Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
From the CP2077 thread, may as well copy paste I guess:
I did my most comprehensive test with Raster vs RT vs PT comparing DLSS Balanced vs DLAA. All but DLAA also compared Ray reconstruction on vs off, and keep in mind that RR is only possible if using DLSS, not DLAA or any other AA implementation, as it uses the upscaling part of the tensor cores to do what it does. It's pretty clear that Ray Reconstruction is the technology to have for improved image clarity as well as lighting and shadow definition whether you are ray traced or path traced and to some degree, even raster as the toggle was available in raster too strangely.
You guys decide which combo of tech looks the best.
Note that:
- Ray Reconstruction offers better details/shadows and reflections than all examples with RR off
- DLSS Balanced offers better quality texture detail and sharpness than native (DLAA). Both DLSS and DLSS were set to the same 0.45 sharpness
- The fps cost using DLAA path traced at 3440x1440 vs DLSS Balanced is exactly 50% - Remember no RR possible with DLAA too.
So what this shows (the same applies to Alan Wake too and all other games that use or will use Ray Reconstruction) is that if you want the best visual quality as well as performance combo, then you need to be using Ray Reconstruction. The exact same findings apply at 5160x2160 using DLSS Performance, just the framerate numbers scale relative to the increase in rendered pixel workload.
At 1080p there isn't enough internal pixel data to upscale all the details that 1440p or greater has, so you lose definition in areas of subtle shadow/lighting. 1440p or above is the point to aim for.
Ray Reconstruction also has a 10fps uplift in this game, other games may vary.
All dll versions for each of the 3 techs used are v 3.7.10 (3.7.0 for RR) using Preset E.
There's something not quite right with all of them.
I don't think this comparison is a good showcase for the tech, as it just displays the compromises - better to keep the punter in the dark (so to speak).
The choice is between bright and gloss to every surface, or more natural and accurate grubbiness of the scene coming through but loss of detail (presumably due to lack of bounces), or RR selectively removing or exaggerating elements of the scene.
They each have their distinctive qualities and do add some pleasing changes in different ways but seemingly at the loss of something else.
There's something not quite right with all of them.
I don't think this comparison is a good showcase for the tech, as it just displays the compromises - better to keep the punter in the dark (so to speak).
The choice is between bright and gloss to every surface, or more natural and accurate grubbiness of the scene coming through but loss of detail (presumably due to lack of bounces), or RR selectively removing or exaggerating elements of the scene.
They each have their distinctive qualities and do add some pleasing changes in different ways but seemingly at the loss of something else.
All methods of technology will have some compromise somewhere but the important thing is when it can provide more good than bad and for most of the gameplay time which is what matters and this is where nvidia tech such as dlss, rr etc. generally excel as being better 90+% of the time compared to the "traditional" methods of denoising, TAA/AA use and so on.
Having played some 500 hours in at least this game and regularly experimenting with settings, updates etc, and al in between playing other games that have similar tech, I can definitely say that RR+PT is the definitive way to experience any game. RR restores otherwise lost detail, I say lost because part of screen space effects where RT isn't used, or denoising when RT is used, results in detail loss and image garbling. RR fixes that by recognising where detail is lost or garbled, and restoring it or cleaning it. So the RR method is the only true and accurate method of rendering where it is available as an option.
Screen soace is the inferior method by far, RR is not possible with non RT pathways so screen space remains left in the dust.
Having played some 500 hours in at least this game and regularly experimenting with settings, updates etc, and al in between playing other games that have similar tech, I can definitely say that RR+PT is the definitive way to experience any game. RR restores otherwise lost detail, I say lost because part of screen space effects where RT isn't used, or denoising when RT is used, results in detail loss and image garbling. RR fixes that by recognising where detail is lost or garbled, and restoring it or cleaning it. So the RR method is the only true and accurate method of rendering where it is available as an option.
Screen soace is the inferior method by far, RR is not possible with non RT pathways so screen space remains left in the dust.
