Jump to content

  • Log In with Google      Sign In   
  • Create Account

Banner advertising on our site currently available from just $5!


1. Learn about the promo. 2. Sign up for GDNet+. 3. Set up your advert!


Graphics baseline for a good-looking PC game?


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
24 replies to this topic

#1 jefferytitan   Crossbones+   -  Reputation: 2433

Like
0Likes
Like

Posted 11 May 2014 - 05:14 PM

I was thinking about a variety of games that I've played. It can be hard (for me) to tell which graphical techniques they use. I would never aim to write a AAA killer (due to the effort/reward ratio), but what sort of techniques are expected for a game not to look dated? A few examples:

  • Textures - simple, with normals, relief-mapped, procedurally generated, sub-surface scattering?
  • Geometry - number of polys, tessellation?
  • Lighting/shadows - forward, deferred, soft shadows, SSAO, global illumination?
  • Effects - HDR, bloom, FSAA, volumetric fog, god rays?
  • Sky - skybox, animated, day/night, weather, seasons?

What would be your baseline must-have for a FPS that won't turn off paying consumers?



Sponsor:

#2 Shane C   Crossbones+   -  Reputation: 1315

Like
4Likes
Like

Posted 11 May 2014 - 05:24 PM

Textures - the detail texture should be of at least size 1024x1024 except for small objects. Although something like 1368x768 is the average or most common screen resolution right now, count on most of your players for a FPS having at least 1680x1050 and being able to see every detail.

 

Geometry - a human character should have at least 10k polys. Less is accepted with normal-mapping, or using tessellation to boost things up.

 

What I can't stress enough is the importance of shaders these days. A good shader writer can potentially turn a dull, drab game into an exciting one, I think.


Edited by Shane C, 11 May 2014 - 05:24 PM.


#3 SeanMiddleditch   Crossbones+   -  Reputation: 9902

Like
7Likes
Like

Posted 11 May 2014 - 05:27 PM

Lighting/shadows - forward, deferred, soft shadows, SSAO, global illumination?


This is the differentiation that's come about most in the last couple of generations. You need high-res textures/models, but by themselves they will still look like crap without good lighting techniques. Even less-realistic games put considerable time into lighting, shading, and so on.

#4 Chris_F   Members   -  Reputation: 2654

Like
3Likes
Like

Posted 11 May 2014 - 05:38 PM


Although something like 1368x768 is the average or most common screen resolution right now

 

My intuition would have been 1920x1080, and that is confirmed by Steam's hardware survey. More than a third of Steam users have a 1080p (or better) display. My guess is that the majority of the people with less than that fall into the category of casual gamers and are less likely to playing the "good-looking PC games". 1080p monitors are so cheap and ubiquitous today that if you were going to buy or build a computer with the intention of playing video games, there would be no sense in getting anything less.


Edited by Chris_F, 11 May 2014 - 05:39 PM.


#5 jefferytitan   Crossbones+   -  Reputation: 2433

Like
0Likes
Like

Posted 11 May 2014 - 05:44 PM

One huge difference I've noticed between really old games and new ones is the lack of repetitive textures. Some games seem to have an impossible texture budget. Do you think this is due to bigger textures, lots of textures plus splat mapping, fancy shaders, procedural textures, something else?



#6 Hodgman   Moderators   -  Reputation: 37958

Like
13Likes
Like

Posted 11 May 2014 - 05:44 PM

The quality of your art matters more :P
Bad art in a state-of-the-art engine will still look bad, but great art can make a very dated engine shine beyond its years.


Back on topic though - the genre that you're targeting will have a big impact on graphical expectations.

#7 Shane C   Crossbones+   -  Reputation: 1315

Like
0Likes
Like

Posted 11 May 2014 - 05:49 PM

One huge difference I've noticed between really old games and new ones is the lack of repetitive textures. Some games seem to have an impossible texture budget. Do you think this is due to bigger textures, lots of textures plus splat mapping, fancy shaders, procedural textures, something else?

