City Of The Future Tessellation demo

[ugaboga232]

It is really only since the 5XXX series do we have full utilization of DX10, DX11 effects are pretty shallow compared to what it could do. Crysis can be hacked to use the very high dx10 locked features in dx9 mode, so that is one example against you.

Consoles really do dictate what developers do. What's stopping someone from making this tech demo into reality? Its beautiful so why don't developers make it.

Also most DX11 titles don't use DX11 for truly new effects outside of some games.

Hint: Has to do with profits.

 

[blastingcap]

I guess you are convinced of this, no matter how many examples I will produce of the opposite. Game developers have always adopted new technology early on. They never choose to cater to the lowest common demoninator. They choose to add optional effects for whatever new hardware there is. This practice is as old as gaming itself.
It's pointless trying to discuss this with you, because you simply will not see the facts.

I heard NV paid Crytek $2MM for TWIMTBP rights, so if there is one game coming out in 2011 that has a chance of tessellating so hard that even NV's GPUs hurt, let alone AMD's, then it's Crysis 2. Oh and I expect Crysis 2 to support GPU PhysX as well. Entirely by coincidence of course.

Aside from that, I have difficulty imagining any other game pushing extreme tessellation in 2011.

 

[Grooveriding]

"I guess you are convinced of this, no matter how many examples I will produce of the opposite. Game developers have always adopted new technology early on. They never choose to cater to the lowest common demoninator. They choose to add optional effects for whatever new hardware there is. This practice is as old as gaming itself.
It's pointless trying to discuss this with you, because you simply will not see the facts."

Lolwut? Seriously, only a few developers (Crytek is one of the more noteworthy) use new technology to such a level as shown in this tech demo or really any tech demo. That is why its a tech demo and not a real engine that can be used for consumer graphics. It takes at least 1 480 GTX (sounds like 2+) to run the tessellation portion, and that leaves out most of nvidia's customers as well. One problem with the PolyMorph is that is scales with the amount of SIMDs, so on lower gpu's, there is less tess performance.

Excellent point and one that pretty much lays waste to a lot of the FUD spam in here. I can think of one developer that actually ever pushes the envelope graphically and that is Crytek, the rest put out stuff that on a graphic level is the same junk we've seen for the past thee years. The only game that came close is Metro 2033, and in that game it was not the DX11, the lighting is what was truly impressive.

There is still not a single game on the market that looks better than Crysis and the game is three years old. So much for developers pushing the envelope. They tell their employees what to do, and the employees do it and we get the same console ports or UE3 or whatever other recycled engine we've gotten for the past however long. Once in a while they tack on some 'feature' that adds little to nothing.

It's a shame, but true.

 

[NoQuarter]

A DX-11 game can be ported to Consoles and vise versa. It is like the VGAs in the PC. The same game can be played in DX-11 when we have a AMD HD5000/6000 or NVs GF1xx equivalent cards but it will play in DX-10 with HD4000 series etc etc.

So we can have a DX-11 PC game that will support Tessellation and at the same time we could ported to Consoles.

A game that uses tessellation to its fullest extent would have low poly meshes though, if you ported it to a system that can't tessellate using the displacement maps you would need to design additional high quality models or continue with what we've been seeing so far - high quality models with tessellation bolted on to reach very high quality.

 

[ugaboga232]

You can do some incredibly awesome things with Cryengine 2 so I am very excited for Cryengine 3. If something is going to push the envelope, that game is.

Also, Scali, I can't seem to find the link you were talking about on tomshardware. If you could link me, that would be great.

Edit: Thanks for the Link, reading now.

 

[Scali]

It is really only since the 5XXX series do we have full utilization of DX10, DX11 effects are pretty shallow compared to what it could do. Crysis can be hacked to use the very high dx10 locked features in dx9 mode, so that is one example against you.

