On FP4 performance, memory bandwidth and power, the two modules have the same figures. The IGX T5000 is a separate SKU from the Jetson T5000, but it has the same form factor and pin layout. nvidia.comnvidia.com ↗“a separate SKU from Jetson T5000 (P/N 900-13834-0080-001) but form-factor and pin-compatible with it”
Jetson AGX Thor T5000 vs. IGX T5000 developer... +1developer.nvidia.com ↗“performance.max_fp4_tflops=2070 TFLOPS”nvidia.com ↗“2070 FP4 TFLOPS (sparse)”
| Attribute | Jetson AGX Thor T5000 | IGX T5000 |
|---|---|---|
| FP4 performance | 2070 TFLOPS | 2070 TFLOPS (listed as sparse) |
| Memory bandwidth | 273 GB/s (128 GB LPDDR5X) | 273 GB/s (128 GB LPDDR5X) |
| Power envelope | 40–130 W | 40–130 W |
Neither module has a Super mode, so these are the only figures.
About your hot enclosure:
- Operating temperature isn't listed for either module, so I can't confirm the ambient temperature range either one supports.
- Jetson AGX Thor T5000 thermal limits: the module has a maximum TTP (thermal transfer plate) temperature of 80 °C and a maximum junction temperature of 115 °C. These are limits for the module itself, not the air temperature it can run in. Your heatsink or cooling has to keep the plate at 80 °C or below at the power level you choose. developer...developer.nvidia.com ↗“thermal.max_ttp_temp_c=80 °C, thermal.max_junction_temp_c=115 °C”
- IGX T5000 thermal limits: TTP and junction limits aren't listed for this module.
- Where to check: the operating temperatures for your design should be in the Jetson Thor series datasheet and the IGX Thor datasheet.
Other differences:
- The IGX T5000 is meant for custom carrier boards and comes with optional functional safety. nvidia.comnvidia.com ↗“for custom carrier boards with optional functional safety”
- The IGX T5000 doesn't support a separate GPU (dGPU). For that, you'd use the IGX T7000. nvidia.comnvidia.com ↗“no dGPU support (use IGX T7000)”