Value of Hardware Unboxed benchmarking

Heck, he even talks about non-RT solutions having their own compromises at 3:30, so what really is your beef?
He doesn't like HUB and thinks they're biased towards Nvidia. That's the beef. Like the noise and artefacts are so very noticeable. Personally speaking, I'd rather have RT relegated to just lighting and shadows until GPUs get more powerful. Reflections are sufficient with either SSR with a good cubemap fallbacks or planar reflections. When most people play a game, they don't run around looking at the floor in an attempt to break SSR. It's not perfect at all but, for performance and image quality sakes, I much prefer it. It's just as of recent that SSR implementations have been so bad. Alan Wake is a prime example of bad SSR. It's like they spent their time enabling RT and just did the bare minimum for their SSR implementation.

If you go back and look at older games, you'll often find that their SSR implementations are more robust that recently released games.
 
Last edited:
Tim mentioned several times that the overall presentation is better with RT enabled. This wasn’t some crucifixion of RT
Context matters here. On it's own the video is fine, but it's a part of a long HU campaign to demonize RT using bad logic and arbitrary standards. Three videos already received heavy criticism for their flawed methodology. It's hard to pretend this last fourth video is done with benign intentions.

I play lots of COD BO6, it has zero ray tracing, yet it noisy as hell at native resolution.

In other games, screen space reflections is often very noisy, ghosty, aliased and unstable. Screen space shadows are the same. Screen space ambient occlusion is also noisy and has halo rings artifacts. Screen space global illumination is noisy too and unstable, software Lumen is noisy and has lots of boiling artifacts, non RT global illumination misses entire objects and areas, with a galore of light leaks and "glowing" objects.

Raster shadows are often jagged, low res, unstable and flicker a lot, especially when shadowing small geometry, most of the time they miss small details and geometry entirely.

He talks about laggy effect? how about a shoutout to all the raster reflections and shadows running at 1/3 or 1/4 the frame rates? These things are so common in gaming I can't believe he never found a single instant of them in his testing.

So his supposed claim of "enjoying games with artifact free presentation" is already bogus as he is intentionally hiding all most of the artifacts of the non RT path. Most of the screen space effects in current games are too reliant on TAA accumulation to bother presenting a supposed clean image.

In the end it comes down to raster being full of missing effects but with noise/flicker/laggy artifacts, or ray tracing with present and enhanced effects but with more or less noisy artifacts. It's that simple.

If he disabled Film Grain effects in many of these RT games, many of his noise problems would disappear. I tested that myself in Cyberpunk 2077 Path Tracing. In Hogwarts Legacy, he can easily use a mod that scales ray tracing resolution above Ultra (this is possible in many games already). In Star Wars Outlaws he is not using the max "Outlaw" settings, so his reflections are still not the highest quality possible from the game.

Notice how the video sometimes had to go to "300% zoom" just to make the problems noticeable. The noise that bothered his so much is often in screen elements that are small to be even noticed without a large zoom, certainly much smaller than entire effects disappearing out of screen space. In Metro Exodus he mentioned very rare examples of noise not even worth mentioning. It's little things like that that shows his argument and logic are being used in a bad faith.

If HU is fair, I expect them to release a video titled Graphics has a flicker/instability problem. But they won't even bother. They are milking the process of selling negativity around ray tracing to their already "primed" audience.
 
Last edited:
I disagree with the part that we "need" all of that.
They are needed if RT is to improve. They aren't needed to justify RT, nor needed until RT is worth it, but no-one was saying they were.
Having better RT h/w is welcome of course, as well as more AI being used for denoising and in more advanced ways. Do we "need" it though to use RT right now? No.
Which is exactly what the video says! :runaway:
 
