Both are Connect Tech rackmount inference servers. The UAGX2U offers more compute: 275 INT8 TOPS and 5.32 FP32 TFLOPS, against 100 INT8 TOPS and 1.88 FP32 TFLOPS on the UNGX2U. Each has 1 Ethernet port.
UAGX2U vs. UNGX2U connecttec... +1connecttech.com ↗“Connect Tech Jetson inference server (AGX Orin); rackmount edge AI.”connecttech.com ↗“Connect Tech Jetson inference server (Orin NX); rackmount edge AI.”
| Spec | UAGX2U - Inference Server | UNGX2U - Inference Server |
|---|---|---|
| Jetson module | Jetson AGX Orin 64GB | Jetson Orin NX 16GB |
| INT8 TOPS | 275 | 100 |
| FP32 TFLOPS | 5.32 | 1.88 |
| Ethernet ports | 1 | 1 |
| Fastest Ethernet speed | 10GbE | 10GbE |
What the figures mean:
- INT8 TOPS: Both are Orin-generation products, so these are INT8 sparse figures. That makes them directly comparable. For the UNGX2U, 100 TOPS is the standard-mode figure. No Super-mode figure is listed for it.
- FP32 TFLOPS: I couldn't confirm whether Connect Tech publishes these numbers itself. They match the figures for the Jetson module inside each server, so treat them as based on the module.
- Ethernet: Each server has one external wired Ethernet port, and its fastest link is 10GbE. SFP cages aren't included in that count. If you need more detail on network interfaces, the datasheets have it.