Both fit a 260-pin SO-DIMM slot: how do the Jetson Xavier NX 8GB and Jetson TX2 NX compare on GPU architecture, CUDA cores, Tensor Cores and NVDLA engines?

Question

Rapidflare · Accepted Answer

Here's the side-by-side comparison of the two 260-pin SO-DIMM modules on the compute attributes you asked about:

| Attribute | Jetson Xavier NX 8GB | Jetson TX2 NX |
|---|---|---|
| GPU architecture | Volta | Pascal |
| CUDA cores | 384 | 256 |
| Tensor Cores | 48 | 0 |
| NVDLA engines | 2 | 0 |

Jetson Xavier NX 8GB uses a Volta GPU with 384 CUDA cores, 48 Tensor Cores, and 2 NVDLA engines.

Jetson TX2 NX uses an older Pascal GPU with 256 CUDA cores and no Tensor Cores or NVDLA engines.

**What this means in practice:** the Xavier NX 8GB is the AI-oriented module of the pair — its 48 Tensor Cores and 2 NVDLA engines drive a stated 21 INT8 TOPS of AI performance, whereas the TX2 NX's catalog entry carries no TOPS figure and its acceleration story rests on the 256-core Pascal GPU alone. The TX2 NX is a reduced-form-factor Pascal module that is pin-compatible with the Xavier NX form factor, making it primarily a general-purpose or lower-cost option rather than a deep-learning workhorse.

Jetson Xavier NX 8GB vs TX2 NX: GPU architecture, CUDA cores, Tensor Cores and NVDLA engines

Sources

Attribute	Jetson Xavier NX 8GB	Jetson TX2 NX
GPU architecture	Volta	Pascal
CUDA cores	384	256
Tensor Cores	48	0
NVDLA engines	2	0