AMD Radeon HD 6970 already benchmarked? Enough to beat GTX480 in Tesselation?

[Scali]

That's total BS. It improves IQ not performance. You have to pay for better IQ.

Uhhhh, no.
http://msdn.microsoft.com/en-us/library/ff476340(VS.85).aspx

Tessellation:

Saves lots of memory and bandwidth, which allows an application to render higher detailed surfaces from low-resolution models. The tessellation technique implemented in the Direct3D 11 pipeline also supports displacement mapping, which can produce stunning amounts of surface detail.
Supports scalable-rendering techniques, such as continuous or view dependent levels-of-detail which can be calculated on the fly.
Improves performance by performing expensive computations at lower frequency (doing calculations on a lower-detail model). This could include blending calculations using blend shapes or morph targets for realistic animation or physics calculations for collision detection or soft body dynamics.
The Direct3D 11 pipeline implements tessellation in hardware, which off-loads the work from the CPU to the GPU. This can lead to very large performance improvements if an application implements large numbers of morph targets and/or more sophisticated skinning/deformation models.

 

[Idontcare]

And who us Andy Grove?

Are you joking me? Ordinarily I'd assume that was a joke post but you didn't use any smilies. Not knowing who Andy Grove is/was is like saying you never heard of Bill Gates or Steve Jobs.

 

[Arkadrel]

The way I see it... your not getting more performance from it. It comes with a performance hit, but thats okay, we just need bigger and bigger tessellation units as time goes on.


Lets say your CPU + normal grafics card without gpu.
Your CPU is stressed, so you pass off some of the workload to a GPU.

Same thing with the GPU your takeing some of its work and makeing the tessaltor unit do it instead. Will this mean one day we ll have seperate tessellation cards? or one day the tesellator will be come bigger than the gpu and maybe get put on its own chip beside the gpu? how knows..

anyways current tessellation isnt applied like that in games... so its a moot point.
Just because something *can* be used 1 way, doesnt mean that it will, and its currently not being used like that.

In todays games, when you turn on tessellation, you dont get faster fps, you get slower.

@Idontcare,
didnt you forget JHH? and Dirk Meyers? just kidding btw.

 

[Gloomy]

I'm going to go out on a limb here and say that DX11 is not being used that way because 90% of the DX11 userbase cannot use it that way. Take that as you will. (not pointing the finger or anything)

()

 

[Skurge]

Are you joking me? Ordinarily I'd assume that was a joke post but you didn't use any smilies. Not knowing who Andy Grove is/was is like saying you never heard of Bill Gates or Steve Jobs.

You have to remember I'm only 19, thanks anyway.

 

[Idontcare]

You have to remember I'm only 19, thanks anyway.

Ah, sorry, I did not know that and I make it a point to never assume people are "young" since that seems to be taken as an insult more often than not.

Andy Grove is a legend, like Jack Welch, in how he managed Intel. "Only the paranoid" survive. He was one of its original founders.

(JHH is in that same league, but Meyers is not, he didn't start a company from scratch, Jerry Sanders otoh)

 

[Skurge]

Ah, sorry, I did not know that and I make it a point to never assume people are "young" since that seems to be taken as an insult more often than not.

Andy Grove is a legend, like Jack Welch, in how he managed Intel. "Only the paranoid" survive. He was one of its original founders.

(JHH is in that same league, but Meyers is not, he didn't start a company from scratch, Jerry Sanders otoh)

Just a read up a bit about him. He did an amazing job at Intel. You really think Meyers has the potential to do something similar?

 

[Arkadrel]

Not sure about Meyers.... I have my doubts about JHH.

Bill gates was a great business man, spots value.
Steve job just has a eye for potential, makes things happend.

These are icons of are times really..

 

[Lonbjerg]

In todays games, when you turn on tessellation, you dont get faster fps, you get slower.

False, but why I am not surprised:

http://www.pcgameshardware.com/aid,...h-Interview-What-DirectX-11-is-good-for/News/

1. One of the main features of DirectX 11 API is hardware tessellation. If you're utilizing this feature, which visual improvements can we expect (e.g. water, NPCs, environment)? Some implementations distort the textures cause of the additional mesh, have you noticed and did you consider this problem?

This feature will be enabled automatically if your hardware has DX11 capabilities. We use tessellation for the Civilization V terrain, which adjusts the mesh's subdivision of the terrain as the user zooms in and out. Not only does it add detail, but terrain tessellation makes the game measurably faster on both Nvidia and AMD hardware (as much as 30% in some cases).
​

Now can you declare that "argument" for debunked and move on?

 

[happy medium]

False, but why I am not surprised:

http://www.pcgameshardware.com/aid,...h-Interview-What-DirectX-11-is-good-for/News/

1. One of the main features of DirectX 11 API is hardware tessellation. If you're utilizing this feature, which visual improvements can we expect (e.g. water, NPCs, environment)? Some implementations distort the textures cause of the additional mesh, have you noticed and did you consider this problem?

