AMD V Nvidia by Richard Huddy

[Janooo]

As I said, that depends on both your source geometry, your camera positions, your screen resolution and your AA settings (aside from the fact that there are various ways to estimate polygon size, distance etc, so there will be different tessellation factor estimates depending on the implementation etc yaddayadda).

Too many unknowns to make any kind of useful estimates whatsoever.

Well, you as a developer have maxtessfactor(x) function available. What would be your x value?
If we talk about screen adaptive tessellation, how many pixels would you recommend for triangles to be effective?

 

[Scali]

Well, you as a developer have maxtessfactor(x) function available. What would be your x value?
If we talk about screen adaptive tessellation, how many pixels would you recommend for triangles to be effective?

Well as I already said, the 1 pixel-per-triangle is the ideal value.
However, in practice hardware is not fast enough for this yet, so you just have to tune it to find the best balance between performance and quality.

 

[Janooo]

Well as I already said, the 1 pixel-per-triangle is the ideal value.
However, in practice hardware is not fast enough for this yet, so you just have to tune it to find the best balance between performance and quality.

Wrong answer. Below 8 pixel the tessellation becomes inefficient.

 

[Scali]

Wrong answer. Below 8 pixel the tessellation becomes inefficient.

I don't see how that makes my answer wrong (unless you don't know what an idea is, in the Plato sense).
I do see that you apparently need an AMD slide to try and point this out though...

 

[Janooo]

I don't see how that makes my answer wrong (unless you don't know what an idea is, in the Plato sense).
I do see that you apparently need an AMD slide to try and point this out though...

It does not matter if the slide is from AMD. The point is that 1 pixel triangles are not efficient.
Check the slide #23 here: http://developer.amd.com/gpu_assets/Tessellation Performance.ppsx

At the end we'll see what AMD did with the tessellation.

 

[artonlangy]

Well, let's see...
I'm a professional graphics software engineer, with no ties to any videocard manufacturer. Aside from my professional work, I have open source projects available on the web, which use this sort of technology.
Richard Huddy is a PR guy (not a developer) for one of the largest GPU manufacturers in the world.



I would say: wrong.
In my experience, PR people generally don't have a formal education in graphics or software development (Huddy does not, as far as I know). Their knowledge of the technology is superficial, and they consult their technical colleagues on what to say, and how to say it. Huddy has often been burnt in the past because he tends to cross these boundaries, and in the process either stabs people in the back or at least puts his foot in his mouth.



The simple answer is that tessellation is not currently one of AMD's strong points, so naturally Huddy will try to downplay its significance.
I'm just saying that his scenario is not an accurate representation of reality. Mine is. As I say: look at how Pixar applies tessellation in their RenderMan software. That's where we should be going with tessellation on GPUs: adaptive subdivision based on the polygon's sizes in screen-space.

I could go in detail on how rasterizing, tessellation, NURBS and AA work, and how it all fits together, but you'd have to ask specific questions, because I don't know what you know, what you don't know, and what you want to know. It's a pretty large subject to cover in a single post.
So please: Ask questions if you want to know something. I know pretty much everything about graphics.



I am willing to go as far as this: He is not an expert in the technical field. He consults experts in his company. That is not the same. You could say he's an expert-by-proxy. He is not to be wholly discounted, but he does not exactly have a flawless reputation, and obviously he has an agenda to push.



That is exactly the danger of the situation. People believe him, even if he is wrong.



I know full well where Huddy is trying to go: he's trying to cover up AMD's poor tessellation performance.

I was reading thru this discussion and it seems to me that you are mistaking the difference between Huddys role at ATI/AMD. Isn't his title "Head of Developer Relations", not "Public Relations" It would seem to me they are vastly different. Would you care to reconcile how you made the jump from one to the other?

 

[Idontcare]

Welcome to the forums artonlangy! :thumbsup:

 

[Scali]

It does not matter if the slide is from AMD. The point is that 1 pixel triangles are not efficient.

That's what I said, isn't it?
I said 1-pixel is the ideal, the thing we are aiming for. Referring back to my earlier post:

Ideally (Pixar/RenderMan/reyes rendering) you want to have one triangle for each pixel (or technically less than one, for supersampling/AA reasons, but you get the idea). This is known as 'micropolygons'. You have 'perfect' geometry this way.
So basically you want to avoid undersampling your geometry, because then you will still not get perfectly round objects nearby... But you also don't want to oversample your geometry and fit too many polygons in a single pixel.

