# Offloading ZED Depth Computation to Jetson DLA / PVA on AGX Orin

**URL:** <https://community.stereolabs.com/t/offloading-zed-depth-computation-to-jetson-dla-pva-on-agx-orin/11777>\
**Category:** Software\
**Tags:** sdk, jetson, depth\
**Created:** [October 8, 2026, 1:02pm UTC](https://community.stereolabs.com/t/offloading-zed-depth-computation-to-jetson-dla-pva-on-agx-orin/11777 "2026-10-08T13:02:58Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Chaejun-Lim](https://community.stereolabs.com/user_avatar/community.stereolabs.com/chaejun-lim/32/6532_2.png) [@Chaejun-Lim](https://community.stereolabs.com/u/Chaejun-Lim)\
**Post date:** [October 8, 2026, 1:02pm UTC](https://community.stereolabs.com/t/offloading-zed-depth-computation-to-jetson-dla-pva-on-agx-orin/11777/1 "2026-10-08T13:02:58Z")

</div>

Hi everyone,

I am currently developing a Vision-Language-Action (VLA) pipeline deployed on an **NVIDIA Jetson AGX Orin**.

Because the downstream policy and vision models demand substantial GPU compute and VRAM, I am looking to maximize the AGX Orin’s dedicated hardware accelerators—specifically the **DLA (Deep Learning Accelerator)** and **PVA (Programmable Vision Accelerator)**—to relieve the GPU load.

I would like to ask:

1. Does the ZED SDK currently support offloading depth estimation (e.g., NEURAL or ULTRA mode) to the DLA or PVA engines?

2. If not natively exposed via `InitParameters`, is there any planned roadmap, experimental flag, or workaround to achieve this?

3. Has anyone in the community successfully routed ZED stereo matching or depth processing to DLA/PVA?

Any insights, benchmarks, or recommended architectural practices would be greatly appreciated.

**Environment:**

- **Platform:** Jetson AGX Orin (64GB)

- **Camera:** ZED X

- **Capture card:** ZED Link (Duo)

- **JetPack:** 7.2

- **ZED SDK:** ZED SDK for JetPack 7.2 (L4T 39.2) 5.5

Thanks in advance!

---

<div class="post-metadata">

**Author:** ![Myzhar](https://community.stereolabs.com/user_avatar/community.stereolabs.com/myzhar/32/11_2.png) [@Myzhar](https://community.stereolabs.com/u/Myzhar)\
**Post date:** [October 8, 2026, 1:40pm UTC](https://community.stereolabs.com/t/offloading-zed-depth-computation-to-jetson-dla-pva-on-agx-orin/11777/2 "2026-10-08T13:40:11Z")

</div>

Hi @Chaejun-Lim,  
Welcome to the [StereoLabs](http://www.stereolabs.com) community.

> [@Chaejun-Lim](#):
>
> 1. Does the ZED SDK currently support offloading depth estimation (e.g., NEURAL or ULTRA mode) to the DLA or PVA engines?

No, the ZED SDK depth processing runs on the GPU only; DLA and PVA are not supported.  
Please note that ULTRA is an obsolete mode based on the old stereo matching pipeline, kept only for backward compatibility with ZED SDK v4.x applications; for new applications, use one of the AI depth modes: `NEURAL_LIGHT`, `NEURAL`, or `NEURAL_PLUS`.

> [@Chaejun-Lim](#):
>
> 1. If not natively exposed via `InitParameters`, is there any planned roadmap, experimental flag, or workaround to achieve this?

There are no experimental flags or workarounds to route the ZED SDK depth processing to DLA or PVA, and I cannot share roadmap details on this topic.

> [@Chaejun-Lim](#):
>
> 1. Has anyone in the community successfully routed ZED stereo matching or depth processing to DLA/PVA?

This is not possible, because the depth engine is internal to the ZED SDK and is not exposed for external deployment.

> [@Chaejun-Lim](#):
>
> Any insights, benchmarks, or recommended architectural practices would be greatly appreciated.

To reduce the GPU load of the depth processing, you can:

- set `InitParameters::depth_precision = DEPTH_PRECISION::INT8` with the `NEURAL` depth mode; the neural network inference runs at lower precision, reducing GPU memory usage and increasing speed with a small accuracy cost, typically 1-2% ([Depth Precision documentation](https://docs.stereolabs.com/docs/development/zed-sdk/modules/depth-sensing/depth-settings#depth-precision))
- use `DEPTH_MODE::NEURAL_LIGHT`, the fastest AI depth mode, if its ideal range (0.3 to 5 m) fits your application; you can compare the computational performance of each mode on embedded devices in the [Depth Modes documentation](https://docs.stereolabs.com/docs/development/zed-sdk/modules/depth-sensing/depth-modes)
- set `InitParameters::depth_stabilization = 0` if you do not need temporal filtering, so the positional tracking module is not activated in the background ([Depth Stabilization documentation](https://docs.stereolabs.com/docs/development/zed-sdk/modules/depth-sensing/depth-settings#depth-stabilization))
- lower the camera resolution and the grab frame rate to the values strictly required by your policy
- set `RuntimeParameters::enable_depth = false` for the frames where depth is not needed

On the other side, you can deploy your own downstream vision models on the AGX Orin DLA engines with TensorRT, in INT8 or FP16 precision, leaving more GPU resources available to the ZED SDK and to the models that cannot run on DLA.
