Specifications
Power
| Power Envelope | 40–130 W |
|---|
Memory
| Memory | 128 GB |
|---|---|
| Memory Type | LPDDR5X |
| Memory Bandwidth (GB/s) | 273 |
Compute
| CPU Cores | 14 |
|---|---|
| GPU CUDA Cores | 2,560 |
| GPU Architecture | Blackwell |
Thermal
| Module Max TTP Temperature | 80 °C |
|---|---|
| Module Max Junction Temperature | 115 °C |
Identity
| Device Type | System-on-Module |
|---|---|
| Jetson Family | Jetson Thor |
Performance
| Max AI Performance (TOPS) | 1,035 TOPS |
|---|---|
| Max FP4 AI Performance (TFLOPS) | 2,070 TFLOPS |
| Max FP32 TFLOPS | 8.064 TFLOPS |
Category: NVIDIA / NVIDIA Hardware / Production Modules
Questions about this product
- Humanoid robots that use Jetson Thor
- Jetson AGX Thor vs IGX Thor Developer Kit: Ethernet speed, SFP ports and FP4 performance
- Jetson module with at least 16 GB of memory and the highest AI performance
- Jetson module with at least 200 TOPS and the highest TTP temperature for an engine bay
- Jetson production modules with 12 or more CPU cores
- Jetson production modules with 64 or more Tensor Cores
- Jetson production modules with the 699-pin B2B (100x87mm) connector
- Jetson T4000 vs AGX Thor T5000: CPU cores, CUDA cores and memory
- Jetson Thor humanoid with the most degrees of freedom
- Jetson Thor module that fits a 75 W power rail
- Lowest-power NVIDIA module with at least 64 GB of memory
- Module with the most GPU CUDA cores in the Jetson Nano's category and its memory
- NVIDIA developer kits with at least 64 GB of memory
- NVIDIA modules and kits built on the Blackwell GPU architecture
- NVIDIA production module with the highest FP32 TFLOPS
- Production modules with more than 200 GB/s of memory bandwidth
- Unitree robots running a Jetson Orin NX