Hitman Episode Two Update Breaks DirectX 12 Support

[the54thvoid]

Gotta love low level APIs that give extra couple of frames per second and break game with the next content update ... devs are used to monolithic dx11 where driver knows the hardware so you don't need extensive testing after every content update once the engine is tested enough.
What happened here is a great example of low level api being a double edged sword.

That flies with what I've read about it. You just can't think DX12 is so easy to use that you don't have to bother trying. It still has rules and even the almighty Async comes with guidelines on how to implement it. If it's done with minimal queuing it wont negatively affect Nvidia. If it's balls out Async (AoS) then it will hamper Nvidia.

Again, shoe is on the other foot now and I don't hear people crying foul about AMD GE titles ramping up Async.

However, going forward, if Pascal isn't good enough with Async and AAA's start using it more (thinking Deus Ex) then my next card may well be Vega.

Still too early to tell though. Allwe are seeing is immature implementation of DX12 in quite a lot of titles.

 

[BiggieShady]

That flies with what I've read about it. You just can't think DX12 is so easy to use that you don't have to bother trying. It still has rules and even the almighty Async comes with guidelines on how to implement it. If it's done with minimal queuing it wont negatively affect Nvidia. If it's balls out Async (AoS) then it will hamper Nvidia.

Again, shoe is on the other foot now and I don't hear people crying foul about AMD GE titles ramping up Async.

However, going forward, if Pascal isn't good enough with Async and AAA's start using it more (thinking Deus Ex) then my next card may well be Vega.

Still too early to tell though. Allwe are seeing is immature implementation of DX12 in quite a lot of titles.

From what I have read, it's all about what happens in a single frame of a game. Best case for nvidia is to batch draw calls and then separately batch async compute commands ... you get parallelism in both separately (but not mixed). Coincidentally that is the worst case scenario for amd. GCN gets best effectiveness if 3D and async compute commands are mixed in the same queue.
Looking at the net result on the level of a whole frame, if done optimally for each architecture, effectiveness should be comparable for both arch. but GCN is certainly easier to code for in this case.
So the issue will be with direct ports from consoles ... as usual, but for nvidia only this time.

 

[corsaro]

First image looks whose then Far Cry 2004!

 

[Prima.Vera]

First image looks whose then Far Cry 2004!View attachment 74116

Still THE BEST FarCry to date, miles away from the junk relesed after. Wish the make a real sequel to that...