Embedded with NVIDIA GPU architecture

AI improves AOI efficiency and greatly reduce manual efforts

The Leadtek AI Inference Device are the compact inference endpoints embedded with the latest NVIDIA Ampere / Turing / Pascal GPU architecture. When applied in an AOI (Automated Optical Inspection) field, the AI Inference Device works as the second screening to differentiate between real defects and false-positives, greatly improving screening accuracy and reducing manual efforts that waste both time and money.

GPU integration
GPU integration
Small
Small
Low energy consumption
Low energy
consumption
Valuable
Valuable
Management
Management
Features :
  • Bring full GPU capacity and optimized NVIDIA CUDA cores to implement AI trained models in its full speed for object detection, AOI screening, content recognition, multi-target tracking, and other AI inference applications
  • Compact device size can even be mounted within the small space constraint of equipment containers
  • Dual Ethernet ports dedicated for communications between the AI inference device and image/video source as well as the AI inference device and AIDMS management software console
  • Power optimization allows the AI Inference Device to be powered the full GPU capacity in lowest energy consumption
  • Features specifically reserved for extended react/control equipment and activities with dual RS-232/422/485, four USB peripherals (USB 3.2/USB 2.0), audio, and additional microSD storage
  • Managed by AIDMS software to update the latest AI trained models and collect inspection results/reports

AI Inference Efficiency Accelerates

Provide 7-10 times faster AI model inference performance Model (Tiny-) YOLO

— More AI Inference Applications —

Automatic Production

Automatic Production

Environmental Monitoring

Environmental Monitoring

Safety Monitoring

Safety Monitoring

Leadtek helps you achieve successful deep learning projects

— Specification Table —

CPU AMD RX-216TD Processor (1.6-3.0 GHz dual-core)
GPU
  • Option 1 : NVIDIA GeForce GTX 1650 (GDDR5) with 896 CUDA cores
  • Option 2 : NVIDIA GeForce GTX 1650 (GDDR6) with 896 CUDA cores
  • Option 3 : NVIDIA Quadro P2200 with 1280 CUDA cores
System memory type Default 2pcs 4GByte of DDR4 SO-DIMM SDRAM memory (up to 1,600 MT/s transfer rate)
Storage 32GByte of SSD device for operating system, software, AI model, and data storage by default
Device dimensions 195.4 mm (width) x 212.6 mm (depth) x 110.0 mm (height)
Device power consumption 35 watts with basic operating peripherals and monitor attached
Power consumption varies depending on the AI application workload
Connectors

(Internal) Graphics Interface: PCI Express 3.0 x8
Display Connectors: depending upon your graphics card's configuration

  • One microSD slot for additional storage
  • Two Type-A USB 3.2 Gen connectors, Two Type-A USB 2.0 connectors
  • Two RS-232, RS-422, or RS-485 connectors
  • One 3.5mm Line-out jack
  • One 3.5mm Microphone jack
  • Two 10/100/1000Base-T copper interfaces with Link and Activity LEDs
  • One 19V, 1x2-pin terminal block power jack
LED Power LED, System Status LED
Button Power button, Reset button
Thermal cooling solution One active fan for device system
Two active fan sink for CPU and graphics card
Operating / Storage Environment
  • Humidity
    Relative (non-condensing) : 10% to 90%
    Storage : 5% to 95%
  • Temperature
    Operational : 0 to 40°C
    Storage : -20 to 70°C
CPU

Intel 8th and 9th generation processors:

  • Option 1 : Intel Core i5-9500E processor CPU (3.0-4.2 GHz, 6 cores)
  • Option 2 : Intel Core i7-8700 Processor CPU (3.2-4.6 GHz, 6 cores)
PCH Intel Q370 Chipset
GPU

NVIDIA Ampere architecture GPU-accelerated options :

  • Up to two GeForce RTX 3060 CLASSIC (CUDA cores 3,584 each)
  • Up to two GeForce RTX 3060 Ti CLASSIC (CUDA cores 4,864 each)
  • One GeForce RTX 3070 AI BLOWER (CUDA cores 5,888)
  • Up to two NVIDIA RTX A4000 (CUDA cores 6,144 each)
  • One NVIDIA RTX A5000 (CUDA cores 8,192)
  • One NVIDIA RTX A6000 (CUDA cores 10,752)
System memory type Default 2pcs 4GByte of DDR4 SO-DIMM SDRAM memory (up to 2,666 MT/s transfer rate)
Storage Default 32GByte of PCIe x4 M.2 2242/2280 SSD device for operating system, software, AI model, and data storage by default
Device dimensions 220.5 (width) x 311.0 (depth) x 181.0 (height) mm
Device power consumption Varies depending on the AI application workload and CPU/GPU configurations
Connectors

(Internal) Two PCI Express x8 interface for graphics
Display connectors vary depending upon your graphics card's configuration

  • One internal PCIe x4 SSD slot
  • Four Type-A USB 3.2 Gen 2 connectors
  • Two RS-232, RS-422, or RS-485 connectors
  • One 3.5mm Line-out jack
  • One 3.5mm Microphone jack
  • Two 10/100/1000Base-T copper interfaces with Link and Activity LEDs
  • One MicroSD slot for additional storage
  • One power cord connector
LED Power LED, System Status LED
Button Power button, Reset button
Thermal cooling solution Two Active fan for device system and CPU
One Active fansink for graphics card(s)
Operating / Storage Environment
  • Humidity
    Relative (non-condensing) : 10% to 90%
    Storage : 5% to 95%
  • Temperature
    Operational : 0 to 45°C
    Storage : -20 to 70°C