Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Section
Column
width50%
Panel
borderColor#CCCCCC
bgColor#FFFFFF
titleBGColor#F0F0F0
borderStylesolid
titleOn the page:

 

Table of Contents

Column

Tip

Hardware requirements

Processors used for detector operation

Hardware requirements for neural analytics operation

If detection tools use GPU for video decoding (see These requirements apply if you select hardware video decoding in the settings of detectors or client display (see Processors used for detector operation), the following hardware requirements must be met:Configuring hardware video decoding for display on the Client).

Note
titleAttention!
  • Video decoding (NVDEC) is used both to power detectors (during hardware decoding of the input stream) and to render the GUI in the client if you enable the Enable hardware decoding parameter (see Hardware-based decoding with NVIDIA NVDEC chips). Decoding requirements are the same for both cases. Video decoding works on GPUs up to and including Kepler/Maxwell/Pascal in all versions of Detector Pack 3.14–3.15 and later.
  • Neural analytics is the execution of neural network algorithms. GPU requirements for inference can vary (see Hardware requirements for neural analytics operation). Starting with Detector Pack 3.15.2, support for Pascal (10 series) and later GPUs has been discontinued.
  • Decoding and neural analytics can run on different GPUs in the same system.

For NVIDIA GPUs (decoding)

ParameterRequirement
Comment
CUDA version
  • 11.1–11.8 (for Detector Pack 3.14–3.15.1)
  • 12.8.1 (for Detector Pack 3.15.2 and later)
  • Updating CUDA does not require replacing the GPU.
  • nvidia-smi displays the CUDA version supported by the current driver. For decoding to work, the driver must support the specified version (or a newer version within the major compatibility range)
Codec supportAvailability of the required codec in NVDEC

Current support matrix: NVIDIA Video Encode and Decode GPU Support Matrix

Compute Capability3.5–12.0 (Kepler
For Nvidia GPUs
  • CUDA 11.1−11.8
  • Support for the required codec in NVDEC (see here)
  • Compute Capability 3.5−8.9

Note
titleAttention!
CUDA is compatible with generations of computers with the following architectures: Kepler (partially)
, Maxwell, Pascal, Volta, Turing, Ampere
(partially)
, Ada Lovelace, Hopper, and Blackwell (see CUDA))
  • Decoding is preserved on all listed architectures in all versions of Detector Pack.
  • Kepler-based GPUs (Compute Capability 3.x) are partially supported. We recommend using a GPU no older than Maxwell (5.x).
  • You can check the
GPU
  • version of Compute Capability
version
  • of your GPU on
the 
  • the manufacturer'
s website.
  • When using Nvidia GPUs, we recommend installing the latest driver from the official website.
  • Number of GPUsUp to four GPUs on one serverIDs:
    To decode video in Axxon One, you can use up to four Nvidia GPUs (see Selecting Nvidia GPU when configuring detectors) installed on the Server (
    Nvidia GPU 0,
    Nvidia GPU
    1,
    Nvidia GPU
    2, 3 (see Selecting Nvidia GPU
    3).
    when configuring detectors
    DriverWe recommend the latest stable version from the NVIDIA official website

    Outdated drivers may not support the required codec

    For Intel integrated GPUs (decoding)

    ParameterRequirement
    Processor generation6th generation Intel Core (Skylake) and higher
    Codec supportAvailability of
    For the integrated Intel GPUs
  • 6th generation CPU or higher

  • Support for
    the required codec in Intel Quick Sync Video