It is a problem in just about every RT game I've played. Some more than others. Unlucky :(
Let me just check my log quickly...
DATW was mentioned already - zero issues, less noise with RT.
LIS Double Exposure - there are two to three scenes with GI boiling in the whole game. Disabling GI results in a significantly worse image. S/w Lumen reflections are better than SSR, no noise.
Returnal - don't remember any RT related noise there.
Still Wakes the Deep - also can't remember any noise but it's using s/w Lumen GI so there may be some. Disabling GI makes the image a lot worse.
RoboCop - this one has some noise on diffuse reflections but it's still better than not having them at all.
Brothers UE5 Remake - can't remember any noise issues in this one.
Dirt 5 - no noise.
Amid Evil - no noise that I can remember.
Hellblade 2 - can't remember any noise.
CP2077 - there is some ghosting with RR but that's about it.
Shadow of the TR - no noise.
The Talos Principle 2 - this one has some GI boiling in some indoor areas. Same as with other UE5 games disabling GI is a lot worse than that.
AW2 - can't remember any issues in this one but I've used RR.
Spider-Man 2 (on PS5) - no noise.
That's about a year back on my list. 14 titles, 3 titles with minor noise issues, 1 title with ghosting, so less than 33% "have a problem" here and out of these 3 which has noise are UE5 s/w Lumen titles.
Which is why I say that the whole premise of the video is completely misleading.

Is your entire judgement of the video based on the title?
No. It's based on what I see with my eyes in games this video claim something about.
 
Because one rendering method has shortcomings, the shortcomings of a different method shouldn't be considered?

From the intro, the video certainly isn't saying RT should be dropped in favour of non RT solutions, nor that alternative solutions are overall better. So I don't really understand your complaint. The main point is about RT noise, something unique to RT solutions.
The conclusion leads to pretty much the opposite though, to quote:
I think this is important to remember when discussing ray tracing as we soon enter a new generation of gpus: Ray tracing isn't always a universal image quality improvement[..]

See, the issues with and potential solutions about RT noise are absolutely a point worthy of discussion (I mean, it *is* the inherent bugbear of RT) - but since this video is apparently not just about RT noise in particular, but to be viewed from the perspective about *universal* image quality in relation to non-RT solutions, I can't help feeling the video being quite slanted, if not downright disingenuous in its comparisons.

I don't have any particular bone to pick with that channel or anything (TBH, I just stumbled on the video browsing around, not even sure if either here or on reddit), but I have to admit this video kinda rubbed me the wrong way in its framing.
 
If you mean the shadows, it's definitely right to remove the aliasing and oversharpening. It should look like it does in the RT shot. In fact I'd wager the RT shadows on that rock are not far off from a reference photograph taken under the same conditions.
Right as in physically correct, I understand that. But my vision is perception is disappointed by the loss of detail. And the difference is so big that it does not make sense to pick this for noise comparison.
 
Right as in physically correct, I understand that. But my vision is perception is disappointed by the loss of detail. And the difference is so big that it does not make sense to pick this for noise comparison.

But you're not losing detail. You're gaining detail. It's like saying 8-bit greyscale is more detailed than 32-bit because it has more contrast between shades.
 
Right as in physically correct, I understand that. But my vision is perception is disappointed by the loss of detail. And the difference is so big that it does not make sense to pick this for noise comparison.
They could turn off PCSS on RT shadows and make it more "detailed" than the shadowmaps could ever be. But it would look very strange, having pixel accurate sharp shadows from a far away tree canopy.

Leaving subjectivity out, the way they did it is correct, as in the light more accurately behaves like it would in the real world.
 
But you're not losing detail. You're gaining detail. It's like saying 8-bit greyscale is more detailed than 32-bit because it has more contrast between shades.
I am. The old renderer gives me a sense of many leaves casting shadows while RT shadow seems to come from hollow branches.

If 8-bit grayscale renders nicer than 32-bit, I would go with that.
 
I am. The old renderer gives me a sense of many leaves casting shadows while RT shadow seems to come from hollow branches.

If 8-bit grayscale renders nicer than 32-bit, I would go with that.
Sounds like you need to go outside during the day and look at some trees and get a better idea of what they're trying to recreate. Shadows cast by leaves several metres off the ground should be blurry
 
I am. The old renderer gives me a sense of many leaves casting shadows while RT shadow seems to come from hollow branches.