Example against me, how?
Even though it may be DX9, it still won't run on a console that way.
You'll see when Crysis 2 comes out. They'll not look as good as the original Crysis on PC, nor as good as the PC version of Crysis 2.

Consoles really do dictate what developers do. What's stopping someone from making this tech demo into reality? Its beautiful so why don't developers make it.

It takes a few years to get from a new technology to a full working game.

Also most DX11 titles don't use DX11 for truly new effects outside of some games.

That's not the point.
Those games are not just vanilla console ports, and there's quite a lot of such games.
It doesn't matter whether these effects are any good, whether the game is actually any fun to play, whether nVidia or AMD sponsored them, or whatever else.
All that matters for this argument is that apparently developers have put effort into adding extra things to the PC version of a game. They didn't wait until consoles supported it.
So why should tessellation be any different?

 

[Scali]

Also, Scali, I can't seem to find the link you were talking about on tomshardware. If you could link me, that would be great.

I edited my post and added it.

 

[Scali]

A game that uses tessellation to its fullest extent would have low poly meshes though, if you ported it to a system that can't tessellate using the displacement maps you would need to design additional high quality models or continue with what we've been seeing so far - high quality models with tessellation bolted on to reach very high quality.

You can pre-tessellate the geometry offline.
But I already said that. Try to keep up.

 

[blastingcap]

It's a shame, but true.

I'm sure the biggest recession in decades isn't helping. Crysis came out 3 years ago when the economy was in far better shape. So I'm not sure if even Crytek will seriously try to push the envelope in Crysis 2 (by seriously I mean making a game that even GTX480 SLI will struggle with at ~30fps). Even fewer people these days are willing or able to blow that much dough on video games. (Reminder: Crytek said Crysis would be their last PC-exclusive (for a while, or maybe permanently) because of disappointing sales. So you'd think that they'd consider the potential size of the buying market when they set the level of graphics.)

In fact the economy is one of the reasons people are talking about re: why Sony/MSFT have declared a ceasefire in the console wars.. they'd both rather milk PS3/XBOX360 for a little longer, knowing that the market for a shiny new $600 PS4 may be even more limited than it was for PS3, and the PS3 had a big advantage with the built-in Blu-Ray player, too.

 

[Scali]

The thing with Crysis is that:
1) It was so far ahead of its time that even the most high-end systems at release were having trouble running it at the highest settings.
2) There has been an incredible impasse in GPU technology during that time. For a long time, the 8800Ultra was the fastest card. New cards introduced by both AMD and nVidia had better bang for the buck and all, but the raw performance was lower. It wasn't until the 5000-series and Fermi that we really saw a nice boost in performance again.

So it's not that strange that nothing has topped Crysis so far.

 

[ugaboga232]

In the end, lets wait for the 6970 and whatever tech demo's AMD releases as well as crysis.

Also, giving the ability to lower tessellation settings benefits nvidia as well. The 6870 beats it in unigine which leaves just the 470 and 480. If AMD can't play the game on the super high tessellation settings, nvidia's version won't be playable either.

I found this at the TH forums : http://www.tomshardware.com/forum/300985-33-more-dirty-tricks-nvidia

and it links this post: http://www.rage3d.com/board/showpost.php?p=1336398700&postcount=1376

But the ultimate test would be, name a game which AMD cards are not playable and nvidia cards are with tessellation enabled (game please, not tech demo, we can all find tech demo's of areas where nvidia does worse).


EDIT: BTW Scali, what do you think of id's new engines that use megatexturing and stuff like that? Do you think it can compete with tessellation or are both necessary for the next level of graphics?

 

[Arkadrel]

The thing with Crysis is that:
1) It was so far ahead of its time that even the most high-end systems at release were having trouble running it at the highest settings.
2) There has been an incredible impasse in GPU technology during that time. For a long time, the 8800Ultra was the fastest card. New cards introduced by both AMD and nVidia had better bang for the buck and all, but the raw performance was lower. It wasn't until the 5000-series and Fermi that we really saw a nice boost in performance again.

