GPUs are great for data processing in a multidimensional data structure. These data structures are inherently parallelizable. Let's see how we can use the GPU for multidimensional data processing in Python:

1. First, let's import the Python packages that are needed:

import numpy as np
import cupy as cp
import time

2. We will be using a multidimensional array in NumPy, which is a traditional Python package that uses the CPU. 

3. Then, we create a multidimensional array using a CuPy array, which uses the GPU. Then, we will compare the timings:

### Running at CPU using Numpy
start_time = time.time()
myvar_cpu = np.ones((800,800,800))
end_time = time.time()
print(end_time - start_time)

### Running at GPU using CuPy
start_time = time.time()
myvar_gpu = cp.ones((800,800,800))
cp.cuda.Stream.null.synchronize()
end_time = time.time()
print(end_time - start_time)

If we will run this code, it will generate the following output:

Note that it took around 1.13 seconds to create this array in NumPy and around 0.012 in CuPy, which makes the initialization of this array 92 times faster in the GPU.