What effects
GPU Ray Tracing Performance Comparisons [2021-2022]
- Thread starter DavidGraham
- Start date
D
Deleted member 13524
Guest
What question am I avoiding?
I asked what you would define as great use of RTX, to which you responded Control.
Control just isn't a great looking game. Making use of RT didn't make it a great looking game, so I don't know how you can argue it's making "great use of RTX". Showing accurate reflections of an average 8th-gen looking game doesn't make it a great use of RTX. This is something even Remedy themselves agree with, considering their latest statements on what to expect for their next 9th-gen focused engine.
I see "great use of RTX" as something with ray tracing that looks great because of raytracing. For example:
The value of accurate reflections is subjective and depends on what you're reflecting. Sure, you cannot do reflections of objects outside the FoV without raytracing (I guess you probably can with portals?, though that might be too costly to put into production), but I think you're overvaluing that feature. It's obviously nice to have accurate reflections but if you need to have something of value to reflect.
No one said anything about gameplay. The relevance of Demon's Souls 2020 is the fact that it looks like this:
Demon's Souls doesn't use raytracing (Bluepoint did state they were planning on using it but I don't know to what extent) and it runs @60FPS in 9th-gen console hardware (i.e. similar to medium-high PC specs) with a decent 4K output.
It's not just raytracing generally killing performance across the board that is making a number of developers hesitant to adopt it on scale and/or users hesitant to turn it on, it's also the fact that rasterization tricks and e.g. Voxel GI are still rising the bar significantly on hardware that many people own (and cannot upgrade at the moment due to shortage and inflated prices).
DavidGraham
Veteran
You can watch the original videos here.
Maybe for Tomb Raider, it uses RT shadows which is difficult to spot the difference for unless side by side, but Minecraft? It's using full path tracing vs the base game which doesn't even support basic lighting or shadowing, how can anybody not spot the massive difference? I mean you have to intentionally sabotage the test scene for people not to notice such a gigantic graphical upgrade!
DegustatoR
Veteran
This game in particular looks almost the same without RT because all its graphical elements usually fit into one screen so all screen space options work rather well here.
yea unfortunately hard to say without seeing how the test and what scenes were conducted. I just don't doubt that casuals can be really casual. By the numbers, there's going to be a large percentage of people that would not be able to figure it out. What may be very basic to you to see and spot will be extremely challenging to others.
It's just a fact that some people are better trained at spotting differences and some people are just terrible at it. Especially when compared to people on this forum. B3D is a really niche population compared to gaming enthusiast forums, and those forums (like Era and Gaf) are extremely niche among the total population of people that play games. We represent like a fraction of a %0.00001 here. There's barely 100 active members here.
Last edited:
I shared screenshots earlier in the thread.
So what you're saying is that in your opinion the transparent environment reflections in Control don't look great? I think you're objectively wrong.
Each feature adds to the final result and in this case those reflections add a lot. Your stance seems to be that if a game uses raytracing and everything else about the game (textures, geometry, art style) doesn't float your boat then raytracing didn't work enough magic so it sucks.
All you're saying is that you like the art work and design of one game that happens to not use RT more than another game that does. But that literally has nothing with the discussion of what RT adds to games.
Lol hopefully there are enough lurkers and readers to make it worthwhile. Otherwise it's just a few monkeys slinging poop at each other over and over.
Accurate reflections from 1998 look at the floor (2:40 if the play from current time doesnt work)
D
Deleted member 13524
Guest
I actually found other blind tests on youtube that show similar results, but I'm trying to be fair and avoid videos from e.g. 2019 because I know the more recent implementations are better.
There's no "objectively wrong" in a subjective opinion. I would be objectively wrong if I said the raytraced reflections aren't more accurate, but what I said was the raytraced reflections don't make Control look that much better than its original 8th-gen console iteration.
No, I never used the term "it sucks". That's just a product of the hyperbole you're applying from an IMO needlessly defensive stance. I also finished Control along with all its DLC, by the way.
Lego definitely doesn't float my boat as a game personally and I just set it as an example of what I think it's great usage of ray tracing.
You're awfully focused on invalidating my subjective opinion, so let's put that out of the way.
