Jetson Xavier AGX computation speed comparison with Intel core i7 3.8GHz

Autonomous Machines Jetson & Embedded Systems Jetson AGX Xavier
1

I’m working with a data structure as follows: QVector<QVector> and using it in computations is very slow on the agx compared to the intel i7 3.8ghz computer. When I change the data structure to struct* instead it is much faster.

I’m also working with pointcloud data stored as vector<vector> and the same computations on the agx is about 10 times slower compared to the x86.

Is this speed discrepancy expected. Should I be using different data structures or thinking/programming differently for the two machines?

3

Hi @kfuseini, it’s hard to say what performance is expected without knowing more about the algorithm.

However if you’ve found that C-style arrays perform better than vector (which incur extra look-up costs and overhead), then it’s safe to say you may want to consider using those instead. Also using C-style arrays would allow you to store the data in CUDA/shared memory and offload computation to the GPU, which is where you will see the real speedup using Jetson.