Yup, RR at launch did have some noticeable drawbacks like the smear/blurring on npcs but I believe that has been resolved? The biggest benefit with RR though is when you're using dlss performance mode, the reflections are noticeably cleaner looking.
I'm stuck on a 3080 so i get some ray tracing here and there. If I focus YES, there is a noticeable difference but not enough for me to care. Happy to leave it on the table for extra performance.
Having played some 500 hours in at least this game and regularly experimenting with settings, updates etc, and al in between playing other games that have similar tech, I can definitely say that RR+PT is the definitive way to experience any game. RR restores otherwise lost detail, I say lost because part of screen space effects where RT isn't used, or denoising when RT is used, results in detail loss and image garbling. RR fixes that by recognising where detail is lost or garbled, and restoring it or cleaning it. So the RR method is the only true and accurate method of rendering where it is available as an option.
Screen soace is the inferior method by far, RR is not possible with non RT pathways so screen space remains left in the dust.
Restoring lost detail feels a strong phrase for it. For the extra delight it brings to the eye it is definitely creating some oddities, subtle as they are, in that Cyberpunk screenshot comparison.
Having a look at PT+RR Off/On, some examples:
RR appears to create extra speckly grime to the yellow background of the 'Out of Order' sign, likewise the 'Insta Food' sign on the vending machine is messy with exaggerated bright spotty reflections.
The paper scraps on the table (bottom right) become high coasters with strong shadow.
The metal bracket on the side of the bench is overly brought out into detail.
The glass counter top cabinet in the centre, when looking through the glass panels, the rear of the bar as seen through the glass is really not quite right.
Can't say I like what it does to the upward curving bar recess below that cabinet, pursuit of depth and shadow just creates a blurry mess, even as it does bring back missing shadows.
The light from the ceiling television screens is being reflected far too strongly on the ceiling above - which could only have come from the floor as the screen angle doesn't allow direct lighting - admittedly an issue in all except PT, but add RR and it brings it back.
It seems to have funny visual aberrations here and there, that subjectively, ought not to be there.
I should say I'm not crapping on it as a technique and saying its no good, just pointing out a few places where it doesn't seem to be doing as good a job as elsewhere in that image.
All methods of technology will have some compromise somewhere but the important thing is when it can provide more good than bad and for most of the gameplay time which is what matters and this is where nvidia tech such as dlss, rr etc. generally excel as being better 90+% of the time compared to the "traditional" methods of denoising, TAA/AA use and so on.
My standards/expectations are perhaps overly high. But RT as implemented now is right now so far short of its potential that I'm not yet impressed. It has still got some way to go, and to me it seems fundamentally a hardware/computational limitation and the schemes can only do so much to compensate.
Thats why I am pretty set on giving it a few years still. It is what it is, it is progress, it is better than nothing I suppose, don't doubt it adds, but for me it's not enough to invest/care about yet.
What you're decrsibed is accurate to how it should look for the most part though, looking at the PT RR on vs off examples (https://imgsli.com/Mjc1ODA0/7/8) we can see that:
RR appears to create extra speckly grime to the yellow background of the 'Out of Order' sign, likewise the 'Insta Food' sign on the vending machine is messy with exaggerated bright spotty reflections.
This difference is so small it could even be chalked up to variance as the lighting in PT is always changing, any NPC or fan blocking an emissive source creaes a difference in indirect illumination on another surface, this is the comparison example you note:
The paper scraps on the table (bottom right) become high coasters with strong shadow.
This is how it should be as there is a spotlight above the table. With RR off, rhe engine is ignoring that there is an emissive light source above the surface of the textures and ignoring the light source distance from the texture. RR is correctly applying these parameters and then placing the shadows where they otherwise should be.
The comparison, look at the lower coaster/saucer thing, its shape dictates that there should be a shadow under it, RR off shows no shadow:
The metal bracket on the side of the bench is overly brought out into detail.
