DX11 vs DX10.1 performance test

[Sylvanas]

Click

A distinct difference with no AA and a few frames with AA, I'd say it's probably GPU limited at this point and seeing as this is a 5770 I'm sure a 5870 would continue the large gap with AA enabled- would be interestng to find out.

 

[mejobloggs]

Pretty decent. I quite like the sound of this, hopefully it's not just Battleforge specific

 

[RussianSensation]

Similar Findings for Battleforge at Firingsquad

 

[nyker96]

looks like a 20% jump, not bad.

 

[AzN]

sweet.

 

[OVerLoRDI]

If this trend is true across all DX11 games not just battleforge that would be awesome. Would be great to see directx not just get new features but also get faster.

 

[Ben90]

Wow complete overload of statistics recently.... 5770 is slower than 4870.... but dx11 is faster.....prices......power consumption.....features.... WTF card do i get?

...glad i dont really have to worry about it cuz im content with my comp till approximately gtx 4/500 & r1k timeframe

 

[MODEL3]

Nice.
I remember also an interview with Stormrise developers that were talking about how faster (and with higher image quality) DX10.1 can be in relation with DX10.

http://www.pcgameshardware.com...e-DX-101-support/News/

Artem Kulakov: We are currently working with AMDs engineers to implement the following DX10.1 features:
? DX10.1 allows you to read back from a Multi-Sampled Anti-Aliased (MSAA) depth buffer, which means it is no longer necessary to render depth out separately. This means that the current Multiple Render Target (MRT) setup can be bypassed for DX10.1 capable HW, and therefore yielding a performance gain.
? DX10.1 introduced fixed sample patterns for MSAA modes, and also allows the Pixel Shader (PS) to output the MSAA Coverage Mask. This enables us to gain full MSAA Alpha Tested geometry, leading to higher visual quality.
? DX10.1 adds a new instruction called Gather, which can gather 4 texture samples at once, at a much lower cost than issuing 4 separate Sample instructions. Consequently we are able to optimize our shadow map technique, and even shoot for higher quality.
? The Gather instruction will also allow us to optimize our Screen Space Ambient Occlusion (SSAO) algorithm - again producing a higher image quality

I would like to see what the performance difference will be between a 5770 and a GTX260 in 6 months from now. (i suppose then we will have at least 5 DX11 titles to test and many others (?) will employ a DX10.1 codepath)