So it's not that strange that nothing has topped Crysis so far.

+1 internets

I completly agree, Im hopeing because of how kickass the new 580/570/560, and the 6990/6970/6950 are, we ll be seeing game developers pushing the games abit harder so we get prettier games.

 

[Scali]

Also, giving the ability to lower tessellation settings benefits nvidia as well. The 6870 beats it in unigine which leaves just the 470 and 480. If AMD can't play the game on the super high tessellation settings, nvidia's version won't be playable either.

Ermmm... it's a TWIMTBP title. If lower tessellation settings would have benefited nVidia, then they would have put them in.
The tessellation settings seem to be tuned to the GTX460. It can get good framerates with everything at the highest settings, at 1920x1080 with AA. The other cards are for the 'overkill' with the highest resolution monitors or multi-monitor setups.
nVidia could have gone higher, if they aimed specifically for the 480.

and it links this post: http://www.rage3d.com/board/showpost...postcount=1376

The guy is wrong... the polys are 6 pixels *wide*, which means they could still easily be more than 16 pixels total area (as per AMD's recommendations), if you also factor in their height (if they're also 6 pixels high, they'd amount to that, more or less... 1/2 * 6 * 6).
Aside from that, he doesn't understand the true problem (the rasterizer problem is the same for both nVidia and AMD, so why would nVidia perform so much better?).

EDIT: BTW Scali, what do you think of id's new engines that use megatexturing and stuff like that? Do you think it can compete with tessellation or are both necessary for the next level of graphics?

It's not really the same thing, hard to compare.
In theory, you wouldn't need textures at all. You can solve everything with real geometry. Textures are just a 'hack' to 'fake' geometric properties.

 

[ugaboga232]

Check out the new slides regarding the 6970. Supposedly it has another gen tessellator (aka 1 more than 68XX and 2 more than 5XXX). Seems somewhat interesting.

Hawx 2 is a bad example as almost every card gets playable fps at full specs. It's just that the 4XX gets much higher fps.

 

[Scali]

Check out the new slides regarding the 6970. Supposedly it has another gen tessellator (aka 1 more than 68XX and 2 more than 5XXX). Seems somewhat interesting.

If it isn't anything like PolyMorph, it's not interesting, as explained above.

It's just that the 4XX gets much higher fps.

That is EXACTLY the point.
Video cards are generally valued by their performance.
Whichever videocard delivers the most frames for your money, is the winner of the review.
Tessellation can turn the tables on which cards are the best bang for your buck, the least powerhungry, and all that.

 

[Leadbox]

If it isn't anything like PolyMorph, it's not interesting, as explained above.

Does it have to be? Does it matter if it can pump out the triangles?
Thx for the info and link earlier btw:thumbsup:

 

[Scali]

Does it have to be? Does it matter if it can pump out the triangles?
Thx for the info and link earlier btw:thumbsup:

It won't be able to pump out the triangles unless it's a setup like PolyMorph.
It's no coincidence that although the 6800-tessellator is twice as fast as the 5000-series tessellator at peak, it has pretty much the same dropoff and bottleneck at higher tessellation factors.
They could double it again, but you will just run into more diminishing returns.

So yes, I say it HAS to be like that. I already explained why earlier, but I see that was in vain.
Put it another way: Your question is much like "Does it matter if you have a single-core CPU, if it is as fast as an octacore one?"
No, it wouldn't matter, but obviously the question itself is ridiculous. The reason we have multicore CPUs is because single-cores cannot scale much higher in performance. So such a single-core processor will never exist.

 

[Seero]

Some say it is tomorrow's technology and will only be useful in the next 5 years. That is a wrong concept. We, as gamers, have a problem today, since yesterday when playing games, and that is, "Where the Fu*k is the bottleneck?" Neither GPU or CPU is at 100% so what is the point of upgrades? Yes, a new CPU or a new GPU will help, but that isn't an answer to why their ain't completely utilized to begin with. It is like buying a race car only to find out that it can only go 40km/h. Yes, you can buy a even faster car and you will be able to get some gain because it requires less time to get from 0 to 40km/h, but that really isn't solving the fundamental problem, which is the speed limit on the road.

