I just spent over an hour watching skin, eyes and hair on the internet.
(Great stuff)
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Game Graphics Technology | 64bit, procedural, high-fidelity debating
- Thread starter Dictator93
- Start date
(Great stuff)
Thx. It should get even more structured in time hopefully! I just wish more people would post
---
GDC 2016 talks are starting to pop up and they are looking already pretty interesting:
Object Space Rendering in DirectX 12 sounds awesome.
Also interesting to see that Timothy Lottes is working on HDR displays.
-----
There will be new, big release of CryEngine version in March, so expect some nice tech trailer from Crytek on GDC.
http://www.cryengine.com/news/cryengine-s-new-release-cycle
Something I've been wondering about for a while is the question of 4k potential in the next gen consoles. I'm of the mind that thinks that in order to hit a $399 price point, native 4k with an expected next gen leap in fidelity would be unlikely. However after getting Rainbow Six: Siege, I think it's MSAA reconstruction technique could be quite beneficial on consoles next gen. Being someone who is primarily a console gamer (can't run anything more complex than Source games on my laptop) I would love to see widespread use of these techniques next gen in order to prevent scaling artifacts so we don't have another gen like the last where everything has to be scaled in some way.
For those of you unaware on consoles Siege renders at a halved resolution in order to hit 60fps in the competitive mode and uses the data from a 2x MSAA sample to output at 1080p on PS4 and 900p on Xbox One. I do not have experience with the Xbox One version but on PS4 the final output is shockingly convincing, even in motion.
Example:
There is some artifacting but it's mostly only visible on high contrast edges and when there is a lot of motion. It's relatively minor and looks more reminiscent of aliasing than a more distracting aberration.
Example (around the gun):
Some of you may remember that Killzone: Shadow Fall's multiplayer mode attempting to do something similar. Guerrilla's technique used alternating lines from temporal sampling in order to reconstruct a 1080p image and I was personally not a fan of it. It didn't save enough performance in order to justify the overall loss in clarity and it was a less temporally stable image than what Siege is doing.
I'd love to get more information on methods like this and their practical application. Dictator93 pointed me to this piece by Matt Pettineo from Ready At Dawn. https://mynameismjp.wordpress.com/2015/09/13/programmable-sample-points/
If anyone has more insight into these or similar techniques I'd like to hear about it, examples would be excellent as well. Do you agree that it would be a useful pursuit for consoles in the next generation? I think it would be, especially with a much higher resolution for the original image.
For those of you unaware on consoles Siege renders at a halved resolution in order to hit 60fps in the competitive mode and uses the data from a 2x MSAA sample to output at 1080p on PS4 and 900p on Xbox One. I do not have experience with the Xbox One version but on PS4 the final output is shockingly convincing, even in motion.
Example:
There is some artifacting but it's mostly only visible on high contrast edges and when there is a lot of motion. It's relatively minor and looks more reminiscent of aliasing than a more distracting aberration.
Example (around the gun):
Some of you may remember that Killzone: Shadow Fall's multiplayer mode attempting to do something similar. Guerrilla's technique used alternating lines from temporal sampling in order to reconstruct a 1080p image and I was personally not a fan of it. It didn't save enough performance in order to justify the overall loss in clarity and it was a less temporally stable image than what Siege is doing.
I'd love to get more information on methods like this and their practical application. Dictator93 pointed me to this piece by Matt Pettineo from Ready At Dawn. https://mynameismjp.wordpress.com/2015/09/13/programmable-sample-points/
If anyone has more insight into these or similar techniques I'd like to hear about it, examples would be excellent as well. Do you agree that it would be a useful pursuit for consoles in the next generation? I think it would be, especially with a much higher resolution for the original image.
Something I've been wondering about for a while is the question of 4k potential in the next gen consoles. I'm of the mind that thinks that in order to hit a $399 price point, native 4k with an expected next gen leap in fidelity would be unlikely. However after getting Rainbow Six: Siege, I think it's MSAA reconstruction technique could be quite beneficial on consoles next gen. Being someone who is primarily a console gamer (can't run anything more complex than Source games on my laptop) I would love to see widespread use of these techniques next gen in order to prevent scaling artifacts so we don't have another gen like the last where everything has to be scaled in some way.
