Cyberpunk 2077 Ultra performance

[K1LaW8tt5]

Actually the the thread is everyone who can play with ray tracing On VS everyone who can't run ray tracing

Actually if you look past Nvidia Vs AMD what some people are saying is that the implementation of RT(along with some other things) in Cyberpunk is rubbish. This then summons the defence squad as for some reason the game is linked to nvidia.

The logic seems to be if you dislike or hate anything in cyberpunk you are an AMD fan/owner/lover......


anyway.
entry level physics apparently. Never seen this in real life thoug.

magic gate in the glass

where is the rest of his body?

 

[TheRealDeal]

Actually the the thread is everyone who can play with ray tracing On VS everyone who can't run ray tracing

No you missed out those of us who are waiting for the game to be the finished article before we play along with getting a next gen card. I really do think when it's ready for Ps5/Xbox series x it will be fit for purpose to play. Maybe then i can grab a decent card at the same time.

 

[nvidiamd]

Here is an example from Nvidia about how RT looks before denoising.



What we see here is that what we call RT atm is more based on denoising performance than on pure RT performance. Image reconstruction will always have more weight than the number of samples you can get. We are not even close to 5spp atm and i don't think the next gen cards will be able to do that. But let's say that in 10 years the tech will go so far that we will be able to do 50spp. Denoising will still have more weight in RT than pure RT power.
I am not saying that the image doesn't look good when it is denoised but RT is nothing like the real life. If it was, we wouldn't need denoising, our eyes/brains will be able to reconstruct the image without any help.
But as long as we still need artists or AI to "imagine" how most of the image would look like if it was fully ray traced or let's say how it will look like if they use 1000 times more rays, then why not hoping at less expensive techniques for increased "realism" ?

 

[Grim5]

Here is an example from Nvidia about how RT looks before denoising.



What we see here is that what we call RT atm is more based on denoising performance than on pure RT performance. Image reconstruction will always have more weight than the number of samples you can get. We are not even close to 5spp atm and i don't think the next gen cards will be able to do that. But let's say that in 10 years the tech will go so far that we will be able to do 50spp. Denoising will still have more weight in RT than pure RT power.
I am not saying that the image doesn't look good when it is denoised but RT is nothing like the real life. If it was, we wouldn't need denoising, our eyes/brains will be able to reconstruct the image without any help.
But as long as we still need artists or AI to "imagine" how most of the image would look like if it was fully ray traced or let's say how it will look like if they use 1000 times more rays, then why not hoping at less expensive techniques for increased "realism" ?

5spp looks like rayvtracing on consoles

 

[Rroff]

what some people are saying is that the implementation of RT(along with some other things) in Cyberpunk is rubbish.

No - many of them are blanket slinging mud at RT to deflect from AMD's limitations there - only slinking back to criticising the implementation in current games, which is woeful, when challenged.

The implementation in CP2077 is disappointing to say the least - they seem to be using a simplified version of the game world for reflections (hence the magic gate) with a lot of effects and objects missing - if you are carrying a body and move close to a reflective surface you won't see your own character and the body is floating in mid air with some parts missing heh.

Largely they have used RT in CP2077 to touch up the situations where traditional techniques fall apart rather than use RT for its strengths to enhance the visuals.

What we see here is that what we call RT atm is more based on denoising performance than on pure RT performance. Image reconstruction will always have more weight than the number of samples you can get. We are not even close to 5spp atm and i don't think the next gen cards will be able to do that. But let's say that in 10 years the tech will go so far that we will be able to do 50spp. Denoising will still have more weight in RT than pure RT power.
I am not saying that the image doesn't look good when it is denoised but RT is nothing like the real life. If it was, we wouldn't need denoising, our eyes/brains will be able to reconstruct the image without any help.
But as long as we still need artists or AI to "imagine" how most of the image would look like if it was fully ray traced or let's say how it will look like if they use 1000 times more rays, then why not hoping at less expensive techniques for increased "realism" ?

No offence but half of that is just regurgitating relevant sound words and either used incorrectly or just doesn't make sense.

Denoising is used to try and mimic having a higher ray count - it is unrelated to our eyes/brains reconstructing the image. With sufficient ray budget and a decent denoising algorithm we can attain visuals that exceed any traditional techniques even if we can't attain the level of fidelity an offline ray tracer could achieve as things stand.

The last couple of years have seen huge strides in denoising techniques and those working on it aren't done yet and as we move through 7 and 5nm processes and the hardware is also refined architecture wise we will be able to significantly increase the ray budget as well.

