German-based Imago Technologies recently unveiled an ultra-high performance General-Purpose GPU vision system designed for intensive applications such as AI and Computer Vision. The fanless Vision Box AI offers access to 12 Arm CPU cores, 64GB of RAM and up to 275 TOPS.

The Vision Box AI integrates the latest Jetson AGX Orin 64GB model from NVIDIA which offers the following features:

• CPU —  12-core Arm Cortex-A78AE v8.2 64-bit CPU (up to 2.2GHz), 3MB L2 + 6MB L3, 
• GPU — Ampere architecture w/ 2048 CUDA cores and 64 Tensor Cores, 275 TOPS (INT8) 
• DL Accelerator — 2x NVDLA v2.0 (up to 2.2GHz)
• Memory — 64GB 256-bit LPDDR5 (204.8 GB/s) 

 
Jetson AGX Orin block diagram (left) and specs table (right)
(click images to enlarge)

For additional storage, the Vision Box AI offers an SD card slot and a M.2 slot with M-key. This device also offers an optional DisplayPort 1.4a with 8K resolution. 

Vision Box preliminary block diagram
(click image to enlarge)

The camera interface combines two Camera Link ports in addition to four 1-GiGE ports for standard GigE cameras. Other I/O ports found on the Vision Box AI include two RS-422, two USB 2.0 ports and an isolated I/O connector. 

The preliminary block diagram also shows an FPGA which interfaces with the AGX Orin module and some of the peripherals, although the FPGA model it’s unspecified. 

• Storage:
  • 1x M.2 with M-key 
  • 1x Micro SD card slot
• Display/Audio:
  • 1x HDMI 1.4a (8K60)
• Camera:
  • 2x Camera Link
• Connectivity:
  • 4x GbE LAN RJ45 port
• USB:
  • 2x USB 2.0
  • 1x USB 3.2 (Type-C)
• I/O:
  • 2x RS-422 (3-In/3-Out) 
  • 1x Isolated I/O port (8-In/8-Out)
• Power:
  • 24V DC in 

 Further information

The company plans to showcase the Vision Box AI and release additional documentation at the Vision trade fair which will happen from October 4-6 in Stuttgart, Germany. Future updates can be found on the Vision Box AI product page located here.