSS_RX_P 163 164 GND
CSI3_D1_N 33 34 CSI2_D1_N GND 165 166 USBSS_TX_N
CSI3_D1_P 35 36 CSI2_D1_P PCIE1_RX0_N 167 168 USBSS_TX_P
GND 37 38 GND PCIE1_RX0_P 169 170 GND
DP0_TXD0_N 39 40 CSI4_D2_N GND 171 172 PCIE1_TX0_N
DP0_TXD0_P 41 42 CSI4_D2_P PCIE1_CLK_N 173 174 PCIE1_TX0_P
GND 43 44 GND PCIE1_CLK_P 175 176 GND
DP0_TXD1_N 45 46 CSI4_D0_N GND 177 178 MOD_SLEEP*
DP0_TXD1_P 47 48 CSI4_D0_P PCIE_WAKE* 179 180 PCIE0_CLKREQ*
GND 49 50 GND PCIE0_RST* 181 182 PCIE1_CLKREQ*
DP0_TXD2_N 51 52 CSI4_CLK_N PCIE1_RST* 183 184 GBE_MDI0_N
DP0_TXD2_P 53 54 CSI4_CLK_P I2C0_SCL 185 186 GBE_MDI0_P
GND 55 56 GND I2C0_SDA 187 188 GBE_LED_LINK
DP0_TXD3_N 57 58 CSI4_D1_N I2C1_SCL 189 190 GBE_MDI1_N
DP0_TXD3_P 59 60 CSI4_D1_P I2C1_SDA 191 192 GBE_MDI1_P
GND 61 62 GND I2S0_DOUT 193 194 GBE_LED_ACT
DP1_TXD0_N 63 64 CSI4_D3_N I2S0_DIN 195 196 GBE_MDI2_N
DP1_TXD0_P 65 66 CSI4_D3_P I2S0_FS 197 198 GBE_MDI2_P
GND 67 68 GND I2S0_SCLK 199 200 GND
DP1_TXD1_N 69 70 DSI_D0_N GND 201 202 GBE_MDI3_N
DP1_TXD1_P 71 72 DSI_D0_P UART1_TXD 203 204 GBE_MDI3_P
GND 73 74 GND UART1_RXD 205 206 GPIO07
DP1_TXD2_N 75 76 DSI_CLK_N UART1_RTS* 207 208 GPIO08
DP1_TXD2_P 77 78 DSI_CLK_P UART1_CTS* 209 210 CLK_32K_OUT
GND 79 80 GND GPIO09 211 212 GPIO10
DP1_TXD3_N 81 82 DSI_D1_N CAM_I2C_SCL 213 214 FORCE_RECOVERY*
DP1_TXD3_P 83 84 DSI_D1_P CAM_I2C_SDA 215 216 GPIO11
GND 85 86 GND GND 217 218 GPIO12
GPIO00 87 88 DP0_HPD SDMMC_DAT0 219 220 I2S1_DOUT
SPI0_MOSI 89 90 DP0_AUX_N SDMMC_DAT1 221 222 I2S1_DIN
SPI0_SCK 91 92 DP0_AUX_P SDMMC_DAT2 223 224 I2S1_FS
SPI0_MISO 93 94 HDMI_CEC SDMMC_DAT3 225 226 I2S1_SCLK
SPI0_CS0* 95 96 DP1_HPD SDMMC_CMD 227 228 GPIO13
SPI0_CS1* 97 98 DP1_AUX_N SDMMC_CLK 229 230 GPIO14
UART0_TXD 99 100 DP1_AUX_P GND 231 232 I2C2_SCL
UART0_RXD 101 102 GND SHUTDOWN_REQ* 233 234 I2C2_SDA
UART0_RTS* 103 104 SPI1_MOSI PMIC_BBAT 235 236 UART2_TXD
UART0_CTS* 105 106 SPI1_SCK POWER_EN 237 238 UART2_RXD
GND 107 108 SPI1_MISO SYS_RESET* 239 240 SLEEP/WAKE*
USB0_D_N 109 110 SPI1_CS0* GND 241 242 GND
USB0_D_P 111 112 SPI1_CS1* GND 243 244 GND
GND 113 114 CAM0_PWDN GND 245 246 GND
USB1_D_N 115 116 CAM0_MCLK GND 247 248 GND
USB1_D_P 117 118 GPIO01 GND 249 250 GND
GND 119 120 CAM1_PWDN VDD_IN 251 252 VDD_IN
USB2_D_N 121 122 CAM1_MCLK VDD_IN 253 254 VDD_IN
USB2_D_P 123 124 GPIO02 VDD_IN 255 256 VDD_IN
GND 125 126 GPIO03 VDD_IN 257 258 VDD_IN
GPIO04 127 128 GPIO05 VDD_IN 259 260 VDD_IN
GND 129 130 GPIO06
PCIE0_RX0_N 131 132 GND

4.1 Operating and Absolute Maximum Ratings

The parameters listed in following table are specific to a temperature range and operating voltage. Operating the Jetson Xavier
NX beyond these parameters is not recommended.

