Considering that DX12 is still in alpha (EAP) and the drivers status are.. welll, everyone could imagine how current drivers are.... I can only say: not bad.

EDIT: AnandTech should NOT have such info like current drivers and information about feature levels :| I waited months to sign the NDA :|

TLDR version; DX12 is basically as fast as Mantle (2 or 3 fps slower on the GPU, 4 or 5ms slower on the CPU) and destroys DX11 both in terms of FPS and CPU time to submit.

Hm, so D3D12 will only work on Windows 10, not even 8.1.

I wonder how much deattached from the world Khronos people are, but right now it looks just like Windows Vista and D3D10 days, they have another chance of bringing a cool API that can work on most OS installs right now. I really hope they're seeing it.

Good thing now is that they don't have the "OpenGL won't work on Windows" thing to deal with right now

They had three options:

- Add WDDM 2.0 to Windows 7/8/8.1 -> really expansive for Microsoft, quite cheap for HIVs;

- Add Direct3D 12 support for WDDM 1.x -> less expansive, but the API will not act like with WDDM 2.0, HIVs had to write drivers for two different driver models;

- Forget about WDDM 1.x OSs -> no additional costs for both Microsoft and HIVs;

Starting with Vista, Microsoft decided to integrate DirectX graphics into the OS (GUI rendering with DWM, video encoding etc.), and the driver model become strictly related to the kernel. So they obtain both a technical and an economical excuse to not upgrade DirectX for previous versions of the OS (Windows Vista was the only big exception, it gains Direct3D 11, WDDM 1.1 and DXGI 1.1, however DWM didn't take any advantage of WDDM 1.1 and DXGI 1.1).

1 year of free upgrade is a good move, Vista licenses were the most expensive licenes of a commercial OS ever seen (600£$€ for Windows Vista Ultimate full reatail!), a lot of people did not upgrade from XP to Vista due hardware requirements of Vista, due bad reputation and FUD (did everyone forget the "palladium" FUD tragicomedy?).

Also, keep in mind that most gamers do not have any problem to pirate the OS, so if they want to upgrade to Windows 10 after the free upgrade period, they will do it.

Finally, I expect that WDDM 2.0 will bring benefits to Mantle and OpenGLNext too. ;)

Considering that DX12 is still in alpha (EAP) and the drivers status are.. welll, everyone could imagine how current drivers are.... I can only say: not bad.

EDIT: AnandTech should NOT have such info like current drivers and information about feature levels :| I waited months to sign the NDA :|

The feature levels stuff is effectively public knowledge by now - you just need to run dxdiag on the latest Windows 10 preview to see them.

IMO this is one of the most interesting parts, and it's relevant to discussion of GL Next too. I wonder how much of D3D12's performance gains come out on lower feature levels? Because if D3D12 can also bring the kind of higher performance reported in that article to D3D10 and 11 class hardware, that's going to set a bar for the ARB: GL Next absolutely cannot have a requirement for the newest hardware only otherwise it's going to have a rocky first few years in the PC/Windows space.

Considering that DX12 is still in alpha (EAP) and the drivers status are.. welll, everyone could imagine how current drivers are.... I can only say: not bad.

EDIT: AnandTech should NOT have such info like current drivers and information about feature levels :| I waited months to sign the NDA :|

The feature levels stuff is effectively public knowledge by now - you just need to run dxdiag on the latest Windows 10 preview to see them.

IMO this is one of the most interesting parts, and it's relevant to discussion of GL Next too. I wonder how much of D3D12's performance gains come out on lower feature levels? Because if D3D12 can also bring the kind of higher performance reported in that article to D3D10 and 11 class hardware, that's going to set a bar for the ARB: GL Next absolutely cannot have a requirement for the newest hardware only otherwise it's going to have a rocky first few years in the PC/Windows space.

In theory only DXGI should be dependent on the WDDM version and D3D itself should be independently implementable; see https://msdn.microsoft.com/en-us/library/bb205075.aspx for example:

Microsoft DirectX Graphics Infrastructure (DXGI) recognizes that some parts of graphics evolve more slowly than others. The primary goal of DXGI is to manage low-level tasks that can be independent of the DirectX graphics runtime.

In practice that hasn't quite worked out, and a new D3D version equates to a new WDDM version, meaning a new DXGI version inbetween. The table at http://en.wikipedia.org/wiki/Windows_Display_Driver_Model#WDDM_1.0 is quite telling and the explicit intention of splitting off DXGI has obviously failed; whether that be for solid technical reasons or on account of something else, I don't know and won't comment. Either way, the current architecture doesn't really reflect the reality of how the APIs are related and interdependent.