 

Just to take a guess, old games used impossibly small texture sizes, like 256x256. New games probably use an average closer to 2048x2048, which gives 64 times the detail of a 256x256 texture like you might see in the GameCube/original Xbox/PS2 days and shortly after. Texture detail has shot up like crazy over the years, and only a few developers haven't really taken advantage of it (Sonic 2006, I'm looking at you).



#8 jefferytitan   Crossbones+   -  Reputation: 2433

Like
0Likes
Like

Posted 11 May 2014 - 06:02 PM

Actually to semi-answer my own question re textures, I was trying to find some good example screenshots, and I think that a huge factor is (a) increased poly/art budget therefore fewer huge flat chunks of land and (b) masking agents such as plants and volumetric grass which prevent seeing huge chunks of bare ground at a time.



#9 MJP   Moderators   -  Reputation: 13440

Like
15Likes
Like

Posted 11 May 2014 - 06:08 PM

IMO high-quality graphics is less about buzzwords, and more about having great interaction between your technology and the artists. You need to have talented artists, talented engineers, and a good working relationship between them. You also need a clear vision of what sort of look your game will have, and engineers that can make the right choices in how to deliver that look. There have been a lot of great-looking games that took low-tech approaches to their game, and they succeeded because they had a consistent vision and their tech was well-suited for the game they were making.

 

Obviously in a lot of cases you will end up using commonly-used techniques since they are well-understood and have good quality/performance characteristics, but that doesn't mean your game will look good just because your engine supports them. The engineers need to make the right choice in which techniques to use, and they need to understand them well enough to explain to the artists how to effectively use them. A good artist can make really crappy visuals with good tech if they don't understand it, or if the tech isn't well-suited to what they're trying to create.

One last critical point is that it's crucial to understand the scope of the game that you can pull off, and make sure that your game design and visuals are well-tailored to the limitations of your team. If you have a small team with limited money, you're not going to be able to pump out the insane amount of unique content that a large AAA studio can. So you won't be creating hundreds of unique meshes per level, along with dozens of unique motion-captured animations for non-interactive sequences. Instead it would be better to take a more minimalist approach, where you can reuse more content or otherwise avoid the need for lots of levels/game areas.



#10 MJP   Moderators   -  Reputation: 13440

Like
8Likes
Like

Posted 11 May 2014 - 06:11 PM

One huge difference I've noticed between really old games and new ones is the lack of repetitive textures. Some games seem to have an impossible texture budget. Do you think this is due to bigger textures, lots of textures plus splat mapping, fancy shaders, procedural textures, something else?

 

Well there's definitely more memory and manpower for creating unique textures, but a lot of it is due to techniques that are able to add unique detail using tileable textures. Many materials in modern games will blend together many maps, which can effectively hide the repetitiveness if done correctly. In our game all environment geometry uses "layer-blended" materials, where different maps and material properties are blended together based on blend weights stored in a  vertex color channel. This gives you the ability to paint unique features and "break-up" even though the textures being used for each layer are tiles across the surface.



#11 lunkhound   Members   -  Reputation: 820

Like
0Likes
Like

Posted 12 May 2014 - 07:56 PM

 

One huge difference I've noticed between really old games and new ones is the lack of repetitive textures. Some games seem to have an impossible texture budget. Do you think this is due to bigger textures, lots of textures plus splat mapping, fancy shaders, procedural textures, something else?

 

Well there's definitely more memory and manpower for creating unique textures, but a lot of it is due to techniques that are able to add unique detail using tileable textures. Many materials in modern games will blend together many maps, which can effectively hide the repetitiveness if done correctly. In our game all environment geometry uses "layer-blended" materials, where different maps and material properties are blended together based on blend weights stored in a  vertex color channel. This gives you the ability to paint unique features and "break-up" even though the textures being used for each layer are tiles across the surface.

 

 

If you've got non-overlapping UV coords, you could store the blending weights in a texture instead of on the vertex colors.  That way your layer transitions wouldn't be tied to vertex density.  You could also put the blend weights in a compressed texture to save a bit on memory.  Anyone know how UE4 does it?



