DX11 generation versus DX10

jww20

Weaksauce
Joined
Jan 9, 2003
Messages
72
http://vr-zone.com/forums/487528/why-directx-11-matters-pic-.html

TWIMTBP? Boxy looking air filter versus a realistic rendering.

dx11renderingcomparison.jpg




Win 7 and DX11 make the 5870 easily outperform dual GPU DX10 cards.

pcamd5870battle1a.jpg
 
The 5870 is faster and has better image quality in games that support DX11. Check.
I'm just waiting for someone to yell, but it doesn't have PhysX!! *palmface*
 
I have an hd4870x2 :confused: . I'm saying that can be very well done without dx11, hell dx9 could do that.
 
With hardware tessellation, it'd definitely not BS.

Radical increase in triangles, to simulate a true curves..

Hardware tessellated games will look a lot better than games beforehand, if the programmers make use of it.
 
I believe that Stalker and I believe it's a TWIMTBP game. I the op added it for the dramatic effect and/or a play on words. :D

Yeah but they could make that look much better without tessellation and just a few more polys. It's not like we've never had round objects in games before :p

I think AVP will be the first game to use it properly. It doesn't make sense to just improve the air filter on a character model when everything else still looks flat and unfeatured (the straps and doohickies on his suit for example are all still just flat textures).
 
Yeah but they could make that look much better without tessellation and just a few more polys. It's not like we've never had round objects in games before :p

Of COURSE you can make it look better by adding more polygons. That's what tessellation does.

The whole point is that it adds more geometric detail the closer the object gets. That way you don't waste GPU power rendering so many triangles when the object gets farther away.

To say "big deal, you can make a model that looks just as good in DX10 if you added more triangles" misses the entire point. The point is that DX11 does this procedurally on-the-fly, which makes it far easier for developers and makes it feasible to have insane polygon counts.
 
We've actually had this technology since about 1997/1998 or so. In fact, Intel had spent some time developing a system for dynamic LOD, but I don't recall it ever having been adopted at any point.

The main difference is that Intel's solution utilized a higher-detail mesh and a lower-detail mesh and simply blended between the two dynamically, whereas tessellation interpolates using a lower-detail mesh.
 
I believe the picture is real, and it looks great :cool: However unless adding tessellation support to a whole game is easy most devs won't bother, same as they've not bothered much with anything else that PC graphics cards can do but consoles can't.

Incidentally I doubt adding tessellation is easy as you'd have to specify how you want everything tessellated (or you'll end up the the truform Michelin man effect), might need different textures too due to tessellated models having different normals (or you'll end up with funny looking textures with normal maps compensating for edges that are no longer there).

Still we can hope ...

Incidentally don't know what the ambient occlusion is meant to be doing - if anything the shadowing looks better in the DX10 picture. That's also not a DX11 effect - been in nvidia drivers for a while now.
 
Last edited:
We've actually had this technology since about 1997/1998 or so. In fact, Intel had spent some time developing a system for dynamic LOD, but I don't recall it ever having been adopted at any point.

The main difference is that Intel's solution utilized a higher-detail mesh and a lower-detail mesh and simply blended between the two dynamically, whereas tessellation interpolates using a lower-detail mesh.

Radeon 8500 :p? TruForm? !?!
 
Didn't QuakeIII start the curved services thing where tessellation scaled?
I vaguely remember lots of round arches and stuff in the gothic architecture. Always wondered why whatever they did never made it into other things.
 
I think truform was a case of ati doing a tech, having it adopted by several game engines, and than never successfully getting microsoft to implement it into directx... which is what they wanted.
 
I think truform was a case of ati doing a tech, having it adopted by several game engines, and than never successfully getting microsoft to implement it into directx... which is what they wanted.

well I guess they got it now :D
 
I believe the picture is real, and it looks great :cool: However unless adding tessellation support to a whole game is easy most devs won't bother, same as they've not bothered much with anything else that PC graphics cards can do but consoles can't.

Incidentally I doubt adding tessellation is easy as you'd have to specify how you want everything tessellated (or you'll end up the the truform Michelin man effect), might need different textures too due to tessellated models having different normals (or you'll end up with funny looking textures with normal maps compensating for edges that are no longer there).

Still we can hope ...

If it works anything like MeshSmooth from 3DSMax then texture/normal maps for the base mesh will be applied correctly to the new topology. Normal mapping should actually be improved, as a denser mesh produces better results. I suspect that it will be reasonably easy to implement, but it cannot be too easy without suffering the same problems you mentioned for TruForm. You need to be able to control the maximum smoothing angle (typically anything 90° or more should be left alone) and number of iterations for each model. Sometimes you need 'hard' edges, even though they would normally be smoothed - in this case extra geometry is required (use 2 edges in very close proximity, which results in an edge with a natural 'width' that shows specular highlights correctly), or a means to mark such edges. I'm sure this was all considered.
I'd bet this will be quite widely adopted, just as normal mapping has been. It's surprising that it has taken so long for it to take off.
 
It's surprising that it has taken so long for it to take off.

I'm pretty sure ATI has been pushing for it for quite a while, since dx8 level hardware, and nvidia has been fighting against it, until dx11.

Until the hardware was in both sets of cards, no chance of it showing up at all.
 
I have a hard time believing that making a perfect 3d circle in a game is something that couldn't be done before DX11...
 
I have a hard time believing that making a perfect 3d circle in a game is something that couldn't be done before DX11...

