Multi-GPU DirectX 12 Results Show R9 Fury X to be The Fastest Single Card in Ashes of the Singularity
Microsoft’s DirectX 12 API provides a number of interesting new features, one of which is the support for the multi-GPU rendering. There are two modes for implementing multiple GPUs under DX12: Implicit Multi Adapter (IMA) and Explicit Multi Adapter (EMA).
The former corresponds to the method which is also used under Direct3D 11, where developers have graphics drivers doing most of the work, with the work allocation usually done by Alternate Frame Rendering (AFR). Explicit Multi Adapter (EMA) on the other hand opens completely new possibilities to the game developers.
As the name suggests, the developer must explicitly trim its code on multi-GPU rendering. It’s up to them how they use the resource and load distribution, which means they should be able to increase performance dramatically – or ruin everything due to an error. Due to the high development costs, however, Explicit Multi-GPU is used by only a few programming Studios.
The latest beta version of Stardock’s real-time strategy game Ashes of the Singularity offers full support for EMA. Which means, for the first time we can stick an AMD and Nvidia card into the same PC and observe immense performance gains. Possible usage scenarios are not only limited to coupling different Radeon and Geforce cards together, but also forging an Intel graphics processor.
Folks over at PC Games Hardware paired a Geforce GTX Titan X with a Radeon R9 Fury X, showing some exciting results. In addition to this, they also had some other graphics cards shootouts through Ashes benchmarks. Here’s the system specifications:
- Intel Core i7-6700K @ 4.5 GHz,
- 2 × 8 GB DDR4-2800,
- Asus Z170.
As obvious, the Radeon R9 Fury X is the fastest single card in Ashes of the Singularity under DirectX 12, when paired with various render partners. Unsurprisingly, both the Radeon R9 390X and Geforce GTX Titan X, when combined with the Fury X, proved to be the most effective in increasing the frame rate significantly – although the measurements were performed only in Full HD.