For those of you unaware on consoles Siege renders at a halved resolution in order to hit 60fps in the competitive mode and uses the data from a 2x MSAA sample to output at 1080p on PS4 and 900p on Xbox One. I do not have experience with the Xbox One version but on PS4 the final output is shockingly convincing, even in motion.
Example:
There is some artifacting but it's mostly only visible on high contrast edges and when there is a lot of motion. It's relatively minor and looks more reminiscent of aliasing than a more distracting aberration.
Example (around the gun):
Some of you may remember that Killzone: Shadow Fall's multiplayer mode attempting to do something similar. Guerrilla's technique used alternating lines from temporal sampling in order to reconstruct a 1080p image and I was personally not a fan of it. It didn't save enough performance in order to justify the overall loss in clarity and it was a less temporally stable image than what Siege is doing.
I'd love to get more information on methods like this and their practical application. Dictator93 pointed me to this piece by Matt Pettineo from Ready At Dawn.
If anyone has more insight into these or similar techniques I'd like to hear about it, examples would be excellent as well. Do you agree that it would be a useful pursuit for consoles in the next generation? I think it would be, especially with a much higher resolution for the original image.
Next trials game will be doing a somewhat similar trick.
They discuss it and the the differences between it and RS:Siege's trick in this thread.
https://forum.beyond3d.com/threads/gpu-driven-rendering-siggraph-2015-follow-up.57240/
Gemüsepizza
Member
Does anyone know if HDR will inflict a performance penality? With DisplayPort 1.3 coming later this year, there will be two new and interesting types of display:
4K HDR with 60hz
4K SDR with 120hz
But 120hz at such a resolution is obviously not really feasible for quite some time, at least not for AAA gaming with very high settings, even with the new GPU generation starting soon. If HDR does not come with a performance penality, this could make those displays very attractive.
4K HDR with 60hz
4K SDR with 120hz
But 120hz at such a resolution is obviously not really feasible for quite some time, at least not for AAA gaming with very high settings, even with the new GPU generation starting soon. If HDR does not come with a performance penality, this could make those displays very attractive.
Games already render to HDR buffers, so the cost of HDR display shouldn't be that big.Gemüsepizza;193837325 said:
What will be interesting to see is how much HDR will magnify aliasing.. (And thus it might need more processing power to get decent overall quality..)
We should know more after GDC and Timothy Lottes presentation.
Gemüsepizza
Member
Thanks, just read a few articles on his blog and saw this:
Ugh, that's so disappointing lol. I just want a nice clean picture, why is this stuff so hard.
Ugh, that's so disappointing lol. I just want a nice clean picture, why is this stuff so hard.
SeeNoWeevil
Member
Am I alone in thinking 4K is going to massively stifle graphics tech? If developers are aiming for 2160p30 on next-gen consoles then I think we're in for a very underwhelming generation indeed. 4K is just a massive burden I personally do not want.
Consoles are struggling to hit 1080p/30fps. Most gaming PCs (based on Steam survey) are now able to handle 1080p/60fps. I honestly don't see next-gen consoles hitting 4K, nevermind 2K, while trying to be affordable and generate a profit for their company. I have a 980Ti and these new games coming out make it hard for me to hit 1440p(2K)/60fps steady - and that GPU cost me what amounts to two current-gen consoles; which doesn't take into account the rest of the computer, components, shipping, packaging, etc. I think next-gen's will prioritize 1080p/60fps gaming, I can see that attainable at a reasonable price to consumers. I wouldn't expect 2K support until next-next-gen.
I was wondering if anyone here with more in-depth knowledge knows when actual full scene raytracing in games will be a possibility? How far are we away from it becoming a regular thing in PC games? Will DX12 make it more manageable? Do you think a GTX 1080 (or whatever it will be called) could do it?