EDIT: The current denoising results are pretty incredible given what they are working with as well - this is the results in Quake 2 before denoising and resultant shading output:

https://imgsli.com/MzcxOTI

 

[gpuerrilla]

Denoising is used to try and mimic having a higher ray count - it is unrelated to our eyes/brains reconstructing the image. With sufficient ray budget and a decent denoising algorithm we can attain visuals that exceed any traditional techniques even if we can't attain the level of fidelity an offline ray tracer could achieve as things stand.

Thanks Rroff, I was enjoying my morning brew when my brain hit processing capacity in one mere sentence! But enthralling reading nonetheless..

 

[K1LaW8tt5]

No - many of them are blanket slinging mud at RT to deflect from AMD's limitations there - any only slink back to criticising the implementation in current games, which is woeful, when challenged.

There's a couple of muppets on either side but dont forget those that seem to think RT in cyberpunk is amazing and refuse to see it for what it is. Which is for the most part what you describe a badly implemented touch up that has just as many problems just different ones.

I'm not against the use of RT but I'm against people glorifying it in this particular game as for me it's not a good use of it.It is as bad of a hack job as normal raster used in other games to get similar end results with just as many errors and limitations.

Edit just to clarify the non-RT version of the screenshots I posted above are just a bad blur with no detail. while the RT option produces an implementation just as inaccurate as what other game engines have been doing for years.

 

[Wrinkly]

Actually if you look past Nvidia Vs AMD what some people are saying is that the implementation of RT(along with some other things) in Cyberpunk is rubbish. This then summons the defence squad as for some reason the game is linked to nvidia.

The logic seems to be if you dislike or hate anything in cyberpunk you are an AMD fan/owner/lover......


anyway.
entry level physics apparently. Never seen this in real life thoug.

magic gate in the glass

where is the rest of his body?

Just a bug, probably due to the gate being incorrectly flagged. If you had travelled up the lift you would have noticed that it works as expected at the top.

 

[K1LaW8tt5]

Just a bug, probably due to the gate being incorrectly flagged. If you had travelled up the lift you would have noticed that it works as expected at the top.

does takemura have abody for you? also I did travel to the top but how do you know thats just not a different pre baked scene at the top?

 

[Wrinkly]

does takemura have abody for you?

I think he did. He just didn't have the jacket in the reflection, which could be explained due to tha character being unique and high poly count. I'd guess there is a threshold that can be raised to display the correct image. I've not played with the info on gamersnexus.

 

[nvidiamd]

No offence but half of that is just regurgitating relevant sound words and either used incorrectly or just doesn't make sense.

Denoising is used to try and mimic having a higher ray count - it is unrelated to our eyes/brains reconstructing the image. With sufficient ray budget and a decent denoising algorithm we can attain visuals that exceed any traditional techniques even if we can't attain the level of fidelity an offline ray tracer could achieve as things stand.

The last couple of years have seen huge strides in denoising techniques and those working on it aren't done yet and as we move through 7 and 5nm processes and the hardware is also refined architecture wise we will be able to significantly increase the ray budget as well.

EDIT: The current denoising results are pretty incredible given what they are working with as well - this is the results in Quake 2 before denoising and resultant shading output:

https://imgsli.com/MzcxOTI

First i don't really care about AMD's performance in RT. I don't own a 6000 series card and i don't plan to get one. From what i am concerned, they can go to hell with their prices and their software support. What i was trying to argue is that joining a RT race when Nvidia controls pretty much everything that is ray traced in games, it will lead nowhere. Forget about AMD let's say you have a videocard company. How would you compete in RT? By multiplying the number of RT cores by 10? It still won't help you too much and as soon as you bring more performance than Nvidia on the table, they will stop sponsoring RT games. And no one will make them for free if most of the people play on 1060 right now.

Now you say that denoising is to try and mimic having a higher ray count. But i said the same thing. Ray Tracing is not the real deal right now because most of what we see in a game is based on denoising. Our eyes work different and with enough rays, you will get rid of denoising. But we will never have enough rays, this is what i said.

I have no doubts that denoising results are incredible and that they will become even better. But let me ask you this: is denoising ray tracing or is it rasterization? It will be amusing if 95% of what we see in a fully path traced game is based on rasterization techniques.
Can RT become more than a good support for rasterization techniques? Do you think we will be able to reach a good enough number of rays so that denoising will count less than 50% of what we see in an image in let's say 10 years?