How much performance you can get if your CPU and GPU both under 80-100% load at all time? I don't know but I know it isn't the case most of the time because it is a technical barrier to developers. To put GPU under load, it requires a huge amount of data. Data can be fed to GPU at initialization phase, aka loading screen. I don't know about you guys but I hate long loading screens and long zoning screens, but without them, GPU can't be utilized. Dynamic Tessellation is the solution for this. Instead of sending huge amount of data to GPU through the bus, send a amount of data to GPU and let GPU computes it on the fly. That reduces the load on sending data from storages to CPU, CPU to Ram, Ram to CPU, CPU to PCIe, and PCIe to GPU.

The easiest way to understand this is file transfer. Many people with some computer background knows that it is a good idea to "zip" the files before sending it through the internet. 1080p youtube videos is only possible because the amount of data sent is relatively small, but requires a lot more processing power to decode the feed. Dynamic Tessellation is exactly the something.

Tessellation itself is not new as all art designers use it like student uses paper. They don't need years of training to make it happen. However, existing arts can't be ported directly, they must be recreated. This is why the image quality changes so much on Unigine Dx11 demo. Technically, the quality is unchanged because Dx10 and Dx11 should use 2 distinct set of arts and Dx10 and Dx11 can't be switched on the fly. Dynamic switching should only be available under wireframe view. Since the demo is done wrong, many gamers who doesn't have a computer degree mis-understood dynamic tessellation as better image quality. Dynamic Tessellation does not increase the quality of the graphics, but capable of displaying high quality graphics at a portion of the bandwidth. That means, more dynamic tessellation = less bandwidth needed, which implies to more utilization on CPU, GPU, RAM, and i/o.

Usually, art design is the last stage of the development. If a game is being developed under Dx10, adding Dx11 tessellation is not hard, depending on how much arts are already done. This is why Civ 5 and Hawx 2 can go crazy on Dynamic Tessellation so quickly. You can expect lots of games requiring lots of tessellation on 2011, not 2015.

The "stop sign" is the number of people using Dx11 cards. Developers can't release a game if it looks like Tekken for most of its potential players. In other words, they can't release a game that is only playable on Dx11 cards. What they really should have done is to first create arts under Dx11, then create a pre-tessellated skin for Dx10. This way, the game will look identical under Dx10/11, but 460/6870 will get 10x FPS compare to 285/4870.

 

[taltamir]

One of the funniest things is how "DX11 is irrelevant" switched sides. It used to be DX11 capabilities were irrelevant due to low market penetration was the argument of nvidia, who only had DX10 hardware... now its "tesselation, the biggest feature in DX11 is irrelevant" because of market penetration is the argument coming from the AMD side since nvidia has drastically more tessellation capability.

In the end, I have to maintain my previous position that this argument is still as valid today for AMD as it was for nVidia last year... this is awesome quality improvement, but it will be years before it will be widely used in games... I will certainly demand polymorph engine in my HD 8770 or GTX680 in 2012-2013... but as utterly amazing a technological achievement that it is, I don't see it making the feature cut for games in the near future.

@Scali, thanks for clarifying some info about how tessellation works.

 

[blastingcap]

Usually, art design is the last stage of the development. If a game is being developed under Dx10, adding Dx11 tessellation is not hard, depending on how much arts are already done. This is why Civ 5 and Hawx 2 can go crazy on Dynamic Tessellation so quickly. You can expect lots of games requiring lots of tessellation on 2011, not 2015.