A minor quibble, but 1920x1080p is actually "2K", but no one uses it since 1080p rose to popularity before 4K was the next big marketing term.
But, I would say you are definitely right in predicting that 4K will not make it consoles in a quite a while. At least in my experience with UE/Unity and the consoles, the renderers use a lot of effects that scale heavily with resolution (it's not just video memory usage) and the step-up in performance requirements is massive, so I doubt that the next generation will be able to cope with it and the more and more complex rendering solutions.
Dictator93
Member
I hope this year the GDC presentations are rather quickly online, last year a number of them never made it even after the wait period :/
Does the vault access for GDC include taped precedings from each session? It would be something I would consider if that were the case.
---
OT - I am curious how HDR screens will actually affect VR, it could increase aliasing, but also combat problems of believability. Do VR games have post processing for lighting in general? Or is that against comfort-guides... For example, would post process bloom be something desirable in VR? Is such a thing a good idea? Especially on LDR monitors? Are things like eye adaptation a no-no? Or perhaps good to prevent eye fatigue?
Dictator93
Member
More GDC talks are coming online and some caught my eye:
Photogrammetry in Star Wars Battlefront
Rendering Mirror's Edge Catalyst
Photogrammetry in Star Wars Battlefront
Rendering Rainbow Six Siege
Rendering Mirror's Edge Catalyst
Temporal Reprojection Anti-Aliasing in Inside
Taming the Jaguar: Insomniac Games
Optimising the GPU pipeline with Compute
Mixed Resolution Rendering in Skylanders
Global Illumination in Tom Clancy's Division
Great thread. Not sure if this is the best place to ask but I'll ask anyway.
Seems like a few console games are using dynamic resolution scaling to maintain 60fps etc. How hard is this to incorporate into an engine? Is this something that 3rd party engines have already or can be modified to use? Or is this something that needs to be planned from the beginning of creating a custom engine?
And finally, any PC games use this? If not, why not? Seems like a good solution to maintain a solid framerate.
Seems like a few console games are using dynamic resolution scaling to maintain 60fps etc. How hard is this to incorporate into an engine? Is this something that 3rd party engines have already or can be modified to use? Or is this something that needs to be planned from the beginning of creating a custom engine?
And finally, any PC games use this? If not, why not? Seems like a good solution to maintain a solid framerate.
All id tech 5 games use it. People generally turn it off, but for low end machines, its quite decent feature. But it definitely needs some additional settings, like how far resolution drop can go for example.
Thanks for the info. Id love to see this become a common option.
id tech 5 is being phased out though right?
Sooo what's happening with the lighting here?
http://xboxdvr.com/gamer/Mr Noobcraft/video/14960248
http://xboxdvr.com/gamer/Mr Noobcraft/video/14960248
Smushroomed
Banned
Cool thread
A.I character pissing about with the light switch.
CyberPunked
Member
Interesting that they're doing a talk on this, considering it's a feature that has been almost completely cut from the game.
These all sound super interesting. I know Naughty Dog is doing a talk at GDC as well. Is it a tech or design talk? I'd love for a look at the final tech of Uncharted 4 shortly before launch.
One tech art (with production tips), one texturing pipeline (with Allegorithmic):
Texturing Uncharted 4: a matter of Substance (presented by ALLEGORITHMIC)
Technical Art Culture of Uncharted 4
Too bad I couldn't fit a trip to GDC to my schedule and budget this year, though I do hope to have a chat with some tech developers in European studios at GDCEU (even if the event itself is merely a shadow of the main event).
You don't really "almost cut off" feature; it either is in or not, and GI itself definitely seems to be in the game even if the effect itself isn't that noticeable. In any case, it's an area that definitely needs more research in and hopefully the talk will shed some light into how the complete feature compares against the early implementation for the vertical slice.
Coolcoolcool
CyberPunked
Member
Not just the verticle slice. Footage from early last year showed a much stronger presence of GI. Then again it's not as though we don't know why it was cut :/
Treechopper
Member
Man, I was planning to go to GDC this year and had to cancel. It was going to be my first trip
Dictator93
Member
It seem sa lot of the DX12 and Vulkan presentations have AMD partnetships (Quantum Break as well!)
