Ray casting can be pretty cool. Variable rate shading has some nice applications. Deep neural networks inside games can be powerful.
Deep learning super sampling however is the holy grail. Or at least a version of it will be. Hence from here on I'll just say deep learning max settings emulation (DLMSE).
At this point we've all seen examples of neural networks increasing the resolution of pictures or interpolating frames, but the secret sauce of DLMSE is that the neural network is trained per game.
The setup is simple. Develop a variant of the game that renders the game to two separate screens/files. One screen has the game running at max settings beyond what the best gaming computers can do, while the other screen has the game render at average or even lower settings. Worse resolution, worse anti-aliasing, worse textures, worse geometry, worse shadows, etc. Could even lower the framerate if its not too much work on the devs.
Thousands of hours of gameplay are then recorded and shipped off to the cloud for neural net training. The net has the clearly defined goal of making the lower quality video look like the max quality video. It may take a lot of time/processing power but this is something that neural nets are fairly good at.
Once trained, the neural network can run on player computers and consoles to bump up the game graphics with only a minor performance hit. A PC with a 2 teraflop GPU could make the game look better than a 10 teraflop GPU running the game natively.
It even has applications for game streaming. The cloud can stream players a low bandwith lower resolution video and local hardware using DLMSE could bump it back up to highest quality 4k pixels.
For this reason, I predict next next generation consoles will include dedicated hardware for deep neural networks, just like new Teslas.
This doesn't sound too applicable considering how many hours it needs to sample.
This doesn't sound too applicable considering how many hours it needs to sample.
Why not?
You could just have a few people playing the game like the Quality Assurance people, and simply feed that into the network.
This doesn't sound too applicable considering how many hours it needs to sample.
It's already a thing.
Even in its current form where only resolution and aliasing are altered, it's extremely effective. Plus like Pi says it would be possible to integrate it with parts of QA and/or automate it.
Yeah I hope games go with 1440p Dlss next gen rather than native 4k. Native 4k is just not worth it. The performance gains are just insane and the picture quality is extremely impressive.
I don't think people know how much QA time is necessary for a project as large as a game...
Plus do you want the amount of glitches that pop up early development to get put into the A.I.?
They'd have to do it once QA is finished.
I don't think people know how much QA time is necessary for a project as large as a game...
Plus do you want the amount of glitches that pop up early development to get put into the A.I.?
They'd have to do it once QA is finished.
Yeah it would not be used during all of QA. That wouldn't even make sense considering a lot of QA isn't playtesting and happens during development.
We're only talking about the very last qa moments.
Yeah it would not be used during all of QA. That wouldn't even make sense considering a lot of QA isn't playtesting and happens during development.
We're only talking about the very last qa moments.
Yeah, that's normally the last month or so. This adds on quite a bit more than a month from the sounds of it.
Yeah, that's normally the last month or so. This adds on quite a bit more than a month from the sounds of it.
Why would it?
It'd be done in parallel.
Why would it?
It'd be done in parallel.
Do what in parallel? You'd still have to have people play through the game an immense amount of hours. It's gonna take longer than the final QA itself.
Do what in parallel? You'd still have to have people play through the game an immense amount of hours. It's gonna take longer than the final QA itself.
How much time does QA generally take for a big game?
Xbox Series X it is already mentioned that there's hardware to support machine learning for improving visuals.
Wonder if PS5 will support something similar.
How much time does QA generally take for a big game?
Ideally QA is time and a half. For IT. For games? I can't say for sure. Just going by hearsay, I would imagine a year in a 3 year dev cycle. Final polish I would imagine is 6 months. I think Killzone Shadowfall was something like that. They had finished the game 6 months before release it I recall. They spent the rest of the time polishing.
How much time does QA generally take for a big game?
i don't know games, but for my software QA is around 40 percent of the R&D budget.
Ideally QA is time and a half. For IT. For games? I can't say for sure. Just going by hearsay, I would imagine a year in a 3 year dev cycle. Final polish I would imagine is 6 months. I think Killzone Shadowfall was something like that. They had finished the game 6 months before release it I recall. They spent the rest of the time polishing.
