The IGX T7000 is rated higher for FP4 (up to 5581 TFLOPS, sparse, versus 2070 for the T5000). It lists 2 SFP ports, where no SFP count is given for the T5000. They are also different types of device: the T5000 is a system-on-module and the T7000 is a board kit.
IGX T5000 vs. IGX T7000 nvidia.com +4nvidia.com ↗“performance.max_fp4_tflops=2070 TFLOPS”nvidia.com ↗“performance.max_fp4_tflops=5581 TFLOPS”nvidia.com ↗“io.sfp_ports=2”nvidia.com ↗“identity.device_type=System-on-Module”nvidia.com ↗“identity.device_type=Board Kit”
| IGX T5000 | IGX T7000 | |
|---|---|---|
| FP4 performance | 2070 TFLOPS (sparse) | Up to 5581 TFLOPS (sparse) |
| SFP ports | Not listed | 2 |
| Device type | System-on-Module | Board Kit |
More detail on each:
- FP4 performance: Both figures are sparse FP4. The T7000's figure is given as "up to" 5581 TFLOPS. The T7000 supports a discrete GPU (RTX PRO 6000 or 5000 Blackwell) and the T5000 doesn't. The specs don't say whether the 5581 figure assumes a discrete GPU is installed. nvidia.com +2nvidia.com ↗“up to 5581 FP4 TFLOPS (sparse)”nvidia.com ↗“dGPU support (RTX PRO 6000 or 5000 Blackwell)”nvidia.com ↗“no dGPU support (use IGX T7000)”
- SFP ports: No SFP port count is listed for the T5000. Its description does mention 4x 25GbE. The T7000 has 2 SFP ports, and its description lists ConnectX-7 with 2x QSFP112 200GbE. nvidia.com +1nvidia.com ↗“4x 25GbE”nvidia.com ↗“ConnectX-7 (2x QSFP112 200GbE)”
- Device type:
- The T5000 is a module for custom carrier boards, with a 699-pin B2B connector (100x87mm). It's a separate SKU from the Jetson T5000 but uses the same form factor and pinout. nvidia.com +1nvidia.com ↗“for custom carrier boards”nvidia.com ↗“form-factor and pin-compatible with it”
- The T7000 is a MicroATX board kit with functional safety and a BMC. nvidia.comnvidia.com ↗“IGX Thor MicroATX board kit”
Both datasheets are available here.