WARNING: Exceeding the listed conditions may damage and/or affect long-term reliability of the part.
The Jetson Xavier NX module should never be subjected to conditions extending beyond the ratings listed
below.

Table 28: Recommended Operating Conditions

Symbol Parameter Minimum Typical Maximum Unit

VDDDC VDD_IN 4.75 5.0 5.25 V
PMIC_BBAT 1.65 - 5.5 V

Absolute maximum ratings describe stress conditions. These parameters do not set minimum and maximum operating
conditions that will be tolerated over extended periods of time. If the device is exposed to these parameters for extended
periods of time, performance is not guaranteed, and device reliability may be affected. It is not recommended to operate the
Jetson Xavier NX module under these conditions.

Table 29: Absolute Maximum Ratings

Symbol Parameter Minimum Maximum Unit Notes

VDDMAX VDD_IN -0.5 5.5 V

PMIC_BBAT -0.3 6.0 V

IDDMAX VDD_IN Imax - 5 A

VM_PIN Voltage applied to any -0.5 VDD + 0.5 V VDD + 0.5V when CARRIER_PWR_ON high
powered I/O pin and associated I/O rail powered.
I/O pins cannot be high (>0.5V) before
CARRIER_PWR_ON goes high.
When CARRIER_PWR_ON is low, the
maximum voltage applied to any I/O pin is 0.5V

DD pins configured as -0.5 3.63 V The pin’s output-driver must be set to open-
open drain drain mode

TOP Operating Temperature -25 See Note °C See the Jetson Xavier NX Thermal Design
Guide for details.

TSTG Storage Temperature -40 80 °C

(ambient)

MMAX Mounting Force - 8.2 kgf1 Maximum force applied to PCB. See the Jetson
Xavier NX Thermal Design Guide for additional
details on mounting a thermal solution.

Note: kgf stands for kilogram-force.

Table 30: Jetson NX Reliability Report

Test Conditions Reference Standard Results

Temperature Humidity 85°C / 85% RH, 168 hours, Power
JESD22-A101 PASS
Biased ON
-40°C to 105°C, 250 cycles,
Temperature Cycling JESD22-A104, IPC9701 PASS
non-operational
25°C to 65°C, 93% RH,
Temp/Humidity Cycle NV Standard PASS
six cycles
140G, half sine,
Mechanical Shock – one shock/orientation,
JESD22-B110 PASS
140G Non-Op six orientations total,
non-operational
50G, half sine,
Mechanical Shock –
Three shocks/orientation, IEC 600068 2-27 PASS
50G Op
six orientations total, operational
Connector Insertion
400-pin connector, 30 cycles EIA-364 PASS
Cycling
3G, 10-500 Hz,
Sine Vibration – 3G two sweep/axis, three axes total, IEC60068-2-6 PASS
non-operational
Random Vibration – 2G 10-500 Hz, two Grms,
IEC60068-2-64 PASS
Non-Op one hour/axis, non-operational
10-500 Hz, 1 Grms,
Random Vibration – 1G
30 min/axis, IEC60068-2-64 PASS
Op
operational
Power ON/OFF, ON for 150 sec
OFF for 30 sec
Hard Boot 2000 cycles at 25°C NV Standard PASS
1000 cycles at -5°C
1000 cycles at 45°C
Operational Low Temp -5°C, 24 hours, operational NV Standard PASS
45°C, 90%RH, 336 hours,
Operational High Temp NV Standard PASS
operational
Telcordia SR-332,
Controlled Environment (GB), T = 2,629,471
MTBF / Failure Rate ISSUE 3 Parts Count
35°C, CL = 90% hours
(Method I)
Telcordia SR-332,
Uncontrolled Environment (GF) 1,493,890
MTBF / Failure Rate ISSUE 3 Parts Count
T = 35°C, CL = 90% hours
(Method I)

4.2 Digital Logic

Voltages less than the minimum stated value can be interpreted as an undefined state or logic level low which may result in
unreliable operation. Voltages exceeding the maximum value can damage and/or adversely affect device reliability.