So this is what we *want*, our *ideal*.

Then I said:

However, in practice hardware is not fast enough for this yet, so you just have to tune it to find the best balance between performance and quality.

So apparently I am not *using* 1-pixel as a value, because it is not fast enough on current hardware in practice (what *is* fast enough depends on what videocard you're using, what resolution, and what kind of scenes you are rendering, it's not like you can just pin-point a single value that 'just works'... aside from the differences between nVidia and AMD when it comes to tessellation, there's also a significant performance difference between low-end Radeon 5000-series and high-end Radeon 5000-series).
Efficiency is not the reason by the way. As long as you have enough brute force to pull off what you want, the efficiency is really not important.
I mean, if theoretically we have a GPU that can handle a full game with 1-pixel polygons at 60+ fps in the highest resolutions, quality etc... who cares that it would be more efficient with larger polygons? If you go the 'efficiency' route, you shouldn't be using most forms of shading anyway, as just basic single-texturing and per-vertex lighting is far more efficient than all this per-pixel multitextured mumbo-jumbo. It's just that nobody cares about games running at thousands of fps. They want as much eyecandy as they can get, with playable framerates.

 

[Scali]

I was reading thru this discussion and it seems to me that you are mistaking the difference between Huddys role at ATI/AMD. Isn't his title "Head of Developer Relations", not "Public Relations" It would seem to me they are vastly different. Would you care to reconcile how you made the jump from one to the other?

Huddy is a PR guy because as head of developer relations, he's the one always talking about anything devrel-related in the press. About what new games they have been working on, how their latest hardware makes these games even better etc.
Most of his activities are press-related. The rest is directed at game developers, trying to spread the gospel of AMD's Radeons and all that.

 

[Lonbjerg]

Huddy is a PR guy because as head of developer relations, he's the one always talking about anything devrel-related in the press. About what new games they have been working on, how their latest hardware makes these games even better etc.
Most of his activities are press-related. The rest is directed at game developers, trying to spread the gospel of AMD's Radeons and all that.

Fuddy is PR.
He is AMD's bullhorn to posting FUD.
Like when he claimed that NVIDIA paid for devs to use PhysX.
Or that they willfully hobbled PhysX (x87).

He is a mouthpiece, often way of base and in conflict with reality.

 

[Grooveriding]

Huddy is a PR guy because as head of developer relations, he's the one always talking about anything devrel-related in the press. About what new games they have been working on, how their latest hardware makes these games even better etc.
Most of his activities are press-related. The rest is directed at game developers, trying to spread the gospel of AMD's Radeons and all that.

Since you know how he spends his time.

What percent over 50%, since you say most of his activities are pres-related, does he spend on press-related pursuits ?

Then, what percent is spent on dev relations ?

Break it down for us, since you are speaking in absolutes and clearly know what you are saying for fact.

 

[flopper]

quote" Richard Huddy who is pointman for his company’s DevRel team" end quote
http://www.kitguru.net/components/g...e-constant-smell-of-burning-bridges-says-amd/

Huddy define the gaming as and I quote
Overall, the polygon [Triangle - Ed] size should be around 8-10 pixels in a good gaming environment”, said Huddy

That is what he define as a good gaming enviroment.
BUT what he is doing in his interview according to Scali is that AMD is actually limiting gaming production as the AMD cards cant do tessealtion properly as adaptive tesselation.

So a game developer has choices to make, do I make a console game, a PC game and if I make a PC game can I add the detail I want with good fps on current hardware OR upcoming hardware?
Any developer would answer that with, where do I get profit from?
If I make a game that is limited by the card that AMD releases tomorow, why should I add detail as adaptive tesselation and make it a better image quality wise as the only company that can play my game properly is NVidia?
We get games that is now going to be worse in quality than what should have been as AMD which Huddy is trying to cover up with his dogma that their tesselation engine isnt up to adaptive tesselation as it should be.

That is what I oppose, sure I enjoy my 5850, I sure enough will enjoy my upcoming 6970 if it adds some speed, but I so dislike that a game developer is stuck in a moment 22 situation where they simply cant do the game they would like for us due to AMD is limiting their options with their so called optimal tesselation engine that dosnt cut it.

True, current games isnt in need of more tessealtion BUT upcoming wont either as AMD foces the developer to not implement it. Is that any different than Nvidia making sure AA run bad on AMD hardware?
Talk about burning bridges, BUT Huddy burned mine and the PC gamers would wide with that crap if it is what Scali state it is.

