Open kphillisjr opened 5 years ago
Having a chapter on this is an interesting idea, but unfortunately I'm not able to personally write it at this time since I don't have a computer with multiple Vulkan capable GPUs.
Maybe I could give an hand with testing as I have a double GPU configuration (2 same cards).
I think the writer of a chapter on this would have to be the same person as the one with multiple GPUs otherwise there'll be too much back and forth.
If so, I will need a lot more exercising before producing anything :smile:
I too would like to see a chapter on this, an interesting approach might be an example using the DGPU for a graphics workload and the IGPU for a compute workload in the same application.
I too would like to see a chapter on this, an interesting approach might be an example using the DGPU for a graphics workload and the IGPU for a compute workload in the same application.
this is a good use case
I was looking over the various Documents on the tutorial, and saw that there is no information on how to get started working with multiple graphics cards. Some of the use cases I know of for this are as follows...
Alternate frame Rendering using two identical cards
Split Frame Rendering ( horizontal) - This is where the screen is segmented into multiple sections and the work is split among all the graphics cards.
Split Frame Rendering (Vertical ) - This is important for VR Demos since the scene is rendered in a fashion where each graphics card renders Half of the VR Scene.
Advanced Rendering Mode (aka Explicit Multi-gpu)- This is where the developer renders independent items to each graphics card, and at the end manually synchronizes the resulting frame. An OpenGL Example for this involves the Extensions of either WGL_AMD_gpu_association or GLX_AMD_gpu_association. graphics card, and then composite the end result together
Mixed Mode Explicit Multi-GPU - This is where multiple sources are used. Although the main targets would be dedicated graphics and integrated graphics sources.
Note: Most of the tutorial is about transferring data between the graphics cards, and a lot of users may have a dedicated card ( AMD or NVidia), and a integrated card ( Intel, or amd). The basic use case in this fashion would be to offload stuff like physics, particles, or similar to the integrated card and do most of the complex rendering on the dedicated card.
Edit: Added a few another mode, and a quick note.