The "stop sign" is the number of people using Dx11 cards. Developers can't release a game if it looks like Tekken for most of its potential players. In other words, they can't release a game that is only playable on Dx11 cards. What they really should have done is to first create arts under Dx11, then create a pre-tessellated skin for Dx10. This way, the game will look identical under Dx10/11, but 460/6870 will get 10x FPS compare to 285/4870.

(The TL;DR version of what I'm about to say: Businesspeople decide whether or not to include DX11 into a game and how much of it to include, not engineers, so don't expect extreme tessellation to pop up all over the place now or even in 2011.)

In case that was directed at me, I want to reiterate my thoughts more clearly:

Although the number of games that use DX11 effects is rising, many games won't use DX11 at all through 2011.

1. There will be an increasing number of PC games taking advantage of DX11

2. However, from a business perspective it doesn't make sense for some gamedevs to put ANY resources in DX11 in their games. Many games are not sold for their graphics, but rather their gameplay. E.g., Plants v Zombies and World of Goo will see little need for DX11. They need optimized DX9, Flash, etc. to run on low-spec machines. Also, there are lots of people who know how to work with DX9 out there whom you can hire for cheap. I can't believe that it would be cheaper to hire DX11-savvy people instead of or in addition to the DX9 team. Keep in mind that we are waist-deep in one of the deepest recessions of the last 100 years with no real end in sight.

Of the games that use DX11 through 2011, it's unlikely that many of them will use extreme tessellation. (The best chance of a breakout is probably Crysis 2, which has received NVIDIA money and may have lots of DX11 effects AND GPU PhysX.)

1. Even if a game uses DX11 effects, the game still needs to look good on DX9. Why? Market size. If I'm a sane businessperson, then I'm going to give DX9 priority for resources. Compared to the massive size of the DX9 market (including consoles), there are not many rigs that have a combination of a) DX11 operating system + b) gaming-quality CPU + c) gaming-quality DX11 GPU. Keep in mind that the economy is horrible, which doesn't help in rapidly increasing DX11 rig counts.

2. Even among games that use DX11 effects (perhaps for marketing reasons), gamedevs know that AMD owns ~90% of the DX11 market, so a large fraction of DX11 rigs will have a relatively weak tessellator. Am I going to use extreme tessellation so that performance tanks on that large fraction of DX11 rigs using AMD GPUs? Especially since the number of Dx11 rigs is already so low to begin with? Not unless I get a lucrative deal from NVIDIA to make up for lost $ales due to frustrated AMD-GPU owners refusing to buy my game.

3. Note that Civ V and HAWX2 run fine on AMD GPUs. We have yet to see a game that totally tanks on AMD GPUs thanks to extreme tessellation. This is because businesspeople have final say on how much tessellation to use in a game.

4. The above is why I said that we won't be seeing extreme tessellation in major games anytime from now through 2011, and possibly later. There might be some weirdo indie gamedev out there who makes a playable tech demo using extreme tessellation, but I'm excluding fringe cases like that. You can quote me on this, too: other than *perhaps* Crysis 2, there will be no major game titles in 2011 that can't run in DX9.

I have no idea where 2015 came from so that must be referring to someone else. I make no concrete predictions past Dec. 31, 2011. I even hedged on whether there is some top-secret console launch due in 2011 that we don't know about.

 

[happy medium]

I heard NV paid Crytek $2MM for TWIMTBP rights, so if there is one game coming out in 2011 that has a chance of tessellating so hard that even NV's GPUs hurt, let alone AMD's, then it's Crysis 2. Oh and I expect Crysis 2 to support GPU PhysX as well. Entirely by coincidence of course.

Aside from that, I have difficulty imagining any other game pushing extreme tessellation in 2011.

So is tessellation important now? It seemed real important to you when the rumers that the new 6000 series was gonna match Nvidia's ability.

Whats next, Physx will be important when AMD catches up?

Come on now Blastingcap.

