Of the two module families the HZHY-AI700 supports, Jetson AGX Orin has the higher memory bandwidth. Its GPU is Ampere. hzhytech.comhzhytech.com ↗“compatibility.supported_jetson_families=Jetson AGX Xavier, Jetson AGX Orin”
AGX Orin vs. AGX Xavier modules nvidia.com +5nvidia.com ↗“memory.memory_bandwidth_gb_per_s=204.8”nvidia.com ↗“Ampere SoM; up to 200 INT8 TOPS (sparse), 1792 CUDA + 56 Tensor cores, 8-core Cortex-A78AE, 32GB LPDDR5”nvidia.com ↗“Industrial-grade Ampere SoM”developer.nvidia.com ↗“memory.memory_bandwidth_gb_per_s=136.5”developer.nvidia.com ↗“Volta SoM; up to 32 INT8 TOPS, 512 CUDA + 64 Tensor cores, 8-core Carmel, 32GB LPDDR4x”developer.nvidia.com ↗“Industrial Volta SoM”
| Module family | Modules | Memory type | Memory bandwidth | GPU architecture |
|---|---|---|---|---|
| Jetson AGX Orin | AGX Orin 64GB, AGX Orin 32GB, AGX Orin Industrial | LPDDR5 | 204.8 GB/s | Ampere |
| Jetson AGX Xavier | AGX Xavier 64GB, AGX Xavier 32GB, AGX Xavier Industrial | LPDDR4x | 136.5 GB/s | Volta |
Memory bandwidth is the same within each family, no matter which module size you choose.
The HZHY-AI700's listed performance is 275 TOPS and 5.32 TFLOPS FP32. These figures are based on the module inside. The 275 TOPS figure matches the Jetson AGX Orin 64GB module (275 INT8 TOPS, sparse). With an AGX Xavier module, AI performance tops out at 32 INT8 TOPS (dense). hzhytech.com +2hzhytech.com ↗“performance.max_ai_tops=275 TOPS, performance.max_fp32_tflops=5.32 TFLOPS”nvidia.com ↗“up to 275 INT8 TOPS (sparse)”developer.nvidia.com ↗“up to 32 INT8 TOPS”