⋮    ⋮  

DirectX 12 Explicit Multi-GPU Tested In Ashes of the Singularity


Ashes of the Singularity DX12 Updates


Stardock released Beta 2 of its upcoming RTS title Ashes of the Singularity today.  This update brings something that many pc gamers have been curious about and anxious to try for themselves, Explicit Multi Adapter Support and full Asynchronous Compute.

“With this update, players can improve game performance by adding another video card,” said Dan Baker, Chief Scientist at Oxide Games. “As long as the two cards are reasonably similar in performance, they can be of any model or brand including mixing AMD and NVidia cards together. Players just insert the additional video card and enable it via the game’s video options panel.”

What is Explicit Multi-Adapter?  EMA is the ability for DX12 and your system to not only recognize that there are multiple gpus installed but use them, regardless of vendor or class.  So you could pair any number of combinations of Radeon and Geforce gpus.  Theoretically you could also combine Intel’s igpu with a dedicated graphics card, but I had no luck with that on either our z87 test bench or our z170 setup as of now, but upcoming releases may make this possible.  Stardock mentions the importance of keeping the cards similar in performance when mixing them as they’re using Split-Frame Rendering and will be using the gpus in parallel, so one being significantly stronger can have a very negative impact on performance as we’ll see later.

This benchmark should be accessible through the newly updated display preferences option menu in the game.

Another DX12 implementation that has made its way into this update is Multi-Threaded Command Buffer Recording.  With the rise of multi core processors in the market it’s time we see better use of those cores.  Typically with DX11 a single core handles most of the driver and API interpretation leaving little room for the processor to handle the actual game code.  DX12 is utilizing the command buffer to overhaul this in five ways.

  • Overhead is significantly reduced by moving the driver and API code to any available CPU thread
  • The absolute time required to complete complex CPU tasks is notably reduced
  • Game workloads can be meaningfully distributed across more than four cores
  • New “bandwidth” on the CPU allows for higher peak draw calls, enabling more detailed and immersive game worlds
  • All available CPU core man now “talk” to the graphics card simultaneously

This could end up being a great feature for those still holding on to their six and eight core FX series processors, but in reality everyone will benefit from it.