#12 jefferytitan   Crossbones+   -  Reputation: 2433

Like
0Likes
Like

Posted 13 May 2014 - 02:22 PM

Any opinions out there - does tesselation make techniques like relief mapping obsolete? I would assume (and correct me if I'm wrong) that you could use the same kind of texture to add geometry detail, and because it's geometry it would benefit from the standard lighting/shadowing/etc without the relief mapping problems with .aliasing and oblique angles. Also can you stack standard vertex/fragment shaders on top of tesselation? That would make it easier to combine effects.



#13 Vilem Otte   Crossbones+   -  Reputation: 1960

Like
2Likes
Like

Posted 14 May 2014 - 03:10 AM

Technically taken, it is possible to use tessellation along with F.e. QDM or POM or relief mapping (anyone actually tried it? If so, could you share a screenshot?) ... but once you can tessellate to the level where your triangles are at size of pixel (which isn't a problem on modern GPUs), why would you need & want to do some offset mapping - it won't help the model to look better at that point.


My current blog on programming, linux and stuff - http://gameprogrammerdiary.blogspot.com


#14 phantom   Moderators   -  Reputation: 8521

Like
4Likes
Like

Posted 14 May 2014 - 04:26 AM

but once you can tessellate to the level where your triangles are at size of pixel (which isn't a problem on modern GPUs)


Except where it is a MASSIVE performance problem because you are now issuing one wave front (32 or 64 threads) to do one pixel's worth of work as pixel workloads are dispatched in groups causing massive under utilization of the hardware and wasted resources all over the place.

Pixel sized triangles are the devil.

Do.
Not.
Want.

#15 Hodgman   Moderators   -  Reputation: 37958

Like
4Likes
Like

Posted 14 May 2014 - 04:37 AM

but once you can tessellate to the level where your triangles are at size of pixel (which isn't a problem on modern GPUs)

Except where it is a MASSIVE performance problem because you are now issuing one wave front (32 or 64 threads) to do one pixel's worth of work as pixel workloads are dispatched in groups causing massive under utilization of the hardware and wasted resources all over the place.Pixel sized triangles are the devil.Do.Not.Want.
AFAIK, a 64-wide wavefront can be filled with 16 2x2 pixel quads - not necessarily within the same triangle (or a 32-wide can be willed with 8 quads).
i.e. Pixel-sized triangles are as bad as they've always been (25% performance), but not worse.

Is there a reference (or published experiment) where I can confirm either way? Do all 3 vendors perform the same way?

#16 phantom   Moderators   -  Reputation: 8521

Like
2Likes
Like

Posted 14 May 2014 - 05:26 AM

Is there a reference (or published experiment) where I can confirm either way? Do all 3 vendors perform the same way?


I think they do it differently; NV do perform better at low tessellation than AMD (which is why they shout about it the most) and Intel I've no idea about.

As normal NV are pretty silent on their internal workings.
AMD do have this document however; http://t.co/zOyz5DFa6D (APU13 talk "The AMD GCN Architecture: A Crash Course") - slide 61 onwards pertains to this topic but the whole thing is a nice chunk of information to have.

#17 Ohforf sake   Members   -  Reputation: 1954

Like
2Likes
Like

Posted 14 May 2014 - 09:01 AM

Another thing to keep in mind is that with parallax mapping, you only pay the price of the difficult texture lookup once. With tesselation, you pay the price of the increased amount of vertices that need to be transformed once for the z pre pass, then for every shadow map and then again for the actual rendering.

Something that I never liked about tesselation is the resulting topology of the mesh. A good artist can squeeze out a lot of curvature from a small amount of triangles by aligning the edges correctly. Most tesselation techniques however just subdivide and displace, which results in far more triangles then an artist would have needed. Has anyone worked out a solution to this, or seen some work that goes in that direction?

PS: To tie this back in with the original question of the OP: You don't need tesselation for a good looking FPS. But you might want to have animations, particle effects and decals (only mentioning it because all three are missing from the list).

#18 Migi0027 (肉コーダ)   Crossbones+   -  Reputation: 3202

Like
2Likes
Like

Posted 14 May 2014 - 03:46 PM

Another thing to keep in mind is that with parallax mapping, you only pay the price of the difficult texture lookup once. With tesselation, you pay the price of the increased amount of vertices that need to be transformed once for the z pre pass, then for every shadow map and then again for the actual rendering.