 

[artonlangy]

Huddy is a PR guy because as head of developer relations, he's the one always talking about anything devrel-related in the press. About what new games they have been working on, how their latest hardware makes these games even better etc.
Most of his activities are press-related. The rest is directed at game developers, trying to spread the gospel of AMD's Radeons and all that.

Ah, so because he does interviews he is a "PR Guy". Is that like how Steve Jobs, or even Jen-Hsun Huang do interviews so they are just a "PR guy's" as well? Discount the actual work they do based on a few interviews.

 

[Genx87]

Ah, so because he does interviews he is a "PR Guy". Is that like how Steve Jobs, or even Jen-Hsun Huang do interviews so they are just a "PR guy's" as well? Discount the actual work they do based on a few interviews.

He interfaces with the public, the consumer, and addresses any questions or concerns about AMDs product line. What do you define as a job responsiblity for somebody within a "public relations" position?

 

[Idontcare]

Welcome to the forums flopper! :thumbsup:

 

[Grooveriding]

He interfaces with the public, the consumer, and addresses any questions or concerns about AMDs product line. What do you define as a job responsiblity for somebody within a "public relations" position?

By that definition most any high level employee in an accomplished corporation is in public relations. As all are called on to interface with their target market at some point or another.

As he said, is JHH pr because he gets up on stage and discusses nvidia cards several times a year ?

God, this topic is abysmal. It's devolved into fanboys back and forth arguing over pointless semantics and has been ridden wildly off-topic into conjecture and finger-pointing with absolutely nothing concrete available to anyone about the original subject.

These new Radeons need to get released and the the dust needs to settle so this garbage is over.

 

[Genx87]

By that definition most any high level employee in an accomplished corporation is in public relations. As all are called on to interface with their target market at some point or another.

As he said, is JHH pr because he gets up on stage and discusses nvidia cards several times a year ?

God, this topic is abysmal. It's devolved into fanboys back and forth arguing over pointless semantics and has been ridden wildly off-topic into conjecture and finger-pointing with absolutely nothing concrete available to anyone about the original subject.

These new Radeons need to get released and the the dust needs to settle so this garbage is over.

In a way JHH is a PR guy. However his primary responsibility is not to answer questions about Nvidia product lines from consumers. He interfaces for the company in terms of long term vision and financial information. Huddy interfaces with customers, consumers, developers, and answers questions regarding the product line at AMD. He is definately a PR guy.

If you think this is a bad topic by all means dont participate in it.

 

[Lonbjerg]

Fuddy is PR, look at his "interviews":
http://www.bit-tech.net/bits/interviews/2010/01/06/interview-amd-on-game-development-and-dx11/
http://news.softpedia.com/news/AMD-NVIDIA-Promotes-PhysX-Games-Through-Bribery-136949.shtml
http://www.kitguru.net/components/g...e-constant-smell-of-burning-bridges-says-amd/

He speaks more of NVIDIA than AMD's own products...says it all dosn't it?
(And often using false claims as "facts")

Infact he speaks more to the press than PR guys at NVIDIA (eg. Ken Brown)

 

[Scali]

Ah, so because he does interviews he is a "PR Guy". Is that like how Steve Jobs, or even Jen-Hsun Huang do interviews so they are just a "PR guy's" as well? Discount the actual work they do based on a few interviews.

I would certainly say that Steve Jobs and Jen-Hsun Huang have a a pronounced public function in their company yes.
They 'sell' the products to the audience. Especially Jobs is a master at it. And most of the time it is Jobs himself making public statements, answering questions, doing interviews etc.
It may not be their ONLY function, or even their PRIMARY function, but they are doing PR, and a lot of it.

 

[Seero]

quote" Richard Huddy who is pointman for his company’s DevRel team" end quote
http://www.kitguru.net/components/g...e-constant-smell-of-burning-bridges-says-amd/

Huddy define the gaming as and I quote
Overall, the polygon [Triangle - Ed] size should be around 8-10 pixels in a good gaming environment”, said Huddy

That is what he define as a good gaming enviroment.
BUT what he is doing in his interview according to Scali is that AMD is actually limiting gaming production as the AMD cards cant do tessealtion properly as adaptive tesselation.

