According to a new rumor NVIDIA’s upcoming Pascal architecture will not be significantly better at Async compute than its predecessor. We do know that Pascal will be coming out sometime this year and will definitely have far better FP64 among other things. Without proper Async NVIDIA hopes to rely on raw performance numbers alone.
For those who do not know Asynchronous compute has been very successful with AMD-based graphics cards since DirectX 12 hit the stage. AMD graphics cards are currently leading all DirectX 12 benchmarks available on the market right now, and this is due to the fact that AMD GPUs can process ASync concurrently. Another interesting thing is that NVIDIA GPUs actually perform better with Async turned off. This is probably due to the fact that Nvidia has disabled ASync from their driver suite due to the fact that its GPUs cannot process ASync concurrently on the hardware level, rather they need context switching which is expensive in terms of frame rate. Geforce GPUs instead rely on a technique called pre-emption. Which is why you see frame rates actually being unpredictable when ASync is forced.
Now on the other hand Maxwell GPUs are the only discrete architecture on the market that supports DirectX 12 feature level 12_1. This does allow GeForce cards to use advanced rendering techniques like VXGI/VXAO and Hybrid Ray Traced Shadows. So you have an upside and downside if you are in either the red or green camp. Both Polaris and Pascal are supposed to bridge this gap and unify full compatibility with DirectX 12. You do have to keep in mind that the entire idea behind ASync is to maximize the use of a GPU’s resources to allow the maximum possible performance.
Now if Pascal cards do have Async problems then NVIDIA will rely heavily on driver optimizations and game developers kindness. This could mean that GameWorks optimizations could play a large role in the company’s strategy. Maybe this is the reason that NVIDIA recently made the GameWorks source code publicly available?