Is AMD's tessellation still way behind Nvidia?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Carfax83

Diamond Member
Nov 1, 2010
6,841
1,536
136
As for the OP, its not gonna make much a difference. I believe by the time tessellation and the new consoles make an impact, AMD will be caught up with (or surpass) nvidia in their tessellation capabilities

Right, and NVidia is going to be sitting still twiddling their thumbs while AMD catches up with or surpasses them :D

AMD still has a long way to go before they're up to par with NVidia..
 

f1sherman

Platinum Member
Apr 5, 2011
2,243
1
0
The interesting part is that evil Nvidia is not using their advantage at extreme tessellation to bully AMD,

which can't be said about AMD and their GPGPU advantage.
Global Illumination anyone? So much oomph used and for literally nothing.

I wonder if that changes with upcoming TWIMTBP titles or NV continues the gentlemen's fight
 

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
i cannot agree that tessellation is a form of compression. I think your looking at it completely the wrong way.

Non tesselation: send tons of tiny triangles
Tesselation: send a few large triangles, tell the GPU to turn those into many small triangles.

It is literally compression

The interesting part is that evil Nvidia is not using their advantage at extreme tessellation to bully AMD,

which can't be said about AMD and their GPGPU advantage.
Global Illumination anyone? So much oomph used and for literally nothing.

I wonder if that changes with upcoming TWIMTBP titles or NV continues the gentlemen's fight

Every major recent TWIMTBP title (Metro:LL, Crysis 2, Batman) have all used tesselation. Maybe Nvidia could use their leverage better if TWIMTBP actually had many games coming out....
 
Last edited:

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
Non tesselation: send tons of tiny triangles
Tesselation: send a few large triangles, tell the GPU to turn those into many small triangles.

It is literally compression



Every major recent TWIMTBP title (Metro:LL, Crysis 2, Batman) have all used tesselation. Maybe Nvidia could use their leverage better if TWIMTBP actually had many games coming out....

Used tessellation yeah because it's a standard DX11 feature. They have not flooded the game with Tessellation to cripple AMD cards in any way.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
The problem is we are only seeing a halfway implementation of tessellation. We are seeing it used to add detail. It can also be used to reduce detail when not needed and save compute resources when not needed. Once the consoles are using DX11 we should start to see better use of tessellation and see the performance gains that can come from it.

The interesting part is that evil Nvidia is not using their advantage at extreme tessellation to bully AMD,

which can't be said about AMD and their GPGPU advantage.
Global Illumination anyone? So much oomph used and for literally nothing.

I wonder if that changes with upcoming TWIMTBP titles or NV continues the gentlemen's fight

LOL Please use the appropriate sarcasm tags. :D
 

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
Used tessellation yeah because it's a standard DX11 feature. They have not flooded the game with Tessellation to cripple AMD cards in any way.

Metro:LL has a very significant amount of tessellation, just because an AMD card doesn't perform worse than an Nvidia card does not mean it is not using tessellation significantly?

Flooding games with tessellation cripples AMD and Nvidia, just cripples Nvidia less. Not good for any customer in any way shape or form. No thanks!
 

Keysplayr

Elite Member
Jan 16, 2003
21,219
54
91
Metro:LL has a very significant amount of tessellation, just because an AMD card doesn't perform worse than an Nvidia card does not mean it is not using tessellation significantly?

Flooding games with tessellation cripples AMD and Nvidia, just cripples Nvidia less. Not good for any customer in any way shape or form. No thanks!

Same goes for practically every other graphics feature available. This isn't reserved just for tesselation.

Change the bolded word above with any of the following you'd like.

AA
DoF
Dynamic shadows and lighting
High textures
Global illumination

Keep on going.
 

Red Hawk

Diamond Member
Jan 1, 2011
3,266
169
106
Right, and NVidia is going to be sitting still twiddling their thumbs while AMD catches up with or surpasses them :D

AMD still has a long way to go before they're up to par with NVidia..

Maybe, but as pointed out in Anandtech's Radeon HD 6900 review, Nvidia is a little restricted in what it can do. Its tessellators are directly part of each SM in its graphics chips. It can't put more tessellation muscle into a graphics chip than its general speed; AMD can simply add more geometry engines if need be since they are completely separate from its stream processors.