This feature will be enabled automatically if your hardware has DX11 capabilities. We use tessellation for the Civilization V terrain, which adjusts the mesh's subdivision of the terrain as the user zooms in and out. Not only does it add detail, but terrain tessellation makes the game measurably faster on both Nvidia and AMD hardware (as much as 30% in some cases).
​

Now can you declare that "argument" for debunked and move on?

That just made me want to upgrade my gpu. Thanks or funds are in short supply.

 

[Idontcare]

Just a read up a bit about him. He did an amazing job at Intel. You really think Meyers has the potential to do something similar?

Yes. Keyword here is potential.

The boundary conditions are entirely different though, Andy Grove of the 1980's and 1990's never had to compete against and outmaneuver the Intel of 2010 while constrained by the resources of AMD of 2010 like Meyers has to.

Most folks would be happy if Meyers can just get AMD back to being a non-profit organization

 

[blastingcap]

Yes. Keyword here is potential.

The boundary conditions are entirely different though, Andy Grove of the 1980's and 1990's never had to compete against and outmaneuver the Intel of 2010 while constrained by the resources of AMD of 2010 like Meyers has to.

Most folks would be happy if Meyers can just get AMD back to being a non-profit organization

part of the recent decline in AMD's CPU division is due to intel's dirty payola tactics which deprived them of potential revenue that could have gone into R&D

 

[Gloomy]

And a big part of it is that they never did have an answer to the Core 2 when it was pertinent. I'm not expecting them to catch up for some time yet, I have hope for BD but honestly, it's going to be a few years before AMD is truly competitive again...

 

[AnandThenMan]

It's Dirk Meyer no S.

 

[Nemesis 1]

ttttttttttttt

 

[Scali]

Go back to orginal statement were I was steering another poster to the orginal Vista specs. Why do I need link .

You need a link to prove your allegations.

LOL you and I read differantly that article has nothing to do with tess . Intel didn't even care about tess on that generation of IGP . It wasn't a gaming IGP . This was about aero.

I never claimed otherwise, re-read what I said.

DO you work for any Hardware companies

No I don't, and that question ha been asked WAY too many times. I find it deeply insulting, and I want actions to be taken against these people.

You should come over to beyond 3D and play

I did that in the past, they don't want to play anymore.

 

[AnandThenMan]

Why is it insulting to ask if someone works for a hardware company? What did I miss?

Back on topic, lack of credible leaks on the 6970 is frustrating to me.

 

[Scali]

Why is it insulting to ask if someone works for a hardware company? What did I miss?

You missed a lot, apparently.
Who I work for is not relevant in a technical discussion.
The reason why people keep asking is to try and discredit me and my arguments, by trying to imply that I'm being paid to post my 'opinions'. And as such, you shouldn't put any value in these opinions (or even technical facts, even when I provide direct links and quotes from Microsoft or other authoritative sources. Remember, DirectX 11 is Microsoft's standard. The links I posted are directly from the source of DX11 and where tessellation came from. More reliable than anything from nVidia or AMD even).
In other words: it's character assassination.

Ironically enough, these same people take the words of eg Richard Huddy as truth, while in his case it is pretty clear that he is working for AMD, and that marketing and promoting AMD's products is his job, literally. He will always spin things in a way to make AMD's hardware look as good as possible.

I am not going to tell you who I work for. I am going to say (as I said before), that the company I work for is not Intel, AMD or nVidia.
While my company makes hardware, we don't make consumer hardware, and we certainly don't make videocards or anything like that. Aside from that, I work in the software department, I am not directly involved with the hardware development itself.
So my employer has absolutely nothing to do with anything related to nVidia, AMD, Intel or other consumer hardware that we may be discussing on this forum.

 

[PingviN]

Whaaa? Wait, are we now really trying to downplay tesselation as a performance improving feature? Properly implemented tesselation can makes things prettier aswell as adding performance, that should be clear by now. I understand why there is room for a discussion about how tesselation is quite some time away from broader implementation, but that doesn't mean it sucks or is a resource hog. Metro 2033 is not tesselation used right. AvP is better at it, but it still uses tesselation only for game characters instead of using it on larger areas (such as terrain).
Tesselation = great when used properly.
Tesselation = not so great when used poorly.

For now, AMDs tesselation performance is adequate, it probably wont be in a few years, but then again, there is a new architecture coming up in 2011, pretty sure there will be a scalable tesselator included.

 

[blastingcap]

Whaaa? Wait, are we now really trying to downplay tesselation as a performance improving feature? Properly implemented tesselation can makes things prettier aswell as adding performance, that should be clear by now. I understand why there is room for a discussion about how tesselation is quite some time away from broader implementation, but that doesn't mean it sucks or is a resource hog. Metro 2033 is not tesselation used right. AvP is better at it, but it still uses tesselation only for game characters instead of using it on larger areas (such as terrain).
Tesselation = great when used properly.
Tesselation = not so great when used poorly.

For now, AMDs tesselation performance is adequate, it probably wont be in a few years, but then again, there is a new architecture coming up in 2011, pretty sure there will be a scalable tesselator included.