QA starts on day 1,.. not just the final months of hardening.
This video is kinda dumb since DLSS looks better than native 4K so of course it'd look better than checkerboard rendering, but it's still a good update on just how amazing dlss is.
Almost all of the tech mumbo coming out of next gen is either stupid or overhyped imo, as I have said for months haha. This is definitely the exception. If one console ends up more capable with whatever this approach evolves into, that console will have better looking multiplats 95% of the time.
This video is kinda dumb since DLSS looks better than native 4K
This is one of those weird statements that I know is basically true, but it sounds like it makes no sense.
This is one of those weird statements that I know is basically true, but it sounds like it makes no sense.
It's basically just because the native 4k render has aliasing. Temporal anti-aliasing is great at fixing that but dlss is even better.
So DLSS 2 doesn't need the A.I. playing the game anymore? X amount of previous frames is enough of an analysis?
So DLSS 2 doesn't need the A.I. playing the game anymore? X amount of previous frames is enough of an analysis?
The AI doesn't learn in real time no. It still needs to be trained.
X amount of previous frames is how it increases the level of detail in the scene. Instead of just scaling up one frame, it has more info to work with.
The AI doesn't learn in real time no. It still needs to be trained.
X amount of previous frames is how it increases the level of detail in the scene. Instead of just scaling up one frame, it has more info to work with.
I was reading it doesn't need to be trained anymore? But then again those were comments from ... Reddit or Era.
I was reading it doesn't need to be trained anymore? But then again those were comments from ... Reddit or Era.
i saw it in that "marketing" video posted by the guys that make DLSS. it was part of their sales pitch.
...but i think it still needs "training",.. just that it doesn't need people to spend time assisting the training. it just learns by itself or something.
i saw it in that "marketing" video posted by the guys that make DLSS. it was part of their sales pitch.
...but i think it still needs "training",.. just that it doesn't need people to spend time assisting the training. it just learns by itself or something.
I think it learns from the previous frames, which is why it needs those massive amounts of tensor cores.
I think it learns from the previous frames, which is why it needs those massive amounts of tensor cores.
Tensor cores are needed because it needs to do a lot of calculations over a lot of pixels.It does not learn while you play.
Sony patents tech similar to Nvidia DLSS. | ResetEra (https://www.resetera.com/threads/sony-patents-tech-similar-to-nvidia-dlss.258075/)
(https://www.notebookcheck.net/fileadmin/Notebooks/News/_nc3/DLSS_2.0_12.jpg) Sony patents tech similar to Nvidia's DLSS, could include it in PS5 consoles to achieve higher FPS counts in 4K games (https://www.notebookcheck.net/Sony-patents-tech-similar-to-Nvidia-s-DLSS-could-include-it-in-PS5-consoles-to-achieve-higher-FPS-counts-in-4K-games.483457.0.html) Sony already has something similar to Nvidia's DLSS with the checkerboard renderer implemented in the PS4 console, but the results are not really on par because of the limited hardware. An improved version integrated with the PS5 could benefit from the more powerful AMD hardware, ensuring that... (https://www.notebookcheck.net/fileadmin/templates/nbc_v4_4/images/logo_nbc_small_shaded.svg) www.notebookcheck.net
[/size]
Edit:
Haven't read the article yet, but I just saw on Era someone said it was about motion tracking not DLSS which would be a weird thing to mix up frankly.
Edit2:
Not DLSS at all.
This is later today.
Now we wait and find out it has nothing to do with a DLSS feature, and it's something totally different.
Edit:
Well sounds like nothing
Twitch streamer ZombaeKillz joins the Tom's Hardware gang live as they talk PC builds for streaming, PS5's DLSS-like future and more
Probably speculation.
i'll bet it has to do with learning what games you enjoy playing so they can better target ads to you..
Era thread: DLSS is better than you even think. (https://www.resetera.com/threads/dlss-is-better-than-you-even-think.280898/)
First image is internally rendered at 720p, and upressed to 4k.