What does tessellation give you that simply drawing more triangles wont :).

Tessellation is simply drawing more triangles.
 

cmdrdredd

Lifer
Dec 12, 2001
27,052
357
126
Metro:LL has a very significant amount of tessellation, just because an AMD card doesn't perform worse than an Nvidia card does not mean it is not using tessellation significantly?

Flooding games with tessellation cripples AMD and Nvidia, just cripples Nvidia less. Not good for any customer in any way shape or form. No thanks!

Way to misrepresent my comments there. Read what I said, it uses tessellation because it's a DX11 feature. Some people here would have to believe they only use it because it makes AMD look bad. If that was the goal, they would force certain things and flood the screen with everything AMD cannot do well. The most they do is add physx and help get the game running optimally on their hardware.

What does tessellation give you that simply drawing more triangles wont :).

Depth...show me a game that doesn't use tessellation that makes brick walls with uneven surfaces look to have depth and not flat textures. I have yet to see it.
 

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
Way to misrepresent my comments there. Read what I said, it uses tessellation because it's a DX11 feature. Some people here would have to believe they only use it because it makes AMD look bad. If that was the goal, they would force certain things and flood the screen with everything AMD cannot do well. The most they do is add physx and help get the game running optimally on their hardware.

Well, Nvidia games tend to use tessellation more than usual, but to each their own... Nvidia and AMD both "nudge" developers to use their own exclusive features, devs won't do anything drastic, they'll just make small implementations that are more favorable for one vendor than the other.

Depth...show me a game that doesn't use tessellation that makes brick walls with uneven surfaces look to have depth and not flat textures. I have yet to see it.

Displacement mapping =/= tessellation, it makes it easier but you can do displacement mapping without tessellation.
 

itsmydamnation

Diamond Member
Feb 6, 2011
3,076
3,908
136
Depth...show me a game that doesn't use tessellation that makes brick walls with uneven surfaces look to have depth and not flat textures. I have yet to see it.

That wasn't the question , nor does it contract my point which is tessellation is nothing but compression.
 

Pottuvoi

Senior member
Apr 16, 2012
416
2
81
It can also be used to reduce detail when not needed and save compute resources when not needed.
This is only possible for some assets and only with a very careful planning.
If used traditionally it cannot reduce vertexes, nor it can be used to truly change the topology of surface. (ie. chain links.)
For this vertex reducing LoD method is needed.

If used with geometry maps one can use them for impostor rendering and such, problem is that it might not be feasible in terms of speed.
http://voxels.blogspot.fi/2012/01/procedural-geometry-maps-tessellation.html

Another problem is the low maximum tesselation factor of 64, which reduces the usage for non pre-tesselated surfaces.
IE, you cannot create decent mountain from single quad or planet from a cube.

http://voxels.blogspot.fi/2012/01/procedural-geometry-maps-tessellation.html
 

sontin

Diamond Member
Sep 12, 2011
3,273
149
106
Non tesselation: send tons of tiny triangles
Tesselation: send a few large triangles, tell the GPU to turn those into many small triangles.

It is literally compression

Lol, this is not compression, this would be expansion. D:

Every major recent TWIMTBP title (Metro:LL, Crysis 2, Batman) have all used tesselation.

So it's bad to use a standard DX11/OpenGL feature, now? :'(
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
This is only possible for some assets and only with a very careful planning.
If used traditionally it cannot reduce vertexes, nor it can be used to truly change the topology of surface. (ie. chain links.)
For this vertex reducing LoD method is needed.

If used with geometry maps one can use them for impostor rendering and such, problem is that it might not be feasible in terms of speed.
http://voxels.blogspot.fi/2012/01/procedural-geometry-maps-tessellation.html

Another problem is the low maximum tesselation factor of 64, which reduces the usage for non pre-tesselated surfaces.
IE, you cannot create decent mountain from single quad or planet from a cube.

http://voxels.blogspot.fi/2012/01/procedural-geometry-maps-tessellation.html

LOL I'm not trying to say we can take simple procedural shapes and turn them into full blown hidef models using tessellation. You can use different levels of tessellation at different draw distances to create a LOD effect, though.
 