It should be an interest next few months regarding vendor performance
It should be an interest next few months regarding vendor performance
Dictator93
Member
A great visual and step by step guide through the very basics of rendering pixels. I found it pretty nifty.
Do you think the Xbox SDK has received some significant improvements as we see more games coming out in 1080p? I initially thought there was a physical limitation (more or less) with the ESRAM having barely enough space for a 1080p buffer, but I'm curious if they have found a way to come around that via software? The Division (open world), Hitman is now also confirmed 1080p with the same assets as PS4, Quantum Break is confirmed 1080p, Gears of War 4 campaign is said to be 1080p/30. Also, speaking of The Division, they said the engine will be used for other Ubisoft games, so could that mean the next Assassin's Creed will be 1080p for both consoles?
Paging Dictator93.
Paging Dictator93.
Dictator93
Member
Heyo!
Sorry everyone that I have not updated the front page in a bit (last time I did was a bout a month ago). Unfortunately real life has made the pursuit of my passions difficult.
And that is where quantum break and the new gears game will probably be made possible IMO. I am really curious about Gears though especially as UE4 is really unproven on new consoles, and I think its base rendering costs for things like realtime shadows and post-processing makes the possibility of a 1080p/30 somewhat questionable. Perhaps they will have to render many elements at lower than output resolution to achieve that (SSAO, SSR, etc). They even apparently have added volumetric lighting to the UE4 branch... lot's of expensive stuff to have at 1.3 TF at 1920X1080! The last game with a similar suite of rendering effects on xb1 that I can think of is Ryse (obmb, global volumetric lighting, visible bokeh, etc).
bokeh
All of which look to be rather high qualty / high internal resolution in the first video they released, which is atypical of what he have seen for similar effects in other UE4 game son xb1, other engines on xb1, or even nearly all games on PS4.
Basically, I am curious to see what compromises will have to be made for both if they do want to hit 1080 on xb1. We already know the compromises of QB to a certain extent (many 720p buffers with a 1080p output).
Do you really think they would toss out all the iterations to Anvil Next and start using Snowdrop?
----
Maybe my post is satisfactory in answering your question, maybe not! It is a lot of speculation of course, but in the end I would conservatively say that it is more reasonable to assume that the xb1's achievements in hitting certain resolutions seems to be up to dev's and their design priorities rather than deft overarching decisions from MS.
Sorry everyone that I have not updated the front page in a bit (last time I did was a bout a month ago). Unfortunately real life has made the pursuit of my passions difficult.
Heyo!
I obviously would not know directly, but I think there are a convergence of things in general where by the largest part would be devs coming to a greater understanding of how to manage for the xb1's shortcomings. I would not doubt that MS has provided guidelines for how to manage the ESRAM depending upon different usecases.
ESRAM should still be one of the main factors getting in the way of a game reaching 1920X1080 if it has a traditionally fatter framebuffer (I think but cannot know completely, but I believe this is the main reason fox engine games are 1280X720 for example). Even then you come into the problem in which games also hit shading and other GPU limitations if they do not have the afforementioned problem. Devs have to know where to cut corners IMO, and that is one of the reasons where native 1080p games start becoming a thing on xb1.
One must not forget that the Division is only 1920X1080 sometimes. So even if they managed to have ESRAM work for their framebuffer size in a performant manner, the oomph is missing elsewhere to maintain that resolution in terms of raw shading / bandwidth performance all the time. This is where I again want to mention the basic thing that has probably been happening: devs understanding where and how to cut corners to mitigate large and visible quality drops. This means often lower resolution internal rendering for lots of effects instead of the final output resolution.
