Can DX11 really only use one CPU core to communicate with the GPU?

[itachipirate]

I was looking into DX12 a bit to see if it would breathe new life into my FX-6300. I heard in some videos that D3D11 could only use a single CPU core to communicate with the GPU. But I've seen my FX-6300 on GTA V receiving full CPU usage across all cores. So in GTA V on my FX-6300, is one CPU core communicating with the GPU while the other 5 do other physics and AI calculations?

 

If DX12 and Mantle allow the CPU to use all cores to communicate with the GPU, and every DX11 game can only use one core to communicate with the GPU,  would that completely remove my CPU bottleneck in GTA V by using all 6 cores to keep the graphics card fed with information instead of just 1? Would it be a theoretical 6x boost over DX11 for CPU-GPU communication in the case of my 6 core CPU?

 

Assuming that the workload is purely the CPU telling the GPU to draw something, on a 6 core processor. Using D3D11 1/6th of the CPU is feeding information to the GPU. While D3D12 would have the CPU being fully utilized for sending information to the GPU.

 

Is my thinking correct?

[Kherm]

Pretty much, that's why you hear 'single core performance' so often

 

However, you won't see a 6x boost, It'll be like this

 

C1(Usage---------------------------------------------------)                 |

C2(Usage---------------------------------------------)                         |

C3(Usage---------------------------------------------)                         |

C4(Usage---------------------------------------)                                |

C5(Usage-----------------------------------)                                     | 

C6(Usage-------------------)                                                      |

[itachipirate]

Pretty much, that's why you hear 'single core performance' so often

 

However, you won't see a 6x boost, It'll be like this

 

C1(Usage---------------------------------------------------)                 |

C2(Usage---------------------------------------------)                         |

C3(Usage---------------------------------------------)                         |

C4(Usage---------------------------------------)                                |

C5(Usage-----------------------------------)                                     | 

C6(Usage-------------------)                                                      |

Could you explain why it wouldn't be a 6x boost in terms of a (theoretical) workload where the CPU is telling the GPU to draw something and no other processing being done

 

EDIT: I think this graph shows what you're saying. It's better distribution but not perfectly divided distribution

[itachipirate]

This seems to really show promise for eliminating the CPU bottleneck on my FX-6300 by allowing all 6 cores to be used for feeding the GPU with draw calls

 

 

http://stackoverflow.com/questions/4853856/why-are-draw-calls-expensive

 

"The main reason to make fewer draw calls is that graphics hardware can transform and render triangles much faster than you can submit them. If you submit few triangles with each call, you will be completely bound by the CPU and the GPU will be mostly idle. The CPU won't be able to feed the GPU fast enough."