Jetson Modules [Linkarrow-up-right]
Jetson Orin Power Mode [Linkarrow-up-right]
Jetson Dev Kit AI (64GB AGX Orin) Performance [Linkarrow-up-right]
AI Performance Study by HAL [Linkarrow-up-right]
The following rough estimate assumes that CPU performance won't be the bottleneck for AI inference.
The AI performance has a linear correlation with the frequency of either the GPU or DLA.
JAOi MAXN Mode: 248 TOPS (2* 52.5 + 143)
JAOi has two DLA, each DLA contributes 52.5 TOPS
The GPU of JAOi contributes 143 TOPS
JAOi 35W Mode: 176.8 TOPS (2* 51.5 + 73.8)
JAOi has two DLA, each DLA contributes 51.5 TOPS (52.5 * 1382.4/1408)
The GPU of JAOi contributes 73.8 TOPS (143 * 612/ 1185.75)
But the above estimataion is different from the number disclosed by NVIDIA [refarrow-up-right]
Last updated 1 year ago