The Jetson TX2 family has four production modules: the Jetson TX2, the Jetson TX2 4GB, the Jetson TX2 NX and the Jetson TX2i. There's also a Jetson TX2 Developer Kit. All of them use the same Pascal GPU with 256 CUDA cores and a 6-core CPU.
Jetson TX2 family production modules developer... +3developer.nvidia.com ↗“Pascal SoM; 1.33 TFLOPS FP16, 256 CUDA cores, dual-Denver + 4-core Cortex-A57, 8GB LPDDR4, 7.5-15W.”developer.nvidia.com ↗“Pascal SoM; 256 CUDA cores, 6-core CPU, 4GB LPDDR4, 7.5-15W; cost-optimized TX2.”developer.nvidia.com ↗“Reduced-form-factor Pascal SoM (Nano/Xavier NX pin-compatible)”developer.nvidia.com ↗“Industrial/rugged Pascal SoM; 256 CUDA cores, 6-core CPU, 8GB LPDDR4 ECC, extended temp/vibration”
| Module | Memory | Power | Form factor | Lifecycle status |
|---|---|---|---|---|
| Jetson TX2 | 8 GB LPDDR4 | 7.5–15 W | 400-pin B2B (50x87mm) | EOL |
| Jetson TX2 4GB | 4 GB LPDDR4 | 7.5–15 W | 400-pin B2B (50x87mm) | EOL |
| Jetson TX2 NX | 4 GB LPDDR4 | 7.5–15 W | 260-pin SO-DIMM (69.6x45mm) | Not listed in the catalog |
| Jetson TX2i | 8 GB LPDDR4 ECC | 10–20 W | 400-pin B2B (50x87mm) | Not listed in the catalog |
What this means for planning replacements:
- Two connector types: The TX2, TX2 4GB and TX2i all use the 400-pin B2B connector. The TX2 NX uses a 260-pin SO-DIMM instead. It's described as pin-compatible with Jetson Nano and Xavier NX, so TX2 NX carriers are likely to offer the easiest upgrade path. developer... +1developer.nvidia.com ↗“mechanical.form_factor=260-pin SO-DIMM (69.6x45mm)”developer.nvidia.com ↗“Nano/Xavier NX pin-compatible”
- Industrial variant: The TX2i is the rugged version. It adds ECC memory and extended temperature and vibration tolerance, so it's worth mapping those units separately if your fleet includes them. developer...developer.nvidia.com ↗“8GB LPDDR4 ECC, extended temp/vibration, ~10-20W”
- Performance baseline: FP32 is 0.666 TFLOPS for the TX2, TX2 4GB and TX2 NX, and 0.63 TFLOPS for the TX2i. That gives you a floor when sizing replacements. developer... +1developer.nvidia.com ↗“performance.max_fp32_tflops=0.666 TFLOPS”developer.nvidia.com ↗“performance.max_fp32_tflops=0.63 TFLOPS”
- Developer kit: The Jetson TX2 Developer Kit is an 8 GB TX2 on a full I/O carrier board. It's also marked EOL. developer... +1developer.nvidia.com ↗“Legacy dev kit on Jetson TX2 (Pascal, 256 CUDA, 8GB)”developer.nvidia.com ↗“identity.lifecycle_status=EOL”
Once you tell me which of these modules your fleet uses and what you need to keep (form factor, memory, power budget or operating temperature), I can list the current modules and partner systems that meet those needs.