Image Quality and Framebuffer Speculations for WIP/alpha/beta/E3 games *Read the first post*

iyYQPSdh2wH7r.png

495ae5-1401031104.jpg

comparing the collar

I see polygon edges here too on his arm.Maybe its just me.
TO1886_WCH_Tracks_23_2014_05_22.jpg
 
Last edited by a moderator:
Can someone tell me why they don't just use tessellation on the collar and sleeve for close up shots? Shouldn't this be an easy scenario to use it.....
 
How big is the performance hit from tessellation on PS4? Might not be worth the tradeoff depending on what else they're doing.
 
It's more of an asset issue, as the type of geometry that works well for standard models will not be good for tessellation. The extra geometry necessary to maintain hard edges will cause efficiency problems if the tessellation is not used at all.

This has already been discussed to death on this forum - the dev needs to decide to build the asset with tessellation required from the start, as there's no way to just flip a switch and get proper results. Just like COD Ghosts had it always on on weapons and human heads, because the assets were created with the feature in mind.

There's still time to optimize the models before release, especially if the game has been pushed to 2015. But if tessellation wasn't a part of the pipeline from the start then it's unlikely to get added as it'd require reworking a lot of the existing assets.
 
It's more of an asset issue, as the type of geometry that works well for standard models will not be good for tessellation. The extra geometry necessary to maintain hard edges will cause efficiency problems if the tessellation is not used at all.

This has already been discussed to death on this forum - the dev needs to decide to build the asset with tessellation required from the start, as there's no way to just flip a switch and get proper results. Just like COD Ghosts had it always on on weapons and human heads, because the assets were created with the feature in mind.

There's still time to optimize the models before release, especially if the game has been pushed to 2015. But if tessellation wasn't a part of the pipeline from the start then it's unlikely to get added as it'd require reworking a lot of the existing assets.

Thanks for the answer, the weird thing is that this doesn't seem to be a problem on the older models. I was just thinking that for a game being released in 2014/2015, tessellation would be part of the asset creation process by now. With the increased clarity that higher resolution allows, these type of thing become more and more noticeable and work against the immersion of the game, but hopefully is like you say and there is still enough time to optimize.
 
I still can't understand this fascination with tessellation. The tech is only really worth using if you can afford to render subpixel triangles and also use high precision displacement mapping (float 1-channel maps) so that you can provide additional geometric detail. Current console hardware is just completely incapable of running that efficiently enough; some monster videocards can use it well enough in DCC apps but then you only work on a single model there with no animation and such.

Without displacement you'll get unnaturally smooth surfaces most of the time and no silhouette detail.

Working for tessellation however presents a significant overhead in the asset creation process, particularly on hard surface models (armor, weapons, SF stuff). I'd say the extra time to build a technical model may easily amount to +100% work. It'll also make for a less efficient use of the polygon count as hard edges need to be beveled to keep the form from smoothing out.

This is what I'd consider a nice tidy model, after subdivided:
molegun_backwireframe.jpg

Notice just how many extra geometry is used to control the subdivision, and just how small some will get on screen.

It's much cheaper and faster and more hw efficient to not use tessellation on complex models at all. Certain pieces may benefit but that'll also complicate the workflow.
 
Yes, i rather them not use tessellation and use that processing power to have better shaders for example. You will notice that improvement much quicker.

In regards to the collar comparison some of us made, like Laa-Yosh said, i understand that the game is still in development and will surely be fixed since there are more upgrades rather than downgrades. Game looks stunning, has that soft cg-look to it. Can't wait for Tuesdays footage hopefully.
 
Yeah the assets have been downgraded slightly in terms of poly count but improvements were made else where too such as better volumetric lighting that creates shadows now, also bloom on the particles.
 
What is volumetric about the lighting on any of the screens?
Bloom on particles is free.

Shaders improvements were probably costly, but bloom? come on.
 
What is volumetric about the lighting on any of the screens?
Bloom on particles is free.

Shaders improvements were probably costly, but bloom? come on.

I am curious to hear your opinion tomorrow on what got upgraded/downgraded when the hq trailer releases and hopefully some png screens like eurogamer got with driveclub this week. So far we can see polygonal edges on the collar of the uniform, anything else? Shaders and post processing are better right?
 
It was mentioned by one of the devs that, now, on top of the 4X MSAA there is a temporal component (TXAA), they are dead serious about the IQ in this game. :oops:
 
Still, the poly count difference on the collar/sleeve is quite striking. Surely the PS4 is able to render those surfaces with a lot more polygons??
 
In that scene, yes, but the model may be used in situations with far more people. It could also be an LOD bug in that build. The only fact at this point is the models exhibit reduced polygon counts, without enough detail to speculate on reason.
 
Screen capture from the trailer thus low quality, but judging from the jaggies on the wooden plank it looks non native 1080p. Anyone else agree?
2544147-2.jpg
 
Back
Top