• 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.

The RT Related Games, Benchmarks, Software, Etc Thread.

Suspended
Joined
30 Mar 2010
Posts
13,068
Location
Under The Stairs!
Double the lol though^^

OK

Daniel_Owen_native_res_does_produce_the_best_output.png
 
Soldato
Joined
18 Feb 2015
Posts
6,485
How bad then? No chance I'm watching that.
Not that bad. The results on PC get skewed a lot by optimised-for-Nvidia-only (and usually by Nvidia devs that major games get for free to help add RTX) that a lot of RT implementations suffer from. Though, of course, that's poor comfort for would-be AMD GPU buyers. It's up to AMD to hustle and throw some devs at studios to make the games better suited to RDNA.
For example, the 4080 is somewhere close to 5x the tflops of an XSX and yet that ratio only holds for one step of the RT pass, with others being close to 2-3x. Plus there's also clear diminishing returns past certain absolute times so inevitably there's only so much speed-up to be done. If you then consider die-area and the costs savings from this hybrid approach, it's easy to see why console makers & AMD went this route. No doubt they can still benefit from adding specialised hw for boosting certain aspects of the RT pipeline (particularly thinking of BVH building) and I'm sure they will, but it's not as doom and gloom as it may seem from certain results on PC. It's not as simple as "just add RT cores bro" that a lot of misguided commentators repeat, because we can see from Intel's GPUs how horrendous those are from a die-area perspective and the results are not commensurate with it. On paper they were even above Ampere in terms of "RT core tech" and yet the performance wasn't there (even accounting for overall GPU performance; talking about ratio wise).

4ER8wiO.png

 
Last edited:
Soldato
Joined
6 Feb 2019
Posts
17,635

:cool:

Who needs texture packs when you can have path tracing to enhance the textures :D



The interesting bit from that video is just how light path tracing actually is, and how heavy denoising actually is.

Enabling path tracing by itself only drops the performance by 10%. But the image has film grain on it, so it needs to be cleaned with the denoising passes and once the denoiser is enable the performance drops a further 60%!
 
Soldato
Joined
14 Aug 2009
Posts
2,822
The interesting bit from that video is just how light path tracing actually is, and how heavy denoising actually is.

Enabling path tracing by itself only drops the performance by 10%. But the image has film grain on it, so it needs to be cleaned with the denoising passes and once the denoiser is enable the performance drops a further 60%!
There isn't a denoiser. He just goes to 4 rays/pixel which decreases noise (from just 1). Denoisers aren't more performance hogs than suficient cast rays or else they wouldn't be used.

RR in Cyberpunk can be significantly faster since is bypassing / replacing some individual denossiers (in some cases), but if you disable that completely and you'd rely on the number of rays cast alone to eliminate noise... can't be done real time yet.
 
Man of Honour
Joined
18 Oct 2002
Posts
100,451
Location
South Coast
RR would still be superior to using cast rays to denoise, effectively RR sets out to replace multiple rays with one AI pass, hence the name Ray Reconstruction, this boosts performance, and boosts image quality to result in a clean image with more detail than what traditional cast rays offers and the denoising it comes with, the only way to get clean non-RR path tracing is to fire off more rays and get more bounces, and that will means a much bigger hit to performance like we saw in Cyberpunk before last patches which meant you cannot modify the rays/bounce count any more since it's being handled by RR.

RR is the future of ray/path tracing.
 
Soldato
Joined
14 Aug 2009
Posts
2,822
RR would still be superior to using cast rays to denoise, effectively RR sets out to replace multiple rays with one AI pass, hence the name Ray Reconstruction, this boosts performance, and boosts image quality to result in a clean image with more detail than what traditional cast rays offers and the denoising it comes with, the only way to get clean non-RR path tracing is to fire off more rays and get more bounces, and that will means a much bigger hit to performance like we saw in Cyberpunk before last patches which meant you cannot modify the rays/bounce count any more since it's being handled by RR.

RR is the future of ray/path tracing.
RR can also give an oily paint look at times and some weird artefacts on characters. The futures still is enough rays shot per pixel. The present is RR :p
 
Man of Honour
Joined
18 Oct 2002
Posts
100,451
Location
South Coast
RR can also give an oily paint look at times and some weird artefacts on characters. The futures still is enough rays shot per pixel. The present is RR :p
That was RR v1, Since ReSTIR GI got implemented (Cyberpunk), this no longer happens and now we get even better path tracing and shadows as a result. All of my screenshots since patch 2.12show no oily skin or any early RR compromises.
 
Man of Honour
Joined
18 Oct 2002
Posts
100,451
Location
South Coast
Makes no difference visually to Cyberpunk, it loads the quality of textures you set in GFX (or mod tweaks) the same in every res, I posted side by side screenshots showing zero visual difference remember, so opted for the one that has the higher framerate.
 

TNA

TNA

Caporegime
Joined
13 Mar 2008
Posts
27,716
Location
Greater London
Makes no difference visually to Cyberpunk, it loads the quality of textures you set in GFX (or mod tweaks) the same in every res, I posted side by side screenshots showing zero visual difference remember, so opted for the one that has the higher framerate.

To you.. This stuff is subjective however and other factors like how far you sit away from the screen and quality of your eyes matter.

So you really can't say that with 100% certainty that it makes no difference.

I almost always see the difference higher resolution makes.

Oh and yes I know you will take the opportunity to post that picture of the guy with his face touching the monitor.
 
Man of Honour
Joined
18 Oct 2002
Posts
100,451
Location
South Coast
It's not subjective when there are literal videos and screenshots showing no difference. If you are seeing a difference then it's placebo or your settings between both resolutions are not the same without any evidence. That's why I always do detailed comparisons showing if or if there isn't a difference.
 
Last edited:

TNA

TNA

Caporegime
Joined
13 Mar 2008
Posts
27,716
Location
Greater London
It's not subjective when there are literal videos and screenshots showing no difference.

You just disregarded what I said. Distance and quality of eyes does matter.

If say I was to play from 4 foot away then yeah, no point in 4K as I would not see the difference. But I don't and I do see a big difference.

I even saw a difference in HFW but the difference was not big enough to play the game at the much lower fps. I needed a higher fps to be able to use frame gen properly. Still could easily see the difference though.
 
Last edited:
Man of Honour
Joined
18 Oct 2002
Posts
100,451
Location
South Coast
But I'm on about sitting a normal distance as per 99% of desktop setups. F|or this purpose it makes no difference and the screenshots and videos posted can be viewed at 100% zoom too as opposed to scaling a native screenshot/video recorded at say 3440x1440 and being viewed on a 32" 4K display which will obviously display back upscaled by the display, whereas if the native 4K version of that screenshot was displayed 100% zoom on the same 32" then that would look crisper, do you see what I'm saying here from a technical point of view?

When just viewing at 100% zoom (not scaled) any video or screenshot the difference isn't there, and likewise my comparisons are exactly that, the same size display viewing both UWA 5160x2160 and UWA 3440x140 - No difference, again, I have posted the comparisons which accounts for the display scaling factor to mitigate the scaling differences between resolutions and display size and that accurately shows no difference.

The only game in recent years that has a difference is Hogwarts and that's due to a flaw in the game engine which only loads very high textures at 4K and nothing below, even if you choose very high textures in GFX settings, this is a game bug not a pro point for 4K being higher quality in that game.
 
Last edited:
Back
Top Bottom