Pottuvoi

Senior member
Apr 16, 2012
416
2
81
LOL I'm not trying to say we can take simple procedural shapes and turn them into full blown hidef models using tessellation. You can use different levels of tessellation at different draw distances to create a LOD effect, though.
True, but for most objects you also need different base meshes for different distances as they have too drastic changes in topology. (backpacks, mouth etc.)
Perhaps then you could use tesselation to hide some of the popping due to the LoD switch.
 
Last edited:

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
Lol, this is not compression, this would be expansion. D:(

Oh my god, the semantics. The compression is taking a lot of polygons and compressing them into a basic polygonal shape which, with additional algorithmic work and extra information, can be decompressed into a very detailed model.
 

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
I think you did not understand Tessellation...

Try again.

coarse_model.jpg


Rough image (low polygons) on the right goes into GPU, algorithmic smoothing is done (turned into MANY polygons essentially), displacement mapping is applied, original ROUGH model is decompressed into a DETAILED final model.

In its most basic form, tessellation is a method of breaking down polygons into finer pieces. For example, if you take a square and cut it across its diagonal, you’ve “tessellated” this square into two triangles.

These triangles are compressed into a square, sent to GPU, decompressed into two triangles.
 

Jaydip

Diamond Member
Mar 29, 2010
3,691
21
81
Try again.

coarse_model.jpg


Rough image (low polygons) on the right goes into GPU, algorithmic smoothing is done (turned into MANY polygons essentially), displacement mapping is applied, original ROUGH model is decompressed into a DETAILED final model.



These triangles are compressed into a square, sent to GPU, decompressed into two triangles.

Why compressed? Area T1+ Area T2=Area Square, there is no compression here.I believe what it is referring to is, you now have more details i.e previously you had 4 vertices but now have 6.
 

sontin

Diamond Member
Sep 12, 2011
3,273
149
106
There is nothing to decompress because the bash mesh has no further information.
The real information comes from the displacement map.

Tessellation creates new information and new data on the GPU and that is no even close to decompression which only extract information from the compressed file or whatever.
 

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
Why compressed? Area T1+ Area T2=Area Square, there is no compression here.I believe what it is referring to is, you now have more details i.e previously you had 4 vertices but now have 6.

Sweet mother of god, compression does not come from adding AREA. My god, you send the GPU a small bit of information in the form of a rough map, the GPU decompresses this information into a LARGE amount of information in the form of a smooth map. You send it a small amount of data, it comes out with a large amount of data.

There is nothing to decompress because the bash mesh has no further information.
The real information comes from the displacement map.

Tessellation creates new information and new data on the GPU and that is no even close to decompression which only extract information from the compressed file or whatever.

GPU takes a small bit of information and makes it into a large amount of information in the manner that the graphics dev intended. This is the textbook definition of compression.

In computer science and information theory, data compression, source coding, or bit-rate reduction involves encoding information using fewer bits than the original representation.

Your end product, the smooth map, has MORE BITS THAN THE ORIGINAL REPRESENTATION.

This is what compression is.
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,605
6
81
Shouldn't it be vice versa? To take a large amount of data and make it into a small amount is compression. But what you said is the other way around:
GPU takes a small bit of information and makes it into a large amount of information in the manner that the graphics dev intended.
 

sushiwarrior

Senior member
Mar 17, 2010
738
0
71
Shouldn't it be vice versa? To take a large amount of data and make it into a small amount is compression. But what you said is the other way around:

Decompression implies that compression was done on the original data being sent. The fact of the matter is simply that tessellation is the equivalency of compression for 3d models. This semantics bulls**t is irrelevant. Tessellation is done to save bandwidth/model size, hence you can make a models head look perfectly smooth (for a baldie :D) instead of looking like a hexagon.
 

Jaydip

Diamond Member
Mar 29, 2010
3,691
21
81
Sweet mother of god, compression does not come from adding AREA. My god, you send the GPU a small bit of information in the form of a rough map, the GPU decompresses this information into a LARGE amount of information in the form of a smooth map. You send it a small amount of data, it comes out with a large amount of data.



GPU takes a small bit of information and makes it into a large amount of information in the manner that the graphics dev intended. This is the textbook definition of compression.



Your end product, the smooth map, has MORE BITS THAN THE ORIGINAL REPRESENTATION.

This is what compression is.

You are completely confused at this moment.Take a moment to read what you are typing.