The best-looking rasterization-only games are considered to look better than any current game that uses RT. Here's a poll from resetera with ~900 participants:
https://strawpoll.com/yp3bodz22/r
Top 3 goes to:
1 - The Last of Us Part 2
2 - Red Dead Redemption 2
3 - Demon's Souls 2020
Cyberpunk is the first RT title on the list being a close 4th (still behind Demon's Souls), and Control is number 20 or something.
Control didn't turn heads as an especially good looking game, and its residual 4 votes put it below Luigi's Mansion and Devil May Cry 5. It's just not a game that is popular for its graphics.
DavidGraham
Veteran
A new test app for Ray Tracing has been released from a group called Tellusium, they compare the performance of the 6700XT vs the 2080Ti.
@900p:
2080Ti is twice as fast as the 6700XT!
@2160p:
2080Ti is almost three times faster than 6700XT!
It's worth mentioning that vs compute shaders the 2080Ti becomes 7 times faster using hardware ray tracing, while the 6700XT runs only 3 times faster.
In compute, without hardware acceleration the 6700XT is 15% faster than 2080Ti.
The memory footprint for RT is different between AMD and NVIDIA, as AMD requires twice as much memory compared to NVIDIA.
https://tellusim.com/rt-perf/
@900p:
2080Ti is twice as fast as the 6700XT!
@2160p:
2080Ti is almost three times faster than 6700XT!
It's worth mentioning that vs compute shaders the 2080Ti becomes 7 times faster using hardware ray tracing, while the 6700XT runs only 3 times faster.
In compute, without hardware acceleration the 6700XT is 15% faster than 2080Ti.
The memory footprint for RT is different between AMD and NVIDIA, as AMD requires twice as much memory compared to NVIDIA.
https://tellusim.com/rt-perf/
Is that specific to the benchmark or is it also the case in RT games ?
So, a c
It shows that HT Raytracing on RDNA2 was never intended by AMD. 2080TI and 6700XT has the same numbers of transistors and yet 2080TI is up to 3x faster. And they explain that "software" bvh travel is inefficient and a huge bottleneck. So using compute shaders for it needs a huge amount of on chip cache to be even in the near full HW Raytracing.
It shows that HT Raytracing on RDNA2 was never intended by AMD. 2080TI and 6700XT has the same numbers of transistors and yet 2080TI is up to 3x faster. And they explain that "software" bvh travel is inefficient and a huge bottleneck. So using compute shaders for it needs a huge amount of on chip cache to be even in the near full HW Raytracing.
A pint rather. This kind of poll is merely a vote for whether or not people know the game and if so, how much those people do like the game.
unfortunately yea. You'd have to baseline the survey by ensuring they've played all these titles at their best possible settings. Not likely. I would agree that this survey is likely more representative of what they've played.
If you have a controlled survey environment, you could give them screenshots to decide by, or videos to watch or even let them play a number of games they are supposed to rate on accordingly equipped systems.
The post that starts with "Is this meant to be a joke?" and shows accurate reflections of the environment and player in control ?
If thats the case I posted a video showing accurate reflections of the environment and player in Unreal (granted its a much simpler environment)
I'm not a RT hater by any means I bought a RTX card because I wanted RT, but ive not seen anything effect wise in RT that hasnt been faked by non RT
Not at all. Everyone is entitled to their own taste in what overall look they like in a game. However, I would be really interested to hear from folks who would prefer missing or inconsistent lighting in their games regardless of the art style.
Yep, I'm struggling to understand the relevance to this topic.
I'm pretty sure your Unreal video didn't have transparent reflections showing reflected dynamic geometry while also showing the objects in the next room behind the glass. I don't think your definition of accurate reflections is accurate
Now if someone says pre-rendered opaque cubemaps is all they need then more power to them as their needs are already met.
I recently finished MoonMa... Deliver us the Moon and liked the game itself a lot. Here's one shot, where I think RT really showed its merits (which i could enjoy thx to a borrowed RT-capable card from a friend, which I hoped he'd sell me for cheap, but then he got wind of the ebay prices... alas.).
It's not always those in-your-face kind of effects, but the more "real" feeling, when small things like the reflection in the glass which becomes much more discernible inside the shadow cast on the glass. But YMMV.
https://imgsli.com/NzY5NDM
It's not always those in-your-face kind of effects, but the more "real" feeling, when small things like the reflection in the glass which becomes much more discernible inside the shadow cast on the glass. But YMMV.
https://imgsli.com/NzY5NDM