And that is where quantum break and the new gears game will probably be made possible IMO. I am really curious about Gears though especially as UE4 is really unproven on new consoles, and I think its base rendering costs for things like realtime shadows and post-processing makes the possibility of a 1080p/30 somewhat questionable. Perhaps they will have to render many elements at lower than output resolution to achieve that (SSAO, SSR, etc). They even apparently have added volumetric lighting to the UE4 branch... lot's of expensive stuff to have at 1.3 TF at 1920X1080! The last game with a similar suite of rendering effects on xb1 that I can think of is Ryse (obmb, global volumetric lighting, visible bokeh, etc).
bokeh
spot light volumetric
what looks like global volumetric lighting
obmb
All of which look to be rather high qualty / high internal resolution in the first video they released, which is atypical of what he have seen for similar effects in other UE4 game son xb1, other engines on xb1, or even nearly all games on PS4.
Basically, I am curious to see what compromises will have to be made for both if they do want to hit 1080 on xb1. We already know the compromises of QB to a certain extent (many 720p buffers with a 1080p output).
I do find the Divisions engine as being really well balanced as the AC engine obviously overshot the gen in its first outing... hence the 1600X900 and barely 30 it managed. Then again, AC has really different requirements than the Division in terms of AI routines and animation (there need to be way more animated characters on screen at once in AC than there are in the division).
Do you really think they would toss out all the iterations to Anvil Next and start using Snowdrop?
----
Maybe my post is satisfactory in answering your question, maybe not! It is a lot of speculation of course, but in the end I would conservatively say that it is more reasonable to assume that the xb1's achievements in hitting certain resolutions seems to be up to dev's and their design priorities rather than deft overarching decisions from MS.
Interesting, so render some elements at lower resolution to save buffer space. Thanks, and it will certainly be interesting to see where this goes, if indeed more games start coming out at 1080p on the Xbox. I'm personally hoping Dues Ex will get to 1080p as I pre-ordered the Xbox version. I can always switch if they do mention the resolution prior to release.
PowerVR at GDC 2016: Vulkan and OpenGL ES ray tracing.
https://www.youtube.com/watch?v=ffHfJuRFF1s
Apartment demo: https://www.youtube.com/watch?v=uxE2SYDHFtQ
http://blog.imgtec.com/powervr/gdc-2016-ray-tracing-graphics-mobile
https://www.youtube.com/watch?v=ffHfJuRFF1s
Apartment demo: https://www.youtube.com/watch?v=uxE2SYDHFtQ
http://blog.imgtec.com/powervr/gdc-2016-ray-tracing-graphics-mobile
Cow Mengde
Banned
I wonder if we'll ever get ray tracing in games. How many years off are we?
Rays can be used in lots of techniques. The Division on PC features Hybrid Frustum Traced Shadows right now.
When VR headsets get eye tracking and foveated rendering, do you think their eye tracking could pass through to benefit old flat-screen games running in virtual cinema mode? e.g. the headset's layer knows where the player is looking on the virtual movie screen, and that gets passed to the driver which handles rendering non-focus regions at lower resolution without the game knowing about any of it.
Could actually be a strong case for virtual cinema mode if that concept works.
Could actually be a strong case for virtual cinema mode if that concept works.
Just to bump this thread since there hasn't been any interesting discussion for a long time. Now we know that Star Wars Battlefront is using Enlighten for its global illumination solution. Here's the link for confirmation:
http://www.geomerics.com/news/enlighten-in-star-wars-battlefront/
Anyway, my question is, is the GI solution real time or baked? Well, I tried to verify this by going under an area with color bleeding in effect. Here's a pic of it, not the best, but I can't take a pic of it now:
I'm referring to the orange hue under the sheet. I went under it and the character was completely unaffected by it. No orange hue is cast on the character. So I'm guessing that means the GI is baked. Can anyone who understands GI better leave some thoughts?
http://www.geomerics.com/news/enlighten-in-star-wars-battlefront/
Anyway, my question is, is the GI solution real time or baked? Well, I tried to verify this by going under an area with color bleeding in effect. Here's a pic of it, not the best, but I can't take a pic of it now:
I'm referring to the orange hue under the sheet. I went under it and the character was completely unaffected by it. No orange hue is cast on the character. So I'm guessing that means the GI is baked. Can anyone who understands GI better leave some thoughts?