If 8-bit grayscale renders nicer than 32-bit, I would go with that.
This is exactly what I mean when I say that some people are so used to old artifacts that they consider them ground truth now.

Here's how forest canopy sun shadows look in real life:

 
If this few seconds long clip of one effect proves that rtrt does not have a noise problem, then would a similar clip that shows a rtrt effect that suffers from noise next to a noise-free rasterized effect be proof that it is a problem? I mean, we all know that this clip would be super easy to produce. Unless we're just playing some silly game again.

Btw, it's kind of mysterious that denoising for rtrt exists in every single game in a world where noise is absolutely not a problem with rtrt. Also, why are better denoising algorithms in active development in a world where noise is not an issue... kinda weird.
 
Last edited:
If this few seconds long clip of one effect proves that rtrt does not have a noise problem, then would a similar clip that shows a rtrt effect that suffers from noise next to a noise-free rasterized effect be proof that it is a problem?
No? Which is kinda my point.
RT doesn't have "a noise problem", that's just complete and utter BS on part of HUB.
Graphics in general has a noise problem, and RT is in fact making that problem less.
 
This isn't a great discussion.

Ah, yes, "the noise problem".
The possibility to produce perfect renderings with RT isn't a debate. What framerate, at what resolution, at what cost of hardware? One could look at someone with an annual income of £250k a year and then look at the cost of energy and conclude there is no cost-of-living crisis. What about people on lower incomes?

Firstly, I'm not sure people should be so up in arms about the word 'problem'. Problems can be big or small. I get some feel there's an agenda at HU - I've no idea if that's the case or not - and that is influencing perception. If we are to just discuss this, the video is really about a limitation in the RT algorithm and hardware we currently have driving it with a questionable choice of title that is very culturally influenced because everything is clickbait these days.

So, to confirm or refute the findings, we should identify what's being talked about. Rather simply, are RT games introducing noise (specifically white noise from sparse sampling, and not other types of signal degradation)? And then determine how much of an issue that is. We need a hardware target - I think the only sane one is the median average RT capable GPU. And we need some operation parameters. If your idea of a good rendering is 2160p120, and RT on on your 4060 creates a noisy game at 1080p45, that be a 'problem' in using RT.

Whatever, there's very little real discussion here.

Right as in physically correct, I understand that. But my vision is perception is disappointed by the loss of detail. And the difference is so big that it does not make sense to pick this for noise comparison.
The non-RT shadows in Indiana Jones are really noisy with a lot of pop and erratic changes. Plus outright wrong with that crazy LOD!

Btw, it's kind of mysterious that denoising for rtrt exists in every single game in a world where noise is absolutely not a problem with rtrt. Also, why are better denoising algorithms in active development in a world where noise is not an issue... kinda weird.
Effective denoising would eliminate the noise problem; so long as the noise isn't affecting the final output, it's not a 'problem'. No-one's saying RT doesn't introduce noise in the image generation. The argument is that that's resolved effectively for what the user sees, either being fine enough that no-one notices, or denoised to a point it doesn't matter.

I'm not entirely convinced by arguments that users get used to it so it's not a problem. A lot of the old rendering issues we tolerate in games, but we'd rather they weren't there and would notice (joyously) when they are banished. SSR looks ghastly at times, but also adds a lot to the game it's used in; we're used to it, it generally makes games better, but it still has a smearing problem. It's definitely possible that there could be noise in RT'd games that (some) people would want gone but can live with for now. It's also possible (inevitable!) that people's tolerances are very different, and the double whammy of denoising and upscaling to get decent framerates and resolutions without RT sparse sample noise might introduce a level of image fidelity degradation that some wouldn't prefer over cleaner, crisp, less-featured visuals. What one person is fine with, another might consider a problem.

A more objective look at this situation, with the video as a launching-point, would be to grab results from different games and hardware targets and identify the relationship between noise and image clarity (framerate, resolution, temporal artefacts).
 
Last edited:
Back
Top