Rendering tech of Infamous Second Son

On a side note, was going through the shadows section, and... it's really quite brute force in approach - 8x8 PCF. In screenshots it looks fine, but the cascade shadow maps make the transitions much more noticeable in motion. Also kind of weird they use so many samples that blur the shadowmap, and then apply a sharpening filter... blobby result at times.

Wonder if they'll look into SDSM for the future.
 
High number of samples may have to do with artifacts, shadow maps can bleed, create splotches or "akne" and so on. In offline CG using shadow maps, its the lighters' job to properly finetune all the variables per shot, based on the camera - but this is not possible in a game, so some sort of compromise have to be found between quality and artifacts. This could be one of these trade offs...

As for the quality issues, I was merely trying to point out that - IMHO - the game looks good but in terms of visible features, it seems to be more about scaling existing technology up to higher detail levels, instead of introducing new things. Not trying to start an argument or anything.
 
Laa-Yosh said:
High number of samples may have to do with artifacts, shadow maps can bleed, create splotches or "akne" and so on. In offline CG using shadow maps, its the lighters' job to properly finetune all the variables per shot, based on the camera - but this is not possible in a game, so some sort of compromise have to be found between quality and artifacts. This could be one of these trade offs...
Click to expand...

Fair enough, but then... to run a sharpening filter is a bit... well... :p 8x8 is a pretty high amount, but I guess it ran within budget (sort of... or not)... Would be curious to know the frame cost of that.

How about using signed distance fields? mm... baked I guess.
 
AlNets said:
Guys, this is a tech thread, not another argument on one's opinions of impressiveness.



I think the argument would help better if there were a comparison of how the buffer precisions or rather attribute precisions would have affected the results. :) (Crytek certainly had a couple when discussing their trade-offs on PS360 for their buffer formats).

Graham wrote up a good post a while back on certain attributes that benefit more from retaining high precision. Are there trade-offs to precision that can be made? Sure. Sebbbi made the case many times. MJP did also make the case for needing higher precision for certain things still. It depends.

Besides, as already mentioned in the presentation (and already reiterated by Sebbbi), decisions were made for ease of development and time. That's it.

So let's stop getting into circular preference arguments please. It's stupid that every thread needs to devolve into such trivial matters in the face of the main topic.
Click to expand...
The problem is that its hard to compare, because of difference in art styles and technology in many games.
I could put out Crysis 2 or Sleeping Dogs shots, but it would not be good comparison overall, because their lack some features like PBR, specular reflection probes or SSR [C2 has those].
Ryse g-buffer setup would be the best for comparison, because its very heavy packed yet still has all Infamous features included, but enormous difference in art styles makes screenshots comparison completely useless. So for real comparison, we need to wait for next Crytek game that covers modern or sci-fi setting, or some other developer like Massive that would disclose their g-buffer setup.
I think the best comparison for today would be Epic's Samaritan demo, although its still lacks PBR.
 
KKRT said:
Laa-Yosh meant in terms of lighting precision probably.
And i agree, i dont see nothing spectacular on those shots in terms of lighting precision that would explain high precision buffers in comparison to games that use lower precision buffers.
Click to expand...

Show me a game on any console which does what Infamous does (the shadows, reflections; even if those are not perfectly correct, particle effects, ambient effects, impressive different number of textures & shaders used etc.) in an open world game by using low precision buffers or "optimized" buffers.

Ergo in a open world game, in 2014, to attain the image quality of Infamous by a middle sized team (if you want to make money from your game) you need 85MB for the g-buffers. It's not luxury and certainly not because developers were lazy.

If you begin by limiting or optimizing stuff, like replacing 32 bit assets by 16 bit, replacing 85MB by 40MB or by replacing the very memory consuming SMAA 2TX (not included in those 85MB by the way) by the similarly "optimized" :rolleyes: FXAA solution...

I bet the final result would be not very different from Infamous like some of you said, law of diminushing returns helping. Developers could use this optimized engine on different hardwares and it could look great. It could look like a prince.

But Infamous is king.
 
Globalisateur said:
But Infamous is king.
Click to expand...

Enough pissing contest please.

It's not luxury and certainly not because developers were lazy.
Click to expand...
No one is arguing this. Again, the pdf was clear on future directions and reasonings.


by replacing the very memory consuming SMAA 2TX (not included in those 85MB by the way) by the similarly "optimized" :rolleyes: FXAA solution...
Click to expand...
No one is arguing this either.

Stop making up irrelevant arguments here.
 
Cyan said:
What are compute queues and what advantages do they have?
Click to expand...

They give you more flexibility in how you feed commands to the GPU. Historically GPU's only read from one command buffer at a time, and depending on the commands issued and the work being done it might have to stall and wait for certain commands to fully complete before moving on. With compute queues the GPU can pull from multiple command buffers simultaneously. This means if the GPU is is sitting on some rendering commands from the main command buffer, it can pull some compute work from the compute queues and execute it on its hardware units. It can help keep your ALU's from sitting idle during a frame when you're bottlenecked by fixed-function hardware.
 
Globalisateur said:
Show me a game on any console which does what Infamous does (the shadows, reflections; even if those are not perfectly correct, particle effects, ambient effects, impressive different number of textures & shaders used etc.)
Click to expand...

In my opinion the general visual feel of a game like GTA V is very close - but of course the devil is in the details and that's where Infamous shows its advantage.

Still, the look of the day and night time lighting, the weather effects with shiny wet roads and neon lights and such, even the view distance - most of the stuff was already there on the previous gen systems with much lesser memory.

This is why I was talking about diminishing returns well before the release of the new systems; if you compare games at 320*200 res with downscaled thumbnails, it'd be very hard to tell which is on PS3 and which is on PS4. Just remember how the PS2 -> PS3 jump was much more noticeable, with proper lighting and shadows and material shaders.