Table 31: CMOS Pin Type DC Characteristics

Symbol Description Minimum Maximum Units

VIL Input Low Voltage -0.5 0.25 x VDD V
VIH Input High Voltage 0.75 x VDD 0.5 + VDD V
VOL Output Low Voltage (IOL = 1mA) - 0.15 x VDD V
VOH Output High Voltage (IOH = -1mA) 0.85 x VDD - V

Table 32: Open Drain Pin Type DC Characteristics

Symbol Description Minimum Maximum Units

VIL Input Low Voltage -0.5 0.25 x VDD V
VIH Input High Voltage 0.75 x VDD 3.63 V
VOL Output Low Voltage (IOL = 1mA) - 0.15 x VDD V
I2C [1,0] Output Low Voltage (IOL = 2mA) (see note) - 0.3 x VDD V
VOH Output High Voltage (IOH = -1mA) 0.85 x VDD - V

Note: I2C[1,0]_[SCL, SDA] pins pull-up to 3.3V through on module 2.2kΩ resistor. I2C2_[SCL, SDA] pins pull-up to 1.8V through on
module 2.2kΩ resistor.

SoC HEIGHT
2.28 ± 0.15MM

Note:

• All dimensions are in millimeters unless otherwise specified.

• Tolerances are: .X ± 0.25, .XX ± is 0.1, Angle ± is 1°

The information provided in this specification is believed to be accurate and reliable as of the date provided. However, NVIDIA Corporation
(“NVIDIA”) does not give any representations or warranties, expressed or implied, as to the accuracy or completeness of such information.
NVIDIA shall have no liability for the consequences or use of such information or for any infringement of patents or other rights of third parties
that may result from its use. This publication supersedes and replaces all other specifications for the product that may have been previously
supplied.

NVIDIA reserves the right to make corrections, modifications, enhancements, improvements, and other changes to this specification, at any
time and/or to discontinue any product or service without notice. Customer should obtain the latest relevant specification before placing orders
and should verify that such information is current and complete.

NVIDIA products are sold subject to the NVIDIA standard terms and conditions of sale supplied at the time of order acknowledgement, unless
otherwise agreed in an individual sales agreement signed by authorized representatives of NVIDIA and customer. NVIDIA hereby expressly
objects to applying any customer general terms and conditions with regard to the purchase of the NVIDIA product referenced in this
specification.

NVIDIA products are not designed, authorized or warranted to be suitable for use in medical, military, aircraft, space or life support equipment,
nor in applications where failure or malfunction of the NVIDIA product can reasonably be expected to result in personal injury, death or
property or environmental damage. NVIDIA accepts no liability for inclusion and/or use of NVIDIA products in such equipment or applications
and therefore such inclusion and/or use is at customer’s own risk.

NVIDIA makes no representation or warranty that products based on these specifications will be suitable for any specified use without further
testing or modification. Testing of all parameters of each product is not necessarily performed by NVIDIA. It is customer’s sole responsibility
to ensure the product is suitable and fit for the application planned by customer and to do the necessary testing for the application in order to
avoid a default of the application or the product. Weaknesses in customer’s product designs may affect the quality and reliability of the NVIDIA
product and may result in additional or different conditions and/or requirements beyond those contained in this specification. NVIDIA does
not accept any liability related to any default, damage, costs or problem which may be based on or attributable to: (i) the use of the NVIDIA
product in any manner that is contrary to this specification, or (ii) customer product designs.

No license, either express or implied, is granted under any NVIDIA patent right, copyright, or other NVIDIA intellectual property right under
this specification. Information published by NVIDIA regarding third-party products or services does not constitute a license from NVIDIA to
use such products or services or a warranty or endorsement thereof. Use of such information may require a license from a third party under
the patents or other intellectual property rights of the third party, or a license from NVIDIA under the patents or other intellectual property
rights of NVIDIA. Reproduction of information in this specification is permissible only if reproduction is approved by NVIDIA in writing is
reproduced without alteration and is accompanied by all associated conditions, limitations, and notices.

ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER
DOCUMENTS (TOGETHER AND SEPARATELY, “MATERIALS”) ARE BEING PROVIDED “AS IS.” NVIDIA MAKES NO WARRANTIES,
EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL
IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE.
Notwithstanding any damages that customer might incur for any reason whatsoever, NVIDIA’s aggregate and cumulative liability towards
customer for the products described herein shall be limited in accordance with the NVIDIA terms and conditions of sale for the product.

