| OS Supports |
Windows Server,
Linux Server (Ubuntu & Red Hat)
|
Mainly for Windows,
Some Linux
|
Ubuntu via NVIDIA JetPack SDK |
Software
Compatibility
|
Broad compatibility with a wide range of open source & commercial
support software package
|
CUDA, TensorRT, DLA, VPI...
NVIDIA Ecosystem
|
| Computing Capability |
Server-grade CPUs with optional discrete GPU, higher performance |
Desktop CPU with optional discrete GPU, high performance |
CPU & GPU integrated in a single SoC, generally lower performance |
| Al Performance |
700 ~ 1,200+ TOPS |
300 ~ 700 TOPS |
20 ~ 300 TOPS |
| GPU Expandability (Power Supply) |
2 Single Deck ~ 10 Double Deck
(500W ~ 2700W)
|
1 Single Deck ~ 2 Double Deck
(80W ~ 850W)
|
GPU embedded |
| Power Consumption |
Large power consumption, require robust cooling solutions |
Higher power consumption, require more cooling solutions |
Lower power usage, energy efficiency satisfied |
| Form Factor |
Mainly Rackmount (1U to 4U) |
Variety in Fan-base & Fanless Box, Wallmount, Rackmount, Tower |
Fanless Embedded Box |
| Size & Mobility |
Large & stationary |
Compact to large, less portable |
Ultra-compact & lightweight, highly portable |
| Applications |
AI AOI, wafer defect detection, web inspection & line scan Al inspection, large-scale inference |
High-speed & high-resolution area scan, Al inspection, robot vision, factory automation |
AMR, high integrated & mobile equipment, Environmental, Health & Safety (EHS) |