Native 4K:
So there's a few things worse in the 720p image.
Text on the Emergency Light is notably more smudged. There is an aliasing effect on the crates/boxes on the sides of the screen.
The FBC on the back of the guys shirt on the left is a little smudged.
I'm guessing the person was moving during the native image. I don't know what to call it, but there's kind of an outline on the right side of the native image's main character. The hair in the 720p image actually looks more filled in.
Just a little look into the quality. It's crazy good.
It would be interesting if we started seeing something like this on consoles. Where developers start scaling back the internal resolution, utilize less of the GPU for "regular" rendering, to utilize the rest for a DLSS type of technique.
I mentioned before that the Xbox Series X hot chips mentioned machine learning for resolution scaling. So it'll be interesting to see if Sony is using something similar, and it'll be interesting to see if/how much this actually gets implemented.
That smudge along Jessie sure is something. Maybe it's a TAA artifact which would be another benefit of DLSS.
In general native 4K looks a lot better to me. Pretty much everything looks smudged and/or aliased to me. Really wish both screenshots were pixel perfect identical while the game was paused.
This tech will only continue to get better though. Wonder if we'll ever get really weird things like rendering alternating frames at different resolutions, or rendering the color buffer at a lower resolution than the motion buffer, or rendering a third buffer that encodes more information about the scene in a novel way specifically for DLSS.
On very close inspection you can tell the native 4K is better but the difference is so small its crazy. As DF keep saying, it seems native 4K is a waste now. If PS5 really is focusing on a "fake 4K" approach like this then that's a big win.
i agree with legend,.. the 4k is quite noticably different once i full screened it. don't get me wrong DLSS did a dang impressive job. I think DLSS did really, really well with the objects in the foreground. but everything further back is in various states of pretty bad. The stairs on the left look like an "impressionist" painting honestly. it's soo undefined.
i still think DLSS has a future but an upscale from 1080p seems far more reasonable. that resolution is already pretty standard these days and shouldn't be hard to hit with any modern hardware.
when the GPU isn't rendering pixals what else can it do? like if developers seriously upscaled 720 renders to 4k with DLSS where would all that power be put to use? or is this just to get 120 fps...
i agree with legend,.. the 4k is quite noticably different once i full screened it. don't get me wrong DLSS did a dang impressive job. I think DLSS did really, really well with the objects in the foreground. but everything further back is in various states of pretty bad. The stairs on the left look like an "impressionist" painting honestly. it's soo undefined.
i still think DLSS has a future but an upscale from 1080p seems far more reasonable. that resolution is already pretty standard these days and shouldn't be hard to hit with any modern hardware.
when the GPU isn't rendering pixals what else can it do? like if developers seriously upscaled 720 renders to 4k with DLSS where would all that power be put to use? or is this just to get 120 fps...
I wasn't saying consoles would render at 720p.
This is just to demonstrate how much quality can be picked up despite a 9x pixel count.
I think 1080p/1440p -> 4K would be most likely. At that point it's a lot closer to native quality.
Ray tracing would be a big thing.
i agree with legend,.. the 4k is quite noticably different once i full screened it. don't get me wrong DLSS did a dang impressive job. I think DLSS did really, really well with the objects in the foreground. but everything further back is in various states of pretty bad. The stairs on the left look like an "impressionist" painting honestly. it's soo undefined.
i still think DLSS has a future but an upscale from 1080p seems far more reasonable. that resolution is already pretty standard these days and shouldn't be hard to hit with any modern hardware.
when the GPU isn't rendering pixals what else can it do? like if developers seriously upscaled 720 renders to 4k with DLSS where would all that power be put to use? or is this just to get 120 fps...
GPU power could be used to render more frames per second, increase the quality of rendered frames (more expensive shaders, raytracing...), or do more compute work (fluid simulation, ai...).
Would be nice if the OP did a better job making a comparison.
-pixel by pixel
-more resolutions so that we can see how 1080p/1440p improve the recreation.