HDMI

HDMI, the HDMI logo, and High-Definition Multimedia Interface are trademarks or registered trademarks of HDMI Licensing LLC.

ARM

ARM, AMBA and ARM Powered are registered trademarks of ARM Limited. Cortex, MPCore and Mali are trademarks of ARM Limited. All
other brands or product names are the property of their respective holders. ʺARMʺ is used to represent ARM Holdings plc; its operating
company ARM Limited; and the regional subsidiaries ARM Inc.; ARM KK; ARM Korea Limited.; ARM Taiwan Limited; ARM France SAS;
ARM Consulting (Shanghai) Co. Ltd.; ARM Germany GmbH; ARM Embedded Technologies Pvt. Ltd.; ARM Norway, AS and ARM Sweden
AB

OpenGL

OpenGL is a trademark of Apple Inc. used under license to the Khronos Group Inc.

Trademarks

NVIDIA, the NVIDIA logo, and Tegra are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries.
Other company and product names may be trademarks of the respective companies with which they are associated.

NVIDIA Corporation | 2788 San Tomas Expressway | Santa Clara, CA 95051 | www.nvidia.com

Advanced ASIC chip synthesis using Synopsys Design Compiler Physical Compiler and PrimeTime 2nd Edition Himanshu Bhatnagar - The ebook in PDF/DOCX format is ready for download now
No ratings yet
Advanced ASIC chip synthesis using Synopsys Design Compiler Physical Compiler and PrimeTime 2nd Edition Himanshu Bhatnagar - The ebook in PDF/DOCX format is ready for download now
57 pages
Phy Interface Pci Express Sata Usb31 Architectures PIPE - Rev6 - 2 - 1
No ratings yet
Phy Interface Pci Express Sata Usb31 Architectures PIPE - Rev6 - 2 - 1
187 pages
Basics of DDR Protocol: Jose Thomas Vellara
No ratings yet
Basics of DDR Protocol: Jose Thomas Vellara
57 pages
Arnold An eFPGA-Augmented RISC-V SoC For Low Power Iot End Nodes
No ratings yet
Arnold An eFPGA-Augmented RISC-V SoC For Low Power Iot End Nodes
14 pages
25568707
No ratings yet
25568707
78 pages
AXI Inter Connect
No ratings yet
AXI Inter Connect
32 pages
Design and Simulation of A PCI Express Based Embed
No ratings yet
Design and Simulation of A PCI Express Based Embed
7 pages
Coresight v3 0 Architecture Specification IHI0029E
No ratings yet
Coresight v3 0 Architecture Specification IHI0029E
280 pages
Axi Prot
No ratings yet
Axi Prot
273 pages
An4581 2
No ratings yet
An4581 2
20 pages
ECRC Calculation
No ratings yet
ECRC Calculation
8 pages
NVIDIA Jetson AGX Xavier Series System-on-Module: Volta GPU + Carmel CPU + 8/32GB LPDDR4x + 32GB eMMC
No ratings yet
NVIDIA Jetson AGX Xavier Series System-on-Module: Volta GPU + Carmel CPU + 8/32GB LPDDR4x + 32GB eMMC
72 pages
USB4 PCIe Tunneling CTS Rev 1.7 - CLEAN
No ratings yet
USB4 PCIe Tunneling CTS Rev 1.7 - CLEAN
53 pages
pg055 Axi Bridge Pcie PDF
No ratings yet
pg055 Axi Bridge Pcie PDF
117 pages
AHB Lite Specification
No ratings yet
AHB Lite Specification
72 pages
CSV Basic Synthesis
No ratings yet
CSV Basic Synthesis
38 pages
Mipi Csi-2 32
No ratings yet
Mipi Csi-2 32
32 pages
PCS White Paper
No ratings yet
PCS White Paper
14 pages
The Veloce Emulator: Laurent VUILLEMIN Platform Compile Software Manager Emulation Division
No ratings yet
The Veloce Emulator: Laurent VUILLEMIN Platform Compile Software Manager Emulation Division
36 pages
Efabless Caravel "Harness" Soc: Preliminary
No ratings yet
Efabless Caravel "Harness" Soc: Preliminary
30 pages
Xilinx Answer 65062 AXI PCIe Address Mapping
No ratings yet
Xilinx Answer 65062 AXI PCIe Address Mapping
11 pages
