We brink low energy AI
GUSTAV QUADRO™
Build for private or public data centers
GUSTAV QUADRO was developed as a logical consequence of the industry. The demand for AI computing power is growing massively every day.
Companies around the world are concerned with how to operate their generated AI models in an energy-efficient manner. In most cases, the operation is limited to 1-2 optimized and trained models for the respective application.
It is precisely this market segment that we are closing with GUSTAV QUADRO.
Performance features
Focused on Low Energy
Every component in GUSTAV QUADRO is optimized for low power consumption.
Up to 4 NVIDIA Jetson Orin NX modules can be run together and 2 boards can be installed in one rack.
The 4 slaves can be controlled by the master. An internal 1000 Mbit switch establishes the communication with just one network interface.
This reduces cable management and again saves a lot of space in the rack. Only 1 switch ethernet port is needed. Each module can be dynamically regulated between 10 or 25 watts. Even a complete shutdown and reboot of the 4 slaves is possible.
Hardware need high end software
Features
From industrial to solution
With GUSTAV QUADRO, we also release the new RB inference engine. This software engine provides high performance visual object recognition, segmentation, tracking and meta data event transfer.
RB-Inference is an engine that comes without frameworks like Pytorch.
GUSTAV QUADRO™
- Up to 2 x GUSTAV QUADRO BOARDS
- Each board host 4 x NVIDIA® JETSON ORIN NX 8 or 16 GB
- 1 x USB 2.0 for Control the device (keyboard + mouse)
- 1 x HDMI 2.0 for KVM monitoring
- 4 x M.2 SSD Slots (One board host 4 x 4 Terabytes SSD maximum
RB-Inference software (ships with GUSTAV QUADRO)
- Supports FP-32,FP-16, Int(8)
- Loads directly trt models
- Supports REBOTNIX *.rvt & *.rvts models
- CSPNet strategy to partition the feature map
- Ultrafast single Stage Object detection
- Retro fit for IP camera anylsing
- Supports NVIDIA Jetpack 6.1 or higher
- Optimized openCV with cuda
- Supports ARM64 Bit Architecture
