I’ve expressed my views and concerns on GameWorks in two past articles; I’m not going to rehash them here. What I will note is that Nvidia’s decision to create a set of closed libraries that developers can’t share with AMD, combined with developer decisions on whether or not to collaborate with AMD on substantive driver improvements in other areas, have created a scenario in which every GW title must be scrutinized to determine if it treats AMD GPUs fairly (to the extent that this is possible).
Much has been written on whether or not Mantle should be viewed as equivalent to GameWorks, and indeed, it’s true that AMD’s DX11 performance relative to Nvidia’s has been somewhat lower in Battlefield 4 and Thief. That’s something I’ve called on AMD to fix — I think the fact that HD 6000 series users or GCN (Graphics Core Next) users who don’t want to use Mantle don’t get fully optimized DX11 performance is problematic.
Watch Dogs wallpaper
The difference here is one of size and scope. Now that GameWorks is embedded in the Unreal 4 engine, it could impact AMD’s competitive performance for years to come. Mantle, in contrast, is shipping in a relative handful of titles. Furthermore, there’s zero evidence that developers who have embraced Mantle are doing so in a manner that would impact Nvidia’s DX11 performance in any way. Some have argued that because AMD can still perform indirect optimization on GW titles, the impact of the program is meaningless — but this ignores the fact that the vast majority of performance profiling in any given game occurs within weeks of its release. If AMD (or Nvidia) can’t push out an optimized driver in a very small window, the majority of game reviews will show a decided advantage for the home player. Only those games incorporated widely into GPU reviews will show the benefit of iterative driver updates.