Enlighten has been pretty much just a baked solution until very recently. GDC they demoed their real time GI in UE4 but as far as I know Frostbite isn't using it yet.Just to bump this thread since there hasn't been any interesting discussion for a long time. Now we know that Star Wars Battlefront is using Enlighten for its global illumination solution. Here's the link for confirmation:
http://www.geomerics.com/news/enlighten-in-star-wars-battlefront/
Anyway, my question is, is the GI solution real time or baked? Well, I tried to verify this by going under an area with color bleeding in effect. Here's a pic of it, not the best, but I can't take a pic of it now:
I'm referring to the orange hue under the sheet. I went under it and the character was completely unaffected by it. No orange hue is cast on the character. So I'm guessing that means the GI is baked. Can anyone who understands GI better leave some thoughts?
That's strange. I thought Enlighten proudly claimed that their GI solution was real time for a long time now. Also, UE4 has real time GI now? Last I heard it was only using Lightmass.
Really? I only remember them talking about how much faster it is to bake with them. Could you give me a link?
Enlighten is compatible with UE4, I don't think it's a part of the engine.
Metalingus5150
Member
I wonder how Naughty Dog was able to achieve their volumetric lighting solution in Uncharted 3 and Last of Us. I thought it looked really good.
Here, this hasn't been updated for a while, I believe:
http://www.geomerics.com/enlighten/
Unless what they mean by real time is still essentially baking indirect lighting in real time, in which case, it confused me.
That test doesn't really test if game has real-time GI, just if there is lightprobe placed under the orange object.Just to bump this thread since there hasn't been any interesting discussion for a long time. Now we know that Star Wars Battlefront is using Enlighten for its global illumination solution. Here's the link for confirmation:
http://www.geomerics.com/news/enlighten-in-star-wars-battlefront/
Anyway, my question is, is the GI solution real time or baked? Well, I tried to verify this by going under an area with color bleeding in effect. Here's a pic of it, not the best, but I can't take a pic of it now:
I'm referring to the orange hue under the sheet. I went under it and the character was completely unaffected by it. No orange hue is cast on the character. So I'm guessing that means the GI is baked. Can anyone who understands GI better leave some thoughts?
Lightmass and pretty much all completely baked methods can pass the test.
Better would have been to see if moving lightsource gives bouncelight. (Which dynamic enlighten can do.)
So another post from me today in my attempt to revitalize this thread briefly.
Now Batman Arkham Knight has a model viewer mode that lets you get up close to the character models to appreciate all the details. The mode is in complete pitch dark with only a spotlight on the character model. So I've always been wondering if the character models here are the gameplay/cutscene models or entirely different models with higher fidelity. To put that to test, I made this comparison:
The first pic is a Photomode shot, which as far as I know, doesn't improve the gameplay model, with DOF cranked up to focus only on Batman. So, in my opinion, at least, both models are very similar, different lighting conditions aside. The model viewer mode probably uses higher quality AA, though. Thoughts anyone?
I also have genuine questions to ask.
Now Batman Arkham Knight has a model viewer mode that lets you get up close to the character models to appreciate all the details. The mode is in complete pitch dark with only a spotlight on the character model. So I've always been wondering if the character models here are the gameplay/cutscene models or entirely different models with higher fidelity. To put that to test, I made this comparison:
The first pic is a Photomode shot, which as far as I know, doesn't improve the gameplay model, with DOF cranked up to focus only on Batman. So, in my opinion, at least, both models are very similar, different lighting conditions aside. The model viewer mode probably uses higher quality AA, though. Thoughts anyone?
So I know double posting is frowned upon but my last post got no replies and I have another question, so....
I've been playing The Witcher 3 and I'm close to the ending, 85 hours in or so. Anyway, the game looks great on PS4 and I think the water looks great, but I'm wondering, does the water use screen space reflections? It certainly behaves like SSR as the reflected objects disappear from the surface when you pan the camera to move away from said objects.
I've been playing The Witcher 3 and I'm close to the ending, 85 hours in or so. Anyway, the game looks great on PS4 and I think the water looks great, but I'm wondering, does the water use screen space reflections? It certainly behaves like SSR as the reflected objects disappear from the surface when you pan the camera to move away from said objects.