Of course it could be done in DX9/10, but it takes more performance compared to DX11 and the tessellation feature. This allows them to give high quality graphics with lower cost, which they can then put to use with A.I., physics, and post-processing features like ambient occlusion. When Nvidia comes out with a DX11 card, they'll get the same benefit and there will be parity(graphic quality wise) between the two IHVs.
 
even though I can get a general estimate, they shoulda included 5870 @ DX10 in that graph, just for comparison.
 
Of COURSE you can make it look better by adding more polygons. That's what tessellation does.

The whole point is that it adds more geometric detail the closer the object gets. That way you don't waste GPU power rendering so many triangles when the object gets farther away.

To say "big deal, you can make a model that looks just as good in DX10 if you added more triangles" misses the entire point. The point is that DX11 does this procedurally on-the-fly, which makes it far easier for developers and makes it feasible to have insane polygon counts.

Sigh, I'm getting tired of all these internet gurus. The point is that for a simple object like that air filter it would be trivial to just send extra polys down the pipe. There will be no wasted GPU rendering power for such a small model. Tessellation is a win only if the polygon increase is non-trivial.

I have a hard time believing that making a perfect 3d circle in a game is something that couldn't be done before DX11...

Exactly. It's a pretty trivial case that has been done for years. People are acting like they've never seen a curved surface in a game before :)
 
Sigh, I'm getting tired of all these internet gurus. The point is that for a simple object like that air filter it would be trivial to just send extra polys down the pipe. There will be no wasted GPU rendering power for such a small model. Tessellation is a win only if the polygon increase is non-trivial.

Isn't that the whole point with Tessellalation? To be able to increase on a large scale without the heavy performance impact? Looks like you two agree though, since smartpatrol is saying tessellation "makes it feasible to have insane polygon counts." . :)



Exactly. It's a pretty trivial case that has been done for years. People are acting like they've never seen a curved surface in a game before :)

I'm not impressed either with this example. Its not like "WOW, now we can have cuved surfaces". :p Seen it before without DX11. I don't like that they use Nvidia vs. ATI either, since this is DX10 vs. DX11. Flamebait is what I call such examples and does no good in discussing the tech itself.
 
The 5870 is faster and has better image quality in games that support DX11. Check.
I'm just waiting for someone to yell, but it doesn't have PhysX!! *palmface*

Eh don't worry Atech will be in here soon to let us know how great PhysX is
 
Eh don't worry Atech will be in here soon to let us know how great PhysX is

LOL! :D

I wonder if he will accept arguments like trinibwoy uses (with edits):

Exactly. It's (PhysX) a pretty trivial case that has been done for years. People are acting like they've never seen a (flying paper, broken glass, banners etc) curved surface in a game before :)
 
so the gtx-295 is still over 30fps? cool no need to upgrade

the GTX-295 is running in DX10 mode, not DX11 like the 5870.

So 30fps in DX10 or 45fpx in Dx11..

I would take Dx11 > Dx10
 
the GTX-295 is running in DX10 mode, not DX11 like the 5870.

So 30fps in DX10 or 45fpx in Dx11..

I would take Dx11 > Dx10

I would too if there was a choice.(I will buy a next gen DX11 card) The 295 does have support for DX11 features and as far as I understood, some of the effects can be downscaled on DX10/10.1:

slide10.jpg

http://www.pcper.com/article.php?aid=783

I just hope that DX11 doesn't simply remove the effects and puts no effort optimizing non-DX11 platforms. That would only hurt gamers.

Dirt 2 went from DX9 directly to DX11:
http://www.pcgameshardware.com/aid,...aphics-on-PC-due-to-DirectX-11-features/News/

I just hope that the features aren't just for pure DX11 cards. That would be bad in my book.
 
Please tell me that you don't believe that 30fps in dx10 will be 45fps in dx11?? I thought we learned our lesson from that horseshit when they were pimping out dx10 to the world back in 06. I believe NOTHING that comes out of their mouths now. NOTHING. Sorta like politicians.

A Little intervew from the treyarch devs from VE circa last year, almost 2 years after dx10 was released:

PCGH: Will the PC version of CoDWaW support DX10?

Cesar Stastny: No.

PCGH: What were your reasons to do without?

Cesar Stastny: The game will be equivalent between PS3, Xbox 360 and PC. This is by design.

PCGH: Will there be a patch after the launch of the PC version that integrates Direct X 10 later?

Cesar Stastny: Maybe if there is a compelling benefit to doing so, such as new superior hardware which requires DirectX 10.

PCGH: From a developer's point of view, can DX 10 be seen as a milestone in game development?

Cesar Stastny: Nope. The reason is that no major consoles including XBox 360 from Microsoft supports it's API. For cross platform developers like Treyarch it doesn't make sense to use DirectX 10 since it is not portable to consoles like DirectX 9 is for Xbox 360.

I have no beliefs that devs think any differently now as many titles are STILL dx9 only almost 3 years after dx10 was released. I wondered why that was. Now I know. Until they update the xbox API support for dx11 then dx9 will reign supreme. Cross platform devs are looking for the least amount of work to port these pigs to all platforms so dx11 will be in a select few titles just like dx10 games are now. DON'T believe the hype,
 
Last edited:
I think tesselation implimentation should be fairly trivial, you just need to flag which models you want to tesselate then decide a coefficient for tesselation which in most cases will simply be distance from viewport.

its very clever, it's like a much more dynamic LOD system which should help large outdoor games a lot, an implimentation in Crysis would probably be very useful, and probably save the developers a lot of time making different quality models.
 
Back
Top