Tesselation-great tech,but not usable in current gen?

I hope it is like that,but I am struglling to understand how you can have better performance with more dense geometry to calculate.
It is impossible to have better performance with tess. than without.
How can anything process faster more polygons(vertices) than the lower res. geometry,please explain?
In avp maybe it(tess) is not in heavy usage and for this reason some dx11 optimization can help.

 

Dx11 is faster than dx9. Tesselation on is not that expensive on ATI is its got dedicated hardware.

www.legitreviews.com/article/1225/1/

Worth a read.

 

this is from Heaven benchmark review - http://www.pcgameshardware.com/aid,...chmarks-Top-article-of-October-2009/Practice/

so what you think its not procesed in shaders but in tessellator so no it doesnt take a hit but yea ati supports only std basic tessellator while nv is more advanced and more flexible which makes it faster..

 

i recall reading somewhere that the using nvidia's tesselator requires a whole shader cluster to be dedicated to tesselation, and removes its from its rendering duties or something, however, this is probably also just a rumour.

and apparently the benchmark that said that nvidia's card performs 40% faster in tesselation than an ati card was apparently with every setting switched off for the nv card except for tesselation. i guess we'll just have to wait a couple of months to see if these rumours held any truth.

 

If tessellation is done right then we should see some pretty desirable results even with the hardware we have available to us now.

 

The implementation we can expect to see at this stage and onwards for the next few years will probably be to simply replace the current use of parallax occlusion mapping with tessellation in DX11 mode.

 

Being that the only person that i've saw mention such a thing is the most anti Nvidia guy on the planet, im sure he would of mentioned it, especially when he claims over 100% better performance with tessellation.

 

Why would they have to disable whole SP cluster when they got own part for tesselation? I think he got it backwards... its ATi that takes a hit with trying to use both heavy SP work and tesselation same time.

 

If each polymorph is integrated in the cluster, i think that is because the cluster only can do one tipe of task, so, or it is tesselation or else.

 

ATI has a dedicated Tesselator seperate from the SP's so I don't know where you got that from? Nvidia uses the SP's for tesselation as well as everything else.

Tesselation is just one part of the hull and domain shadesr so whilst its effectively free the rest of the shaders are not.


http://www.realtimerendering.com/blog/direct3d-11-details-part-ii-tessellation/

My tests with AVP show that tesselation hardly affects performance, but enabling the other stuff such as DX11 shadows and AA (Particularly AA) crank up the heat.

I run AVP at 1080p all DX11 Maxed and 2xaa I get 60fps V Sync locked. Setting 4xaa can knock a third off my FPS.

 

You have no idea what you are talking about so it may be an idea to STFU.

Both NV (Soon.) and ATI support DX11 how MS specified it needs to be supported. No more no less. ATI have been doing tessellation since the 360 xenos GPU was designed, whilst this is Nvidia's first shot, so it reamins to be seen how well they implement it.

 

QFT.

At the end of the day, they both do the same thing, but by slightly different methods to come to the same conclusion.

Nvidia use 16 scalable engines across their shader clusters, which means they can probably choose how many to use in order to maximise performance.

ATI use a dedicated Tessellator seperate from the clusters.

Performance will be compared when we see proper numbers, that's all we know at this moment in time.

 

stalker call of pripyat uses tessellation and my videocard has no issue maxing every one of the settings and still maintaining a playable fps. I really dont know to what extent stalker uses it though

 

and thats what i said wdf maybe you dont uke2:, so what if ati uses it for years, nv made it more advanced and it doesn't have to disable shaders for tessellation since its part of it. Im not saying ati has to scarifies it or anything like that, its just not so complex as nvidia's.

 

If you would actually run your 5850 in an environment where it can breathe and not be bottlenecked by your 3.1ghz C2Q, you'd see that the tessellation performance is better than you think.

I can hear your GPU screaming from over here. "Help, help, my owner does not know what he is doing" :wanker:

 

Not that much, stalker games would look awesome if they'd implemented it on brick walls and such. I don't really care about how well the first gen can do tesselation, if buy a new gfx card then it's because my 8800GTX is finally becoming weak enough to handle newer games, DX9, 10 or 11.

Even if the first cards can't run it properly in a real game, they still offer an increase in performance which can at least let dev's take the DX9 and DX10 effects to the next step.

A nice example would be Fallout 3, even on DX9, I can't run it maxed out.

Also, when implemented right, wouldn't tesselation render older techniques such as bumpmapping rather obsolete? So basically we'd be looking at the performance of the game without bumpmapping and such, which gives us an increase in performance, which is then decreased again by tesselation.

 

It was not the point I was making,even with a full balanced setup the next gen. games will not render with a satisfactory fps even on your setup with 2 gpus.
Why do you think nvidia dedicated a whole segment of the chip logic for tesselation if it is not so demanding,they know very well where the rt graphics is directed,in a sense,towards implementing the nonrealtime-prerendred rendering tecnology in real time like subdivison surfaces(tesselation)GI,volumetric hair,fur,soft bodies and even raytracing.

I work daily with 3d graphics(prerendered) and see that all the features from prerendered are slowly coming to rt as the proccesing power grow and tesselation is a triangle mesh subdivision again taken from prerendered.

And also regarding the post mentioning the bump mapping:
The bump/normal mapping will not be taken out in the creation process of the real time models by the tesselation cause they work best when combined.
The geometry is tesselated for the major geometric forms to be implemented on the mesh via a displacement map and the bump/normal map is used for shading the finer surface details without the need to create a polyhog geometry.
Again I think that tesselation is the greatest quantum leap forward in realtime in years and that the next or even the gen. after will be ready for full(optimal)tesselation with good frame rates.

 

Why are you banging on about disabling shaders? ATI has dedicated tesselation hardware on the gpu completely separate from them. NVIDIA uses shaders to do the tesselation as it doesn't have dedicated hardware.

 

Bull****.

http://www.nvidia.com/object/gf100.html

 

Why do people keep quoting Nvidia and their new cards, they havent been released yet, personally i think the new cards are going to be rubbish.

im not ATi fan i just buy whats best for me at the time.