Are there any performance differences when running the Jetson AGX Orin as either a PCIe Root Complex or Endpoint to an external FPGA?
*** Please note that this reply is generated by LLM automatically ***
The provided context does not directly address the question of performance differences when running the Jetson AGX Orin as either a PCIe Root Complex or Endpoint to an external FPGA. However, I can provide some general insights and comparisons based on the available information.
The Jetson AGX Orin series modules are designed to provide high-performance computing capabilities, and their performance can be affected by various factors, including the configuration and usage of the PCIe interface.
When running the Jetson AGX Orin as a PCIe Root Complex, it acts as the master device, controlling the PCIe bus and managing the communication with other devices connected to the bus. In this configuration, the Jetson AGX Orin can utilize its full processing capabilities, including the GPU and CPU, to perform tasks and communicate with other devices.
On the other hand, when running the Jetson AGX Orin as a PCIe Endpoint, it acts as a slave device, receiving commands and data from the PCIe Root Complex (in this case, an external FPGA). In this configuration, the Jetson AGX Orin’s processing capabilities may be limited by the bandwidth and latency of the PCIe interface, as well as the capabilities of the external FPGA.
In general, running the Jetson AGX Orin as a PCIe Root Complex may provide better performance for tasks that require high processing power and low latency, such as machine learning, computer vision, and scientific simulations. However, running the Jetson AGX Orin as a PCIe Endpoint may be more suitable for tasks that require high-bandwidth data transfer and low-power consumption, such as data acquisition, signal processing, and IoT applications.
It’s worth noting that the performance differences between running the Jetson AGX Orin as a PCIe Root Complex or Endpoint will depend on the specific use case, system configuration, and application requirements. Therefore, it’s recommended to evaluate the performance of the Jetson AGX Orin in both configurations to determine the best approach for a particular application.
In terms of thermal performance, the provided context suggests that the Jetson AGX Orin series modules have a thermal design power (TDP) budget of 40W, and an over-current (OC) limit is set to prevent the module from exceeding this budget. However, it’s unclear how the thermal performance of the Jetson AGX Orin would be affected by running it as a PCIe Root Complex or Endpoint.
In summary, while the provided context does not provide a direct answer to the question, it suggests that running the Jetson AGX Orin as a PCIe Root Complex may provide better performance for tasks that require high processing power and low latency, while running it as a PCIe Endpoint may be more suitable for tasks that require high-bandwidth data transfer and low-power consumption. However, the actual performance differences will depend on the specific use case, system configuration, and application requirements.
*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***
Hi everyone,
I’m working on a PCIe integration where the NVIDIA Jetson AGX Orin is configured as the Root Complex and an FPGA board is acting as the Endpoint. The PCIe link trains successfully and FPGA ↔ FPGA communication shows full x8 performance. However, when measuring throughput, the Jetson → FPGA direction consistently operates at approximately half the data rate compared to FPGA → Jetson.
Has anyone experienced asymmetric throughput on Jetson PCIe interfaces, or have suggestions on tuning, driver configuration, or root-port parameters that may impact performance? Any pointers or best practices would be greatly appreciated.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.