People hate checkerboards but some of the denoising is done through checherboards. People hate blur but some of the denoising is done through blurring a lot of pixels. People hate fake lights or shadows but you use that a lot in denoising. People think rasterization sucks but like 95% of what they get in a "ray traced" game, is done through rasterization techniques. OMG but it looks so cool.

Is it wrong to tell people that only a small percent of the image is in fact ray traced, that the rays are only a good support for image reconstruction? Is it wrong to tell them that they will never get fully path traced games, or at least not in the next 100 years? Or that rasterization still has and will always have the biggest role in games?

 

[Rroff]

First i don't really care about AMD's performance in RT. I don't own a 6000 series card and i don't plan to get one. From what i am concerned, they can go to hell with their prices and their software support. What i was trying to argue is that joining a RT race when Nvidia controls pretty much everything that is ray traced in games, it will lead nowhere. Forget about AMD let's say you have a videocard company. How would you compete in RT? By multiplying the number of RT cores by 10? It still won't help you too much and as soon as you bring more performance than Nvidia on the table, they will stop sponsoring RT games. And no one will make them for free if most of the people play on 1060 right now.

Now you say that denoising is to try and mimic having a higher ray count. But i said the same thing. Ray Tracing is not the real deal right now because most of what we see in a game is based on denoising. Our eyes work different and with enough rays, you will get rid of denoising. But we will never have enough rays, this is what i said.

I have no doubts that denoising results are incredible and that they will become even better. But let me ask you this: is denoising ray tracing or is it rasterization? It will be amusing if 95% of what we see in a fully path traced game is based on rasterization techniques.
Can RT become more than a good support for rasterization techniques? Do you think we will be able to reach a good enough number of rays so that denoising will count less than 50% of what we see in an image in let's say 10 years?

People hate checkerboards but some of the denoising is done through checherboards. People hate blur but some of the denoising is done through blurring a lot of pixels. People hate fake lights or shadows but you use that a lot in denoising. People think rasterization sucks but like 95% of what they get in a "ray traced" game, is done through rasterization techniques. OMG but it looks so cool.

Is it wrong to tell people that only a small percent of the image is in fact ray traced, that the rays are only a good support for image reconstruction? Is it wrong to tell them that they will never get fully path traced games, or at least not in the next 100 years? Or that rasterization still has and will always have the biggest role in games?

Current (or even future path tracing) ray tracing in games doesn't work like offline renderers where every pixel is constructed on a per ray basis - in games you still process the geometry the same way as rasterisation and then shade the texels using the output from the denoiser. It isn't like AI reconstruction of the scene using a partial ray tracer output as you seem to be inferring - the scene is randomly sampled using path tracing to generate input for the denoiser which then tries to build a temporal stable output for shading.

Things like "95% based on rasterization techniques" is irrelevant to games which would use path tracing like the approach in Quake 2 RTX - only stuff like Lumen and CryTek's hybrid approach and hence why they have a lot of limitations and aren't an alternative solution like some try to hype them up to be.

Also things like "fake lights and shadows" aren't a thing in the way you are implying relevant to denoising - that is only relevant to dead-end hybrid approaches like CryTek's and other hybrid uses of RT which really includes Control and CP2077.

 

[ltron]

Actually if you look past Nvidia Vs AMD what some people are saying is that the implementation of RT(along with some other things) in Cyberpunk is rubbish. This then summons the defence squad as for some reason the game is linked to nvidia.

The logic seems to be if you dislike or hate anything in cyberpunk you are an AMD fan/owner/lover......


anyway.
entry level physics apparently. Never seen this in real life thoug.

magic gate in the glass

where is the rest of his body?

Yes there are bugs, I played on GeForce Now to see RT and some of Wakoko's hair and part of her head were white in the reflection. It was jarring.

 

[ICDP]

No - many of them are blanket slinging mud at RT to deflect from AMD's limitations there - only slinking back to criticising the implementation in current games, which is woeful, when challenged.

The implementation in CP2077 is disappointing to say the least - they seem to be using a simplified version of the game world for reflections (hence the magic gate) with a lot of effects and objects missing - if you are carrying a body and move close to a reflective surface you won't see your own character and the body is floating in mid air with some parts missing heh.

Largely they have used RT in CP2077 to touch up the situations where traditional techniques fall apart rather than use RT for its strengths to enhance the visuals.