So a game developer has choices to make, do I make a console game, a PC game and if I make a PC game can I add the detail I want with good fps on current hardware OR upcoming hardware?
Any developer would answer that with, where do I get profit from?
If I make a game that is limited by the card that AMD releases tomorow, why should I add detail as adaptive tesselation and make it a better image quality wise as the only company that can play my game properly is NVidia?
We get games that is now going to be worse in quality than what should have been as AMD which Huddy is trying to cover up with his dogma that their tesselation engine isnt up to adaptive tesselation as it should be.

That is what I oppose, sure I enjoy my 5850, I sure enough will enjoy my upcoming 6970 if it adds some speed, but I so dislike that a game developer is stuck in a moment 22 situation where they simply cant do the game they would like for us due to AMD is limiting their options with their so called optimal tesselation engine that dosnt cut it.

True, current games isnt in need of more tessealtion BUT upcoming wont either as AMD foces the developer to not implement it. Is that any different than Nvidia making sure AA run bad on AMD hardware?
Talk about burning bridges, BUT Huddy burned mine and the PC gamers would wide with that crap if it is what Scali state it is.

As a game developer, video card's limitation is the least to worry about. As long as 6xxx series can play the game, then there are no problems. If consumer's video card can't play at maximum quality out of the game, then it is the consumer's problem.

If a game runs 5 FPS on one series but run 50+ on any other cards, then it is clear that series has problems. This is not the case on AMD cards. Fermi series does tessellation well. Fermi does better than 5xxx(probably 6xxx IMO) series on games that have heavy tessellation. Is Nvidia at fault for selling a product that does something good? or is AMD at fault for selling a product that does something not as good?

In my opinion, the ball should not be at developers' court, nor Nvidia's court either. AMD can't force and shouldn't complain when their product can't play as well as its competitors under some situations. AMD knew about this problem, at least they knew Fermi does tessellation better under some situations. They can spend engineers to fix it, or start generating reasons why those situations ain't practical. Which one of these options benefits gaming as a whole, and which one cost less?

 

[Scali]

But let's not get carried away... The majority of gamers don't even have DX11 hardware in the first place, let alone the latest and most high-end cards.
So in practice, most gamers will either play without tessellation anyway, or at lower detail settings.

 

[Gideon]

I still don't quite understand how can anyone accuse AMD of neglegting tesselation.

Considering that a sort-of-mid-range card as 6870 is noticeably more efficient in tesselation than last years high-end, Cayman should end up being significantly faster than the 5870.

Let's also not forget, that we really haven't seen any games using anywhere near as much tesselation as Heaven or Stonegiant benchmark. IMHO it would really be interesting to see how a GTX 480 would compare to 69xx series when a lot more is going on than simple fly-by demo. I'd really be impressed if GTX 480 tesselation efficency doesn't drop, considering that the hardware, in nVidias case, has to do a lot of other stuff as well.

 

[Genx87]

Didnt the 5770 has similar tesselation performance as a 5870 due to the lack of scaleability in AMDs tesselator in the 5000 series?

 

[Scali]

I still don't quite understand how can anyone accuse AMD of neglegting tesselation.

You would, if you were to look at a chart that also included nVidia's GPUs.
This chart is not very impressive. At its peak, it's twice as fast as the previous gen, which still isn't very fast... and with higher tessellation factors (where the real point of tessellation is) it seems to drop back down quickly, to pretty much the same poor speeds of its predecessor.

Take a look at this for example: http://www.geeks3d.com/20100826/tessmark-opengl-4-gpu-tessellation-benchmark-comparative-table/ ... You can figure out for yourself where a 1.5-2x increase would put AMD... Neglect? I think a solid yes.

 

[badb0y]

You would, if you were to look at a chart that also included nVidia's GPUs.
This chart is not very impressive. At its peak, it's twice as fast as the previous gen, which still isn't very fast... and with higher tessellation factors (where the real point of tessellation is) it seems to drop back down quickly, to pretty much the same poor speeds of its predecessor.

Take a look at this for example: http://www.geeks3d.com/20100826/tessmark-opengl-4-gpu-tessellation-benchmark-comparative-table/ ... You can figure out for yourself where a 1.5-2x increase would put AMD... Neglect? I think a solid yes.

Considering the 6800 is not high end part for AMD, I think it's good that their new mainstream product atleast matches their flagship from last year. Also we have no idea what Cayman will do but from the 5800, this is a solid upgrade you can't take that away from them.