As for the actual technical stuff, it's also obvious to me that the devs should be able to achieve a lot more with more efficient use of the system resources. I fail to understand why there's any argument about this...
 
sebbbi said:
Virtual deferred rendering is not used in Trials Fusion. I am not yet ready to talk about our future plans, but I will do a technology interview about Trials Fusion soon. It will give you technical information about our next gen choices (and trade-offs).
Click to expand...

Fair enough. The other tech interviews I've read by you were very informative. Looking foward to it.
 
Laa-Yosh said:
In my opinion the general visual feel of a game like GTA V is very close - but of course the devil is in the details and that's where Infamous shows its advantage.

Still, the look of the day and night time lighting, the weather effects with shiny wet roads and neon lights and such, even the view distance - most of the stuff was already there on the previous gen systems with much lesser memory.

This is why I was talking about diminishing returns well before the release of the new systems; if you compare games at 320*200 res with downscaled thumbnails, it'd be very hard to tell which is on PS3 and which is on PS4. Just remember how the PS2 -> PS3 jump was much more noticeable, with proper lighting and shadows and material shaders.

As for the actual technical stuff, it's also obvious to me that the devs should be able to achieve a lot more with more efficient use of the system resources. I fail to understand why there's any argument about this...
Click to expand...

But comparing ISS to GTA V exactly shows how much the difference really is and how old last gen systems really are. GTA V at daytime has fundamental IQ problems and was hardly playable for me due to all the shimmering (I moaned about this in the GTA V thread). ISS is much better.

Especially IQ is imo a part of the game tec were we are far from reaching diminishing returns.
 
Billy Idol said:
But comparing ISS to GTA V exactly shows how much the difference really is and how old last gen systems really are. GTA V at daytime has fundamental IQ problems and was hardly playable for me due to all the shimmering (I moaned about this in the GTA V thread). ISS is much better.

Especially IQ is imo a part of the game tec were we are far from reaching diminishing returns.
Click to expand...

Agreed, there's some comparisons here between IF2 (2nd gen release) vs I:SS (launch window release)

http://www.dualshockers.com/2014/02...reenshot-comparison-the-leap-from-ps3-to-ps4/

Now add the frame advantage on top and it's a really nice generational leap - I can't wait to see GTA V on PS4.
 
Billy Idol said:
But comparing ISS to GTA V exactly shows how much the difference really is and how old last gen systems really are.
Click to expand...

That's what I'm talking about - most of the differences are "simply" because of much stronger hardware (and more memory). Of course it's still far from an automatic gain, it takes considerable engineering effort to utilize the newer systems.


Also, this is something I've been talking about for a while now: game engines have implemented most of the offline rendering features by now. However, they're all using various rough approximation techniques instead of doing things the "correct" way, like shadow maps instead of raytracing or gaussian filters on the skin texture instead of actually calculating how light is passing through the volume, or screen space reflections etc. etc.

What the new systems have produced so far in rendering technology is mostly about choosing less simplified approximations while increasing asset detail. I don't really feel as any game has done anything really groundbreaking so far - but there's a lot less room to improve already, and we're also quite early into the new generation.
 
Laa-Yosh said:
That's what I'm talking about - most of the differences are "simply" because of much stronger hardware (and more memory).
Click to expand...
To be fair, isn't that true of every generation? How many generations start with tech unseen of before, where devs create completely new techniques on completely new hardware? And it's even harder now because everything potentially new on consoles has already been done before on PC because they're the same design. We can't really expect Sucker Punch or anyone else to introduce whole new rendering techniques in launch software.
 
PS2 to PS3 has brought a metric ton of feature improvements and many of them were at the beginning. Normal mapping and proper lighting/shadows was the most obvious one, nearly impossible and practically nonexistent in the previous generation and an almost default feature on the new one from the start.

Granted, it was all coming down from the PC and to some extent from the first Xbox and Gamecube, but it was a very big and obvious step up in image complexity or visual fidelity or whatever you want to call it.
Still, there were some truly outstanding games within the first year or so... Remember Heavenly Sword for example?

And you seem to agree with me that it isn't really reasonable to expect anything like this nowadays... So it's not me you're really arguing with, you seem to be in agreement instead ;)
 
ultragpu said:
I gotta ask according to this slide here, what does it mean if it says the engine draws 11 million polygons on a regular basis for the entire scene? Is this the number for the green blocks or including the yellow blocks? Can we say the game renders 11 million polygons per frame?
http://www.dualshockers.com/2014/04...-11-million-rendered-regularly-by-the-engine/
Click to expand...
11M per frame would be a plausible figure, I suppose.

Laa-Yosh said:
Normal mapping ... was the most obvious one, nearly impossible and practically nonexistent in the previous generation
Click to expand...
Well, I see someone didn't play much Xbox. :D
 
Last edited by a moderator:
HTupolev said:
11M per frame would be a plausible figure, I suppose.
Click to expand...

I really doubt that. LoD is quite aggressive on the city and characters geometry and 11m is a lot for those consoles.
Seems like peak situation and in unstable condition. Or maybe its a figure before culling, like in GoW:A performance graphs.

Same goes for 120k polys per character quote, where it only applies to Delsin and maybe few other key characters.
 
KKRT said:
I really doubt that. LoD is quite aggressive on the city and characters geometry and 11m is a lot for those consoles.
Seems like peak situation and in unstable condition. Or maybe its a figure before culling, like in GoW:A performance graphs.

Same goes for 120k polys per character quote, where it only applies to Delsin and maybe few other key characters.
Click to expand...

Please could someone properly tag this thread accordingly! ;)
 
You must log in or register to reply here.
Back
Top