This is one of my major issues with CP2077, the RT is used as a crutch and even then not a very good one. It looks OK but certainly not "life like" and the stock game looks average and cartoonish. If anyone wants to show-off what RT reflections can do then please use Watchdogs legions. The world and the RT reflections in WDL puts CP2077 to shame and even WDL has its limitations.

 

[Rroff]

This is one of my major issues with CP2077, all it takes is to see you are not getting reflections of yourself to ruin the "life like" look some are claiming CP2077 gives. If anyone wants to show-off what RT reflections can go then please use Watchdogs legions. The world and the lighting puts CP2077 to shame and even WDL has its limitations.

Even WDL doesn't really take advantage of what ray tracing can bring - people get too hung up on reflections in windows and puddles and basic shadows - but what really separates ray tracing is the things like scattered/indirect lighting, refraction and more complex use of reflections such as proper specular reflection, etc. stuff that traditional techniques just can't do and/or would have to be hand placed fake instances in every individual use of the effect, etc. some of it is quite subtle but once you run around a scene with it going back to games that don't use RT everything looks so flat and cardboard like.

Some of it doesn't convey well in screenshots either - just the way your brain takes in some of the more subtle accurate interactions of light as you move around makes a big difference.

 

[ICDP]

Even WDL doesn't really take advantage of what ray tracing can bring - people get too hung up on reflections in windows and puddles and basic shadows - but what really separates ray tracing is the things like scattered/indirect lighting, refraction and more complex use of reflections such as proper specular reflection, etc. stuff that traditional techniques just can't do and/or would have to be hand placed fake instances in every individual use of the effect, etc. some of it is quite subtle but once you run around a scene with it going back to games that don't use RT everything looks so flat and cardboard like.

Some of it doesn't convey well in screenshots either - just the way your brain takes in some of the more subtle accurate interactions of light as you move around makes a big difference.

I agree, that WDL is still not perfect as alluded too in my post. It's more that there are better examples of what RT can bring than CP2077. When people use terms like "life like" to describe the lighting and reflections in CP2077, I think they must be visually impaired.

What I also like about WDL is that turning off RT does not degrade the look of the game that much. The RT just adds an extra level of eye candy that give a good approximation (not perfect) of reflections in real life. Something that makes you think "this looks nice". Standing at a set of traffic lights and a bus stops in front of you, you see the reflections of you and everyone else near you as well as any other objects. It just makes you realise what RT should be capable of given time.

 

[Rroff]

What I also like about WDL is that turning off RT does not degrade the look of the game that much. The RT just adds an extra level of eye candy that give s good approximation (not perfect) of reflections in real life. Something that makes you think "this looks nice"

TBH proper use of RT should leave best effort use of rasterisation in the dust - one of the problems with ray tracing at the moment is having to support people who don't have capable hardware so you can't implement features that are fundamental to the rendering engine and/or remove some of the burden of rasterisation where there is feature overlap with ray tracing, etc. something people often don't appreciate with CP2077 for instance that there is a whole load of rasterisation workload as well as the ray tracing workload going on and that if you were to fully path trace the game you can claw back some performance from not having to use traditional rendering for a whole load of stuff.

 

[Wrinkly]

I agree, that WDL is still not perfect as alluded too in my post. It's more that there are better examples of what RT can bring than CP2077. When people use terms like "life like" to describe the lighting and reflections in CP2077, I think they must be visually impaired.

Assuming you acknowledge that RT can render life like images, at what point does the image fail to be life like?

To make it easier, at what stage does the following image become life like?

Here is an example from Nvidia about how RT looks before denoising.

 

[ICDP]

You are asking a loaded question because RT is more about how things look in real time. Games use ray tracing in a very limited form and as such rely on many different aspects to trick the player into an illusion of "real life". They aren't there yet and will be quite a while before RT replaces traditional methods completely. So it is difficult given current RT tech that any game is going to be totally life like in motion.

This is where IMHO CP2077 fails because the ray tracing is OK but not great. Other games have already done it much better.

 

[Wrinkly]

You are asking a loaded question because RT is more about how things look in real time. Games use ray tracing in a very limited form and as such rely on many different aspects to trick the player into an illusion of "real life". They aren't there yet and will be quite a while before RT replaces traditional methods completely. So it is difficult given current RT tech that any game is going to be totally life like in motion.

This is where IMHO CP2077 fails because the ray tracing is OK but not great. Other games have already done it much better.

Of course I'm asking a loaded question. RT is based on physics and models our own vision, hence life like.