Again this is what is expected given the emissive light source directly in that region of scene, the spotlight from the ceiling top right, the light cast from that will be hitting the metal and highlighting detail which the non RR path is ignoring:
The glass counter top cabinet in the centre, when looking through the glass panels, the rear of the bar as seen through the glass is really not quite right.
I don't see this one, what doesn't look right? it looks to be lit in accordance with the emissive light sources that would naturally be hitting that area of the scene. if anything, the RR version is resolving the glass bottles behind the glass door panes properly whereas the non RR version is muddying the details that should be visible as the glass is clear not frosted.
Can't say I like what it does to the upward curving bar recess below that cabinet, pursuit of depth and shadow just creates a blurry mess, even as it does bring back missing shadows.
The RR version is correctly shaded again, the non RR version is actually blocky and does not represent how that recess should look given the scene's neighbouring lighting:
The light from the ceiling television screens is being reflected far too strongly on the ceiling above - which could only have come from the floor as the screen angle doesn't allow direct lighting - admittedly an issue in all except PT, but add RR and it brings it back.
This is normal and how light would react and reflect on a shiny surface in real life, I know this as a wedding photographer who deals with reflections off surfaces, The non RR version has the same reflection, just the ad on the screen has changed so the colour/brightness of the reflection is more muted accordingly to match the screen contents. An argument could be made that when RR is on for the Ray Traced version, then it is brighter, but then again so is everything else surrounding the whole frame as this is the trait of RT vs PT. PT is the most accurate once again in conjunction with RR.
What you're decrsibed is accurate to how it should look for the most part though, looking at the PT RR on vs off examples (https://imgsli.com/Mjc1ODA0/7/8) we can see that:
This difference is so small it could even be chalked up to variance as the lighting in PT is always changing, any NPC or fan blocking an emissive source creaes a difference in indirect illumination on another surface, this is the comparison example you note:
This is how it should be as there is a spotlight above the table. With RR off, rhe engine is ignoring that there is an emissive light source above the surface of the textures and ignoring the light source distance from the texture. RR is correctly applying these parameters and then placing the shadows where they otherwise should be.
The comparison, look at the lower coaster/saucer thing, its shape dictates that there should be a shadow under it, RR off shows no shadow:
Again this is what is expected given the emissive light source directly in that region of scene, the spotlight from the ceiling top right, the light cast from that will be hitting the metal and highlighting detail which the non RR path is ignoring:
I don't see this one, what doesn't look right? it looks to be lit in accordance with the emissive light sources that would naturally be hitting that area of the scene. if anything, the RR version is resolving the glass bottles behind the glass door panes properly whereas the non RR version is muddying the details that should be visible as the glass is clear not frosted.
The RR version is correctly shaded again, the non RR version is actually blocky and does not represent how that recess should look given the scene's neighbouring lighting:
This is normal and how light would react and reflect on a shiny surface in real life, I know this as a wedding photographer who deals with reflections off surfaces, The non RR version has the same reflection, just the ad on the screen has changed so the colour/brightness of the reflection is more muted accordingly to match the screen contents. An argument could be made that when RR is on for the Ray Traced version, then it is brighter, but then again so is everything else surrounding the whole frame as this is the trait of RT vs PT. PT is the most accurate once again in conjunction with RR.
What you're decrsibed is accurate to how it should look for the most part though, looking at the PT RR on vs off examples (https://imgsli.com/Mjc1ODA0/7/8) we can see that:
This difference is so small it could even be chalked up to variance as the lighting in PT is always changing, any NPC or fan blocking an emissive source creaes a difference in indirect illumination on another surface, this is the comparison example you note:
You do need to use the slider, above doesn't do it justice, to see just how overt the difference.
It seems clear in PT+RR on that in a number of cases some noise has been selected for rather than correctly eliminated.
In the case of the sign its not only the yellow background - check the 1st U in 'OUT' and 2nd O in 'OF'. Unless those particular small spots of the sign are exceptionally clean (and the artist creating the scene intended to convey that) it has introduced bright spots about the text. Looking at the smaller font text at the bottom, the shapes of the text is almost lost due to it.