Pcie Intel Specification
No ratings yet
Pcie Intel Specification
9 pages
SCT 4 Platform Design Spec Rev1 4 PDF
No ratings yet
SCT 4 Platform Design Spec Rev1 4 PDF
62 pages
Power Optimization (Part 2) : Xuan Silvia' Zhang
No ratings yet
Power Optimization (Part 2) : Xuan Silvia' Zhang
26 pages
Interrupts
No ratings yet
Interrupts
59 pages
Banana Pi: Specification For PCIE-DVR: Root Complex Architecture Specification
No ratings yet
Banana Pi: Specification For PCIE-DVR: Root Complex Architecture Specification
32 pages
Adv Verif Topics Ch5-Part1
No ratings yet
Adv Verif Topics Ch5-Part1
7 pages
1 Introducing VCS Basic
0% (1)
1 Introducing VCS Basic
6 pages
Video Display Unit (VDU) : Analogue TV & Monitors Computer Monitors
No ratings yet
Video Display Unit (VDU) : Analogue TV & Monitors Computer Monitors
7 pages
03 Building Custom Socs
No ratings yet
03 Building Custom Socs
30 pages
Pci-Express (Peripheral Component Interconnect) : Root Complex
No ratings yet
Pci-Express (Peripheral Component Interconnect) : Root Complex
5 pages
NVIDIA Jetson Nano System-on-Module: Maxwell GPU + ARM Cortex-A57 + 4GB LPDDR4 + 16GB eMMC
No ratings yet
NVIDIA Jetson Nano System-on-Module: Maxwell GPU + ARM Cortex-A57 + 4GB LPDDR4 + 16GB eMMC
38 pages
High-Speed 8B/10B Encoder Design Using A Simplified Coding Table
100% (1)
High-Speed 8B/10B Encoder Design Using A Simplified Coding Table
5 pages
pg138 Axi Ethernet
No ratings yet
pg138 Axi Ethernet
176 pages
Mainframe - Book Details
100% (1)
Mainframe - Book Details
14 pages
Opencores Coding Guidelines
No ratings yet
Opencores Coding Guidelines
28 pages
Pcie Aer
No ratings yet
Pcie Aer
10 pages
PCI Express Basics & Background: Richard Solomon Synopsys
100% (1)
PCI Express Basics & Background: Richard Solomon Synopsys
45 pages
Ascent SW User Manual For Fluoroskan Ascent, CF & FL
100% (1)
Ascent SW User Manual For Fluoroskan Ascent, CF & FL
400 pages
Nvidia Nano Datasheet
No ratings yet
Nvidia Nano Datasheet
41 pages
HPPatch
No ratings yet
HPPatch
83 pages
006 I2C Bus Protocol
100% (2)
006 I2C Bus Protocol
16 pages
PCIe
0% (1)
PCIe
41 pages
Pulpissimo: Datasheet: The Pulp Team
No ratings yet
Pulpissimo: Datasheet: The Pulp Team
101 pages
ApplicationSpecific DRAM Architectures and Designs
No ratings yet
ApplicationSpecific DRAM Architectures and Designs
81 pages
Embedded Vision Demo
No ratings yet
Embedded Vision Demo
10 pages
2. Availability Concepts
No ratings yet
2. Availability Concepts
32 pages
Poweredge hs5620 Technical Guide
No ratings yet
Poweredge hs5620 Technical Guide
80 pages
RW6.5 997-3 EthernetIP_CIP Safe_Config_SPL_v3.0.1
No ratings yet
RW6.5 997-3 EthernetIP_CIP Safe_Config_SPL_v3.0.1
18 pages
S.2 COMPUTER STUDIES Revision Questions by Tulina Dorothy
100% (1)
S.2 COMPUTER STUDIES Revision Questions by Tulina Dorothy
5 pages
Hadoop & Spark
No ratings yet
Hadoop & Spark
40 pages
29 802.1q
No ratings yet
29 802.1q
10 pages
The PCI Bus
No ratings yet
The PCI Bus
25 pages
Palestra 4 Abram Belk
No ratings yet
Palestra 4 Abram Belk
143 pages
Signal Intagrity Simulation of PCIE PDF
No ratings yet
Signal Intagrity Simulation of PCIE PDF
5 pages
HCNA-HNTD Intermediate Lab Guide V2.2
No ratings yet
HCNA-HNTD Intermediate Lab Guide V2.2
175 pages
TM105 Meeting1 Slides
No ratings yet
TM105 Meeting1 Slides
49 pages
SR MR Iov
No ratings yet
SR MR Iov
63 pages
00 Initial Steps For Setting Up Actix For GNG B
No ratings yet
00 Initial Steps For Setting Up Actix For GNG B
62 pages
Synopsys DW
No ratings yet
Synopsys DW
11 pages
Moxa Iologik 4000 Series M Modules Manual v4.1
No ratings yet
Moxa Iologik 4000 Series M Modules Manual v4.1
52 pages
Cortex Mo+ Technical Reference Manual
No ratings yet
Cortex Mo+ Technical Reference Manual
51 pages
AXI To AHB-Lite Bridge Cycle Model
No ratings yet
AXI To AHB-Lite Bridge Cycle Model
22 pages
File: /home/binod/documents/allfmca p/rocket-chip-master/README - MD Page 1 of 7
No ratings yet
File: /home/binod/documents/allfmca p/rocket-chip-master/README - MD Page 1 of 7
7 pages
Data Deduplication
No ratings yet
Data Deduplication
62 pages
PCI SIG Arch Overview
No ratings yet
PCI SIG Arch Overview
37 pages
Embedded Systems Design - 2: Dr. N. Mathivanan
No ratings yet
Embedded Systems Design - 2: Dr. N. Mathivanan
10 pages
Oracle Cloud Infrastructure 2023 Sales Specialist Assessmen
No ratings yet
Oracle Cloud Infrastructure 2023 Sales Specialist Assessmen
5 pages
IT_Partners
No ratings yet
IT_Partners
1 page
Installation: Mevinet-Q2
No ratings yet
Installation: Mevinet-Q2
16 pages
Intelligent High Performance Memory Access Technique in Aspect of DDR3
No ratings yet
Intelligent High Performance Memory Access Technique in Aspect of DDR3
6 pages
5468.amba Axi Ahb Apb
No ratings yet
5468.amba Axi Ahb Apb
2 pages
Cyclone V Avalon-Mm Interface For Pcie Solutions: User Guide
No ratings yet
Cyclone V Avalon-Mm Interface For Pcie Solutions: User Guide
173 pages
AXI4-Stream Upsizing Downsizing Data Width Converters For Hardware-In-The-Loop Simulations
No ratings yet
AXI4-Stream Upsizing Downsizing Data Width Converters For Hardware-In-The-Loop Simulations
3 pages
Using MsiZap
No ratings yet
Using MsiZap
3 pages
Problem Code Es
No ratings yet
Problem Code Es
18 pages
ECDIS V3.xx Upgrade Guide
No ratings yet
ECDIS V3.xx Upgrade Guide
4 pages
Whitepaper-FAQ-Comparing Silver Peak and Riverbed
No ratings yet
Whitepaper-FAQ-Comparing Silver Peak and Riverbed
5 pages
Phy Ip For Pcie 3.0
No ratings yet
Phy Ip For Pcie 3.0
2 pages
Running Opengl On Visual Studio 2005
No ratings yet
Running Opengl On Visual Studio 2005
23 pages
Vce Private Cloud: 1. Install Winscp and Start It
No ratings yet
Vce Private Cloud: 1. Install Winscp and Start It
13 pages
ATM Quality of Service Configuration Guide: Thomson Gateway
No ratings yet
ATM Quality of Service Configuration Guide: Thomson Gateway
44 pages
Multi-Tenancy With Vcloud Director and NSX-T
No ratings yet
Multi-Tenancy With Vcloud Director and NSX-T
10 pages
Block Storage Provisioning and Management
No ratings yet
Block Storage Provisioning and Management
29 pages
18cs442 MC-2
No ratings yet
18cs442 MC-2
2 pages
Ferret Installation Example
No ratings yet
Ferret Installation Example
9 pages
Instruction Op-Code Operand Bytes Machine - Cycles T - States Detail
No ratings yet
Instruction Op-Code Operand Bytes Machine - Cycles T - States Detail
3 pages
Convenient Plug-In Software For Using Laser Scans in Autocad
No ratings yet
Convenient Plug-In Software For Using Laser Scans in Autocad
2 pages
Bare-Metal Embedded C Programming: Develop high-performance embedded systems with C for Arm microcontrollers
From Everand
Bare-Metal Embedded C Programming: Develop high-performance embedded systems with C for Arm microcontrollers
Israel Gbati
No ratings yet
Application-Specific Integrated Circuit ASIC A Complete Guide
From Everand
Application-Specific Integrated Circuit ASIC A Complete Guide
Gerardus Blokdyk
No ratings yet