Speculation on the polycount or quality of nextgen character models.

ultragpu

Banned
Now as the current gen consoles are seeing their twilight days I think it would be interesting to discuss and speculate with educated guesses on what the nextgen can offer us in terms of INGAME character models.

This generation of making ingame character models with its vast utilization of Normalmaps extracted from high res source models, have been both efficient and effective judging by the result. But there are still tons of improvements to be made and technical hurdles to overcome. If you look at the hero model from Epic's Samaritan realtime demo, then compare to something like Nathan Drake or Battlefield soldier model, you can clearly see a big difference in quality.

So assuming a nextgen consoles is as powerful as or close to a 580 sli setup, do you think we can standardize tessellation with displacement maps on most of the nexgen in game characters? Do you think Samaritan quality is representative of what would be shown on nextgen consoles? Or do you think most would simply keep the old Normalmap route but only increase the polycount as they find power would be better used elsewhere? I wonder how close we can get to the zbrush source model with nextgen.
 
no point in going much higher than now, just enough to nail some of the silhouette problems, of which realtime tessellation may be able to take care of.

higher res textures, and better shaders are mostly whats needed right now, far more beneficial to just throwing more polys at a scene.
 
Now as the current gen consoles are seeing their twilight days I think it would be interesting to discuss and speculate with educated guesses on what the nextgen can offer us in terms of INGAME character models.

This generation of making ingame character models with its vast utilization of Normalmaps extracted from high res source models, have been both efficient and effective judging by the result. But there are still tons of improvements to be made and technical hurdles to overcome. If you look at the hero model from Epic's Samaritan realtime demo, then compare to something like Nathan Drake or Battlefield soldier model, you can clearly see a big difference in quality.

So assuming a nextgen consoles is as powerful as or close to a 580 sli setup, do you think we can standardize tessellation with displacement maps on most of the nexgen in game characters? Do you think Samaritan quality is representative of what would be shown on nextgen consoles? Or do you think most would simply keep the old Normalmap route but only increase the polycount as they find power would be better used elsewhere? I wonder how close we can get to the zbrush source model with nextgen.

You have got no hope with that matey...

With the next xbox all but confirmed as running a 7000 series GPU it could end up with less power then a single GTX 580.

Samaritan also used Tri-SLI GTX 580's so it could be a while before we see console games on that level.
 
With the next xbox all but confirmed as running a 7000 series GPU...

Really? I think I missed that.

It would be strange to see MS going from having a cutting edge architecture in the Xbox and Xbox 360 (both were months ahead of comparable parts in the PC space) to having one that's a couple of years out-of-date at launch. Even Sony were only 9 months behind on the 7900 series.

it could end up with less power then a single GTX 580.

But the results obtained from it will probably be a lot better.

Samaritan also used Tri-SLI GTX 580's so it could be a while before we see console games on that level.

Dual 6800 GT dev kits couldn't hope to keep pace with the real 360 when pre-launch 360 games were shown off on them; if history repeats in a similar manner then it probably won't be long till the tech is surpassed. Tech demos normally look good long after the tech is surpassed though. The nVidia Dawn demo still looks surprisingly great especially when run at native res with SSAA on.
 
You have got no hope with that matey...

With the next xbox all but confirmed as running a 7000 series GPU it could end up with less power then a single GTX 580.

Samaritan also used Tri-SLI GTX 580's so it could be a while before we see console games on that level.

Though you do have to remember that raw power isn't everything, consoles are closed boxes, and everyone knows how much further a closed box can be pushed compared to PC enviroment
 
That all depends on what you want to speculate as being the next-gen console baseline. Is it the dreamers' hardware console configuration or is it the people based in reality's hardware console configuration? Is it the Wuu or is it a PS4/XB3? Is it the Wuu that can magically stuff Tri-SLI GTX680s in that tiny case?

Until you can get those down your answers will be all over the place.
 
We'll have to wait and see if Wuu has an API (let alone hardware tech) that can make use of tessellation to the same degree as DX11 before we speculate on tessellation uptake. And then it'll be a matter of whether devs care for adding that complexity to their art asset pipeline. A couple devs have made some progress, but not the big ones yet. It'll be a long while...
 
no point in going much higher than now, just enough to nail some of the silhouette problems, of which realtime tessellation may be able to take care of.

higher res textures, and better shaders are mostly whats needed right now, far more beneficial to just throwing more polys at a scene.

One of the biggest problems I still see even in high end pc games is clothing and hair have primitive or no realistic animation.

That's one of the big 3 things that should hopefully be solved*(others being essentially perfect IQ, and very realistic Lighting).
 
There's a lot that can be done to improve the looks of a model with a Nathan Drake polycount. The skin is still very plasticy and smooth, and would benefit from subsurface scattering, blended normal maps (a la GoW3), and higher resolution textures.

