why cpu faster than gpu?
    7 views (last 30 days)
  
       Show older comments
    
My matlab version is 2022a. I ran the DDPG of 'SimplePendulumWithImage-Continuous' for 50 epidodes with 
criticOptions.UseDevice = 'cpu';actorOptions.UseDevice = 'cpu';
and 
criticOptions.UseDevice = 'gpu';actorOptions.UseDevice = 'gpu';
Surprisingly, it took 712 seconds for cpu and 1202 seconds for gpu. How did this happen?
The cpu is R5800H and the gpu is 3070 max-q.
3 Comments
  Joss Knight
    
 on 14 Sep 2022
				We're looking into it. It may be that this example isn't complex enough to get a benefit from the GPU.
Answers (1)
  Vidip Jain
    
 on 20 Feb 2023
        In the case of the DDPG algorithm for the 'SimplePendulumWithImage-Continuous' environment, the performance may be influenced by the size and complexity of the model, the number of episodes, and the batch size used during training. It is possible that the CPU in your system is better suited for this specific task than the GPU. This can happen if the GPU is not fully utilized during training due to factors such as small batch size or insufficient parallelization. In such cases, the CPU may perform better because it is optimized for sequential operations and can better utilize the available resources for this specific task.
0 Comments
See Also
Categories
				Find more on Parallel and Cloud in Help Center and File Exchange
			
	Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!