Something that I never liked about tesselation is the resulting topology of the mesh. A good artist can squeeze out a lot of curvature from a small amount of triangles by aligning the edges correctly. Most tesselation techniques however just subdivide and displace, which results in far more triangles then an artist would have needed. Has anyone worked out a solution to this, or seen some work that goes in that direction?

PS: To tie this back in with the original question of the OP: You don't need tesselation for a good looking FPS. But you might want to have animations, particle effects and decals (only mentioning it because all three are missing from the list).

 

There is a solution, if I read your answer right, some methods do it on the fly, others do it offline, generating a "density" of where tesselation is required, as if you had a hill and a flat area after the hill, the density map ( texture maybe ), would have a high intensity around the hill and practically zero around the flat area, then use this intensity value to determine the tesselation factor on the fly. In this way computational power would be reduced, with the cost of an extra texture ( if using the offline method ).


Hi!

Game-Dev Journal: www.gamedev.net/blog/1882-the-cuboid-zone/

Wordpress Journal: cuboidzone.wordpress.com/


#19 Alessio1989   Members   -  Reputation: 2674

Like
2Likes
Like

Posted 14 May 2014 - 05:12 PM

 

Is there a reference (or published experiment) where I can confirm either way? Do all 3 vendors perform the same way?


I think they do it differently; NV do perform better at low tessellation than AMD (which is why they shout about it the most) and Intel I've no idea about.

As normal NV are pretty silent on their internal workings.
AMD do have this document however; http://t.co/zOyz5DFa6D (APU13 talk "The AMD GCN Architecture: A Crash Course") - slide 61 onwards pertains to this topic but the whole thing is a nice chunk of information to have.

 

 

As far I know both AMD VLIW and GNC architecture perform good at low & mid-low tessellation factors...

 

This is the "old" sample of the legacy Directx SDK..

 

http://www.pcgameshardware.de/screenshots/original/2011/12/HD-7970-DX11-Tessellation.png

 

I got similar performance droops on all tessellation demos I tested (not only from Microsoft) at mid and mid-high tessellation level on both my system (the first has a VLIW4 6970, the second a cheap GCN 7730 mobile).

 

EDIT: for GCN max recommended tessellation factor value is 15. http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2013/05/GCNPerformanceTweets.pdf

 

EDIT2: "AMD: tessellation factors above 15 have a large impact on performance" http://developer.amd.com/wordpress/media/2013/04/DX11PerformanceReloaded.ppsx

 

...

 

x7/x8 should be the max recommended value for VLIW4/5 GPUs (make sense since the doubled the tessellation unit in the GNC architecture)..


Edited by Alessio1989, 17 May 2014 - 08:03 AM.

When you're upset, just imagine a T-Rex making a bed


#20 Migi0027 (肉コーダ)   Crossbones+   -  Reputation: 3202

Like
3Likes
Like

Posted 14 May 2014 - 11:43 PM

 

.......

 

There is a solution, if I read your answer right, some methods do it on the fly, others do it offline, generating a "density" of where tesselation is required, as if you had a hill and a flat area after the hill, the density map ( texture maybe ), would have a high intensity around the hill and practically zero around the flat area, then use this intensity value to determine the tesselation factor on the fly. In this way computational power would be reduced, with the cost of an extra texture ( if using the offline method ).

 

 

If you're interested, theres an interesting paper that adresses it: ( Around Page 44 )

 

https://developer.nvidia.com/sites/default/files/akamai/gameworks/events/gdc14/GDC_14_From%20Terrain%20to%20Godrays%20-%20Better%20Use%20of%20DirectX11CantlayTatarinov.pdf

 

Even though this paper has way more information than you need tongue.png . ( Even though it's interesting )


Edited by Migi0027, 14 May 2014 - 11:43 PM.

Hi!

Game-Dev Journal: www.gamedev.net/blog/1882-the-cuboid-zone/

Wordpress Journal: cuboidzone.wordpress.com/





Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS