NCA Edge AI & Embedded Systems 2 — Questions and Answers
Question 1: Which NVIDIA Jetson module is specifically designed for entry-level edge AI applications with the lowest power envelope in the Jetson lineup?
- Jetson AGX Orin
- Jetson Orin NX
- Jetson Nano (Correct answer)
- Jetson TX2
Correct answer: Jetson Nano
Jetson Nano targets entry-level edge AI with a ~5–10W power envelope, making it ideal for cost-sensitive embedded deployments.
Question 2: What is the primary role of NVIDIA DeepStream SDK in an edge AI pipeline?
- Training deep learning models on edge devices
- Streaming optimized inference on multi-sensor video inputs (Correct answer)
- Compiling CUDA kernels for embedded GPUs
- Managing OTA firmware updates for Jetson devices
Correct answer: Streaming optimized inference on multi-sensor video inputs
DeepStream is a streaming analytics toolkit that enables real-time, multi-stream video inference on NVIDIA edge hardware.
Question 3: When deploying a TensorRT engine on a Jetson device, why must the engine be rebuilt on the target device rather than transferred from a desktop GPU?
- TensorRT engines are encrypted per GPU serial number
- Engine files are architecture-specific and not portable across different GPU architectures (Correct answer)
- Jetson devices use a different version of CUDA than desktop GPUs
- TensorRT does not support cross-compilation for ARM targets
Correct answer: Engine files are architecture-specific and not portable across different GPU architectures
TensorRT engines are optimized for the specific GPU architecture and are not portable; an engine built on an Ampere desktop GPU will not run on a Jetson Volta or Ampere variant without rebuilding.
Question 4: Which power mode on Jetson devices allows users to cap the TDP and limit active CPU/GPU cores to reduce thermal output?
- MAXN mode
- nvpmodel modes (Correct answer)
- Jetson Clocks mode
- CUDA MPS mode
Correct answer: nvpmodel modes
nvpmodel lets you select predefined power modes that constrain TDP, CPU frequency, and active core counts on Jetson modules.
Question 5: In an edge AI camera pipeline, what does the term 'sensor fusion' refer to?
- Merging firmware from multiple camera vendors
- Combining data from multiple sensor modalities (e.g., RGB, lidar, radar) to produce richer scene understanding (Correct answer)
- Fusing GPU and CPU processing on a single SoC
- Updating camera ISP parameters via AI inference
Correct answer: Combining data from multiple sensor modalities (e.g., RGB, lidar, radar) to produce richer scene understanding
Sensor fusion integrates outputs from heterogeneous sensors to improve perception accuracy beyond what any single sensor can provide.
Question 6: What is the function of the NVIDIA Jetson Linux Board Support Package (BSP)?
- Providing pre-trained AI models for Jetson applications
- Supplying the OS kernel, drivers, and bootloader needed to bring up Jetson hardware (Correct answer)
- Optimizing TensorRT plans for Jetson GPU architectures
- Managing Docker containers on Jetson devices
Correct answer: Supplying the OS kernel, drivers, and bootloader needed to bring up Jetson hardware
The BSP includes the Linux kernel, device tree, NVIDIA drivers, and bootloader required to initialize and run Jetson hardware.
Question 7: Which communication protocol is most commonly used to connect edge inference nodes to a central cloud platform for telemetry and model updates?
- CAN bus
- MQTT (Correct answer)
- PCIe
- NVLink
Correct answer: MQTT
MQTT is a lightweight publish-subscribe protocol widely used for IoT and edge-to-cloud telemetry due to its low bandwidth overhead.
NCA Edge AI & Embedded Systems 2
Which NVIDIA Jetson module is specifically designed for entry-level edge AI applications with the lowest power envelope in the Jetson lineup?