 

[blastingcap]

I think you have me confused with someone else. I don't think I ever said tessellation was unimportant. If anything I was salivating over rumors that Barts and Cayman were going to have beefed up tessellators. As Scali has rightfully pointed out, though, AMD's kludgey solution is not as elegant or "futureproof" as NV's. I have voted with my wallet to get a GTX460-768 rather than a disappointingly-priced 6850, with the idea to get a Kepler next year if it can do single-GPU 3-way monitors. Actions speak louder that words, so I have nothing to prove re: my impartiality.

PhysX is another topic and if we're going to discuss that it should be in another thread, rather than threadjacking Scali's thread even more.

So is tessellation important now? It seemed real important to you when the rumers that the new 6000 series was gonna match Nvidia's ability.

Whats next, Physx will be important when AMD catches up?

Come on now Blastingcap.

 

[Seero]

(The TL;DR version of what I'm about to say: Businesspeople decide whether or not to include DX11 into a game and how much of it to include, not engineers, so don't expect extreme tessellation to pop up all over the place now or even in 2011.)...

No it isn't directed to you.

Yes, it is a business discussions. In terms of hardware, Intel decided to resurrect dynamic tessellation and made it theirs. AMD and Nvidia followed and released Dx11 hardwares. Some games have decided to go with Dx11, and the easiest way is to use dynamic tessellation.

I never said AMD can't do tessellation, I said dynamic tessellation is a good thing, regardless of rather it is AMD or Nvidia. I disagree when people say "Nvidia tessellated too much, it isn't necessary." If AMD can't handle high degree of dynamic tessellation, than the ball is on their court. They can rationalize the problem as their product is fine as extreme dynamic tessellation isn't the strong point, but enough for 1080p user, and efficient for most games now and near future. On the other hand, they can fix the tessellation unit so that it is scalable.

Dynamic tessellation is a good thing, but it depends on how people implement/utilize it. If the result is better quality at a lower frame rate, then they have implement it wrong. This isn't direct to you or AMD.

 

[blastingcap]

No it isn't directed to you.

Yes, it is a business discussions. In terms of hardware, Intel decided to resurrect dynamic tessellation and made it theirs. AMD and Nvidia followed and released Dx11 hardwares. Some games have decided to go with Dx11, and the easiest way is to use dynamic tessellation.

I never said AMD can't do tessellation, I said dynamic tessellation is a good thing, regardless of rather it is AMD or Nvidia. I disagree when people say "Nvidia tessellated too much, it isn't necessary." If AMD can't handle high degree of dynamic tessellation, than the ball is on their court. They can rationalize the problem as their product is fine as extreme dynamic tessellation isn't the strong point, but enough for 1080p user, and efficient for most games now and near future. On the other hand, they can fix the tessellation unit so that it is scalable.

Dynamic tessellation is a good thing, but it depends on how people implement/utilize it. If the result is better quality at a lower frame rate, then they have implement it wrong. This isn't direct to you or AMD.

I agree.

 

[Grooveriding]

I heard NV paid Crytek $2MM for TWIMTBP rights, so if there is one game coming out in 2011 that has a chance of tessellating so hard that even NV's GPUs hurt, let alone AMD's, then it's Crysis 2. Oh and I expect Crysis 2 to support GPU PhysX as well. Entirely by coincidence of course.

Aside from that, I have difficulty imagining any other game pushing extreme tessellation in 2011.

I had the same bad feeling that physx would be in Crysis 2 when I read about that $2 million NV paid Crytek to get their name on it.

Someone replied to me with some info about Crytek having already put a list of specifications of Cryengine 3 out there and it made specific mention of their own physics implementation.

I think it is a safe bet there will be no physx in Crysis 2. It will be the usual TWIMTBP logo on startup and of course the game will use DX11.

Crytek designs games so well I have my doubts they would ever add something to the detriment of their game, or moreso, their engine, just to satisfy NV. I'm sure there will be tessellation but I'll bet it will be to the benefit of the game rather than making an excessive use of it for the sake of marketing.