My approach to SLi or Xfire is definitely a wait and see. For example a lot of UE4 demos do not support multiple gpus (it depends on temporal accumulation for certain things so AFR breaks their implementation).
With upcoming dx12/Vulkan APIs, all the responsibility for implementing multi-gpu is going to reside with game developers and not the IHV. On one hand it is better as you can implement better solutions like SFR, also the IHVs won't have to hack around engines just to get multiple gpus working. On the other hand game devs are going to have to put in extra work to implement full multi-gpu support (versus the current approach where they have to do very little work), where all this extra work is on top of the greater workload to use these low level APIs.
So basically tl;dr Multi-GPU support may improve, but it is just as likely to become far worse.
Log in to comment