We can also look at the other render options to gain some insight into the object.
The RR version is correctly shaded again, the non RR version is actually blocky and does not represent how that recess should look given the scene's neighbouring lighting:
Not the bottle wall cabinet, the countertop cabinet sitting on the bar (on the right side in the above image), viewing through the glass case, you can see here RR makes a hash of it, in a way that none of the 8 other options do, although some get close.
The curved recess I mentioned is underneath the countertop cabinet - subjective on my part (but isn't it all!) its just overly exaggerated the shadow.
Since we are focusing on that wall bottle cabinet, see the horizontal lines on the lowest shelf panel. Again PT+RR is amongst the worst visually of the various options.
God know what it was trying to do to the shadow below the wall cabinet, and whatever the wall marking/line is meant to be, but it produced a mess.
Again this is what is expected given the emissive light source directly in that region of scene, the spotlight from the ceiling top right, the light cast from that will be hitting the metal and highlighting detail which the non RR path is ignoring:
Bracket: That spotlight is above the far side of the table over the bench corner, there should be no light directly hitting the metal bracket from it. At best the holo TV is casting some weak direct light.
You can tell the form of metal bracket it is intended to be from the other render options. PT+RR outputs amongst the worst, virtually loses its shape and structure.
RR makes it more prominent and since you know what it is 'meant' to be, your brain fills in for you at a glance, but I don't think it can be argued its doing a good job, even if the enhancement (which has added noise) appears more rich superficially to the eye in passing. As you approach the accuracy will increase and will resolve the item, so you could argue the jobs a good'un in the grand scheme, though that isn't compelling. RR should do better and hopefully can and will.
This is how it should be as there is a spotlight above the table. With RR off, rhe engine is ignoring that there is an emissive light source above the surface of the textures and ignoring the light source distance from the texture. RR is correctly applying these parameters and then placing the shadows where they otherwise should be.
The comparison, look at the lower coaster/saucer thing, its shape dictates that there should be a shadow under it, RR off shows no shadow:
Id be intrigued to know whether they are saucers or coasters, or just paper scraps. Don't know if anyone knows this scene?
Using the slider and viewing both objects its clear both have had their shadows simply exaggerated.
I suspect its confused a dark shallow vertical coaster edge, with a hint of shadow, and a grubby discoloured surface and selected/weighted wrong and over emphasised.
This is normal and how light would react and reflect on a shiny surface in real life, I know this as a wedding photographer who deals with reflections off surfaces, The non RR version has the same reflection, just the ad on the screen has changed so the colour/brightness of the reflection is more muted accordingly to match the screen contents. An argument could be made that when RR is on for the Ray Traced version, then it is brighter, but then again so is everything else surrounding the whole frame as this is the trait of RT vs PT. PT is the most accurate once again in conjunction with RR.
PT is the only source of accuracy, RR when cleaning the image, as well as reducing noise can clearly also select for and amplify it in corner cases - but the corner cases would be part the battleground for something like RR, so its perfectly valid to point out its own flavour of oddities, which can be improved.
I can't deny that it introduces richness and depth to the scene that would be otherwise lacking with the amount of PT/RT alone.
You can't whack spending that amount on a monitor and still being classed as a peasant. Time to contact the bank to see what they are willing to give you.
My G7 puts me in a drug den with a needle hanging out my arm.
You can't whack spending that amount on a monitor and still being classed as a peasant. Time to contact the bank to see what they are willing to give you.
My G7 puts me in a drug den with a needle hanging out my arm.
Mate, I'm too far from stupid to infer the 3080 can't run raster, that claims on you, bud.
Glad we've reached agreement by lack of anything RT related from your end at all on the topic of the 3080 now only being able to run some RT in 2024.*thumbsup
Mate, I'm too far from stupid to infer the 3080 can't run raster, that claims on you, bud.
Glad we've reached agreement by lack of anything RT related from your end at all on the topic of the 3080 now only being able to run some RT in 2024.*thumbsup
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.