AMD claims scalable tessellator in by November in Cayman, not 2011. The "scalability and off-chip buffering" of the tessellator was on an AMD slide. Go look for it or ask around if you are really curious. http://forums.anandtech.com/showpost.php?p=30642607&postcount=1

 

[beserker15]

Yeah, I also think people should not downplay the benefits of tesselations. Especially in the future. The ability to get awesome image quality of very low resolution textures is pretty awesome. Isn't the problem right now however is that we are stuck in between two methods of rendering a good image? High res textures and shaders versus high tesselations and low res textures? If a developer goes full throttle with DX11, would that not mean that the developers would need to basically make two games (one with high res textures, shaders, and one with tesselations) or else DX10 and lower cards will make the game look worse than Quake 3?

We are stuck in a middle ground now right? Where games will use low res models + tessellations but keeps the high resolution backgrounds and shaders yielding small benefits. Or not even using it right and have high resolution textures and shaders on everything while also adding tesselations in as a feature, making everything run worse?

For the technical people out there, I guess my question is...is it easy to implement a truly advance DX11 game right now, but also make it so it doesn't look like crap on non-dx11 hardware once tesselations is turned off?

 

[Scali]

AMD claims scalable tessellator in by November in Cayman, not 2011. The "scalability and off-chip buffering" of the tessellator was on an AMD slide. Go look for it or ask around if you are really curious. http://forums.anandtech.com/showpost.php?p=30642607&postcount=1

I find those two statements extremely vague... 'Scalable'? Yea, everything is scalable. Some things are just very poorly scalable, such as the current implementation from AMD. Are they admitting that it's pretty much not scalable?
And how scalable will the new implementation be? Why will it be scalable? What does it mean?

As for 'off-chip buffering'? In the context of a tessellator, I don't get this at all. The idea of tessellation is to generate triangles on-the-fly, and pipeline the lot. You don't want to write them back into a temporary buffer, and read them back later. If you're going to do that, you may as well just store them into memory directly, and skip the tessellation altogether. With a simple cache you can also do 'off-chip buffering' of the geometry that way. Then again, geometry is already buffered on-chip.

 

[Scali]

Yeah, I also think people should not downplay the benefits of tesselations. Especially in the future. The ability to get awesome image quality of very low resolution textures is pretty awesome. Isn't the problem right now however is that we are stuck in between two methods of rendering a good image? High res textures and shaders versus high tesselations and low res textures? If a developer goes full throttle with DX11, would that not mean that the developers would need to basically make two games (one with high res textures, shaders, and one with tesselations) or else DX10 and lower cards will make the game look worse than Quake 3?

Not entirely.
Currently games already need various levels of geometry detail, to handle different levels of hardware performance, and to scale objects down in the distance.
So developers already need to make a pretty large variety of geometry detail, textures, shaders and all that (if you look closely you can see objects popping into less/more detail in most games, even Crysis).
In DX11 mode, they can just reduce it to using one set of models, textures and shaders, and have the hardware scale it up and down automatically.
So it can probably be combined with backward-compatible DX9/DX10/console versions very easily.

If you start fresh with a new game, you could approach it from the other way: you start by making the models for tessellation, and then you precalc the tessellation for the models at a few different detail levels to generate the static detail levels for DX9/DX10/consoles.
I think this is the approach that developers will be taking in the future.

The PolyBump technology in Crysis is very similar, it also allows them to scale the detail up and down quite easily. Tessellation will make it pretty much seamless (you can just scale your displacement maps, and adjust your tessellation factors).

 

[Idontcare]

I find those two statements extremely vague... 'Scalable'? Yea, everything is scalable. Some things are just very poorly scalable, such as the current implementation from AMD. Are they admitting that it's pretty much not scalable?
And how scalable will the new implementation be? Why will it be scalable? What does it mean?

I saw that marketing slide and did the forehead slap thing...it soooo reminded me of the marketing aspect of the "double vias" saga.

Take something that has always been true (to some degree) and claim it to be a wholly new feature of a new product line.

At the end of the day however we really don't care how AMD marketing team leverages the terminology in their slides, we really care about how their engineers leveraged the techniques in the formulation of their products.

I'm vastly less interested in what scalability means to the marketing dept, much more interested in what it meant to the guys who engineered the chip. They are the ones we need to be worried about knowing what needs to be done. Its the same way with AMD and bulldozer...who cares if AMD marketing butchers the CMP/CMT analogies and performance estimates, what really matters is whether or not the design engineers knew what they were doing.

 

[Scali]

At the end of the day however we really don't care how AMD marketing team leverages the terminology in their slides, we really care about how their engineers leveraged the techniques in the formulation of their products.

Yea, but that's what bothers me.
When nVidia revealed their Fermi architecture, they didn't just say "We have 'scalable' tessellation".
No, they just immediately went down into the technical details, and then just slapped the marketing label "PolyMorph Engine" on it.

AMD hasn't gone into detail yet, nor have they even bothered to figure out a nice marketing label for whatever it is they may have cooked up.

what really matters is whether or not the design engineers knew what they were doing.

Exactly, I hope we get some real info on what they've been doing regarding tessellation soon, and even better: some benchmarks as a proof-of-concept of their efforts.