Jedipottsy
Member
A minor quibble but 2k is a defined resolution - 2048x1080 as defined by DCI. 4k gets a little complicated as DCI refer to 4k as 4096x2160 - double 2k, while UHD-1 is 3840x2160. Depending on you point of view both are technically 4k, but from different bodies, whereas 2k only ever referred to a resolution of 2048x? (There are variations of 2k with differing horizontal resolutions, such as 2048x1556 anamorphic), none of which included 1080p as '2k'. Although many still consider 4k the DCI standard of 4096x2160, and TVs are UHD 4k.
Dictator93
Member
Hey Javin,
sorry for not responding at all to your last quesiton!
Yeah, TW3 uses SSR... just seemingly for water surfaces (which is arguably a pretty hard area to do it, since your are projecting them on a transparency). TW3 very rarely has extra smooth glossy metal surfaces to see if the game uses SSR elsewhere, but I do not think it does.
Sand Bar reflected
Sand Bar not reflected
And like you said, the best way to check (beyond looking for occlusion errors and stuff) would be to just pan the camera down below the object in question along the surface to see if its reflection disappears.
So another post from me today in my attempt to revitalize this thread briefly.I also have genuine questions to ask.
Now Batman Arkham Knight has a model viewer mode that lets you get up close to the character models to appreciate all the details. The mode is in complete pitch dark with only a spotlight on the character model. So I've always been wondering if the character models here are the gameplay/cutscene models or entirely different models with higher fidelity. To put that to test, I made this comparison:
/snip
I am not sure if the AA is different, rather if just the lighting conditions and lack of sub-pixel detail in the distance make the AA coverage appear that much better. But yeah, I think the model is the same in gameplay and elsewhere. You can especially tell in the intro seuqence of the game when you get the new bat suit and it quickly zooms in on batman's suit then cuts immediately to gameplay.
Thanks a lot, man! I was quite sure it was SSR in The Witcher 3, but I wanted confirmation. I tried looking around the Internet, but I couldn't find a definitive answer, so I decided to ask it here. And you came to the rescue, thanks again!Hey Javin,
sorry for not responding at all to your last quesiton!
Yeah, TW3 uses SSR... just seemingly for water surfaces (which is arguably a pretty hard area to do it, since your are projecting them on a transparency). TW3 very rarely has extra smooth glossy metal surfaces to see if the game uses SSR elsewhere, but I do not think it does.
Sand Bar reflected
Sand Bar not reflected
And like you said, the best way to check (beyond looking for occlusion errors and stuff) would be to just pan the camera down below the object in question along the surface to see if its reflection disappears.
I am not sure if the AA is different, rather if just the lighting conditions and lack of sub-pixel detail in the distance make the AA coverage appear that much better. But yeah, I think the model is the same in gameplay and elsewhere. You can especially tell in the intro seuqence of the game when you get the new bat suit and it quickly zooms in on batman's suit then cuts immediately to gameplay.
Also, yeah, The Witcher 3 has very few, if any, glossy surfaces around the world and I haven't seen SSR anywhere else besides the water surface. I'm actually curious to see if a better reflection technique will come that removes the biggeet limitation of SSR.As for Batman Arkham Knight, your observations do make sense. The character models appear cleaner overall, but there are still some nasty shimmering on some of them. The Man Bat's character model, especially. Hair is also a victim of shimmering. In any case, pretty impressive that the gameplay models look that good.
Also, if you don't mind, there will be a new trailer for Uncharted 4 later and it would be nice if you could post your impressions on the visual fidelity here.
Classic planar reflections and realtime cubemaps don't suffer from SSR's limitation, although they have applicability and efficiency challenges.
Some modern games (Tomorrow Children is a decent example) have been playing around with generating low-frequency volumetric irradiance maps in real time, which can allow for a pretty general-purpose solution, although it basically always takes a lot of power and memory to get such things to offer particularly sharp reflections.