But that would give "just" an N64 to Dreamcast degree of improvement. If we've got bigger poly budgets, we could have clothes layered onto bodies instead of a unified skin. Clothes would have physics and collide realistically with stretching, bunching, and creasing properties that we're not accustomed to seeing in games today. Of course more polys can be thrown into hair, as well.
 
I agree that the next consoles incorporating what is right now a cutting edge PC graphics card is unlikely, too many problems, cost, power, heat. I also agree that you can get so much more out of a console GPU because it is a fixed and known quantity and graphics API's are very thin.

Next generation is going to be interesting, the jump power from PS2->PS3 era was in part consumed by high definition. 720p in theory though actually sometimes not quite that. Sure next generation we'll see more 1080p games but for me it's about quality of image rather than number of pixels. It's the overall experience, content, post processing, image quality, good frame rate etc..

This next generation we get all the power the jump provides for content, the questions are:
1. Do we have the memory to store it?
2. Do we have the art budgets to generate it?

We got a 13x increase in memory from PS3 to PS3, 38mb (32mb main ram, 4mb GS memory, 2mb IOP memory) to 512mb (256mb VRAM, 256mb main memory). I don't see that we'll get a 13x increase this time around as that would exceed the 4gb barrier and mean 64 bit addressing or some sort of memory paging system. I would like to see 4gb or more but I wouldn't put money on it. Console development has always been about extremely tight memory constraints. Fast storage for streaming would be good but without having to install games (which sucks)

Fortunately technologies such as tesselation and displacement mapping actually reduce the memory footprint per vertex. Unfortunately it's textures that eat most of the video memory. We will get higher quality texture compression next generation but it will cost more memory.

Programability will take a great leap forward next generation, no longer are we going to be constrained by the classic vertex shader->pixel shader->render target pipeline. I think we're going to see more diverse "how did they do that" rendering / effects, lots of new technologies.

The real problem is artists budgets and money. People have moaned for years about piracy but I think the 2nd hand games market has done more to damage finances than piracy has ever done.
 
I agree that the next consoles incorporating what is right now a cutting edge PC graphics card is unlikely, too many problems, cost, power, heat. I also agree that you can get so much more out of a console GPU because it is a fixed and known quantity and any more.
Yes, PC's have bigger overhead for feeding the GPU due to OS and heavy layering but once the stuff is already running on the GPU the speed of executing the shaders is pretty much the same (ignoring console-specific fine tuning). Once you start doing crazy things a'la software shading on Cell things get more complicated obviously
 
Yes, PC's have bigger overhead for feeding the GPU due to OS and heavy layering but once the stuff is already running on the GPU the speed of executing the shaders is pretty much the same (ignoring console-specific fine tuning). Once you start doing crazy things a'la software shading on Cell things get more complicated obviously

It may be true but there's no denying budgets. Budgets place a unique cap in the pc platform, affecting both the quality as well as the quantity of unique content of most exclusive titles.(This has affected iphone and ipad too, as even the original psp games run around most iApps content, despite the superior capabilities of the newest apple platforms.)

A very interesting thing is that we essentially seem to be stuck with primitive or no clothing and hair physics as I mentioned earlier. It is probably the thing that stands out the most beside SS skin in human characters, comparing real time to prerendered computer animations.

One of the things that I hope improves in the final wii U zelda is the quality of link's clothing animation(In my opinion clothing animation was only acceptable in link's hat).
 
There's a lot that can be done to improve the looks of a model with a Nathan Drake polycount. The skin is still very plasticy and smooth, and would benefit from subsurface scattering, blended normal maps (a la GoW3), and higher resolution textures.

Subsurface scattering is not really possible in realtime. All advanced game engines have some kind of specialized skin shader though, even Uncharted 2-3.
Not sure what you mean by blended normal maps...

If we've got bigger poly budgets, we could have clothes layered onto bodies instead of a unified skin. Clothes would have physics and collide realistically with stretching, bunching, and creasing properties that we're not accustomed to seeing in games today. Of course more polys can be thrown into hair, as well.

Multilayered cloth is insanely expensive in computations, especially if you want it to be robust enough for an interactive environment. We have part of our renderfarm running simulations 24/7 at the end of a CG movie project and even then we usually just fix small intersection problems by hand instead of tweaking and re-running the simulation.

Hair is another important and problematic issue, as mentioned by others...
 
GraphicsCodeMonkey said:
as that would exceed the 4gb barrier and mean 64 bit addressing or some sort of memory paging system.
We've already had 64bit addressing this generation in one console until the developer community (thankfully)booed it out of the SDK - ie. it's hardly an issue "IF" there was an actual use for it, all hardware starting with PS2 (except original XBox) was addressing memory above >4GB just fine already.

Whether that will fit financial constraints is another issue though.
 
Back
Top