ASUS TUF.
2560x1440 but I get the same rays per second at 4K.
Not sure, didn't check at the time.
GPU Ray Tracing Performance Comparisons [2021-2022]
- Thread starter DavidGraham
- Start date
ASUS TUF.
2560x1440 but I get the same rays per second at 4K.
Not sure, didn't check at the time.
D
Deleted member 13524
Guest
Ascent: another "RTX-ON" title with a serious case of "emperor is just wearing a thong" while killing performance.
I see a visible difference with reflections. The other two are perhaps "too well implemented" in rasterization and the difference seems practically negligible to me.
RT Effects:
https://www.pcgameshardware.de/comm...ew.cfm?article=1377602&page=1&draft=-1&rank=2
RT Shadows:
https://www.pcgameshardware.de/comm...ew.cfm?article=1377602&page=1&draft=-1&rank=4
RT Ambient Occlusion:
https://www.pcgameshardware.de/comm...ew.cfm?article=1377602&page=1&draft=-1&rank=3
At 1440p, performance with RT vs. rasterizer is 38% on a 3080 Ti and 28% on a 6800 XT, and then it gets worse on lower end models.
The game is also super dependent on CPU performance and RT + 720p doesn't help normalize it like higher resolutions usually do.
I see a visible difference with reflections. The other two are perhaps "too well implemented" in rasterization and the difference seems practically negligible to me.
RT Effects:
https://www.pcgameshardware.de/comm...ew.cfm?article=1377602&page=1&draft=-1&rank=2
RT Shadows:
https://www.pcgameshardware.de/comm...ew.cfm?article=1377602&page=1&draft=-1&rank=4
RT Ambient Occlusion:
https://www.pcgameshardware.de/comm...ew.cfm?article=1377602&page=1&draft=-1&rank=3
At 1440p, performance with RT vs. rasterizer is 38% on a 3080 Ti and 28% on a 6800 XT, and then it gets worse on lower end models.
The game is also super dependent on CPU performance and RT + 720p doesn't help normalize it like higher resolutions usually do.
Huge difference, AMD really needs to improve with RDNA3 to be more competitive.
The other two have negligible cost on 3090, RT AO - 1.36 ms, RT Shadows - 0.3 ms @ 1440p.
AO and Shadows are practically free in comparison with reflections.
Computerbase tested in 1440p and 2160p without reflections: https://www.computerbase.de/2021-08...t/3/#diagramm-the-ascent-2560-1440-raytracing
The 3070 is still >50% faster than the 6700XT.
The 3070 is still >50% faster than the 6700XT.
Isnt the 3070 a competitor to the 6800xt though? Anyway, i see not much talk on the mobile front, hows a 3070 mobile doing against a 6700m/6800m for example?
Their performing way below their desktop variants, a 3070 mobile is close to a 2070S in raw performance for example. Which aint bad by any means for a laptop, but it doesnt really make sense to name it '3070' without the 'mobile' prefix perhaps. Curious how they perform in ray tracing, they pack less tensor cores as their dGPU variants aswell?
Jawed
Legend
UE4 + tiny dev team effect?
Ye but thats not what i was talking about, MSRP doesnt mean anything these days.
I've not played for a while with UE4, so I wonder whether it has simplified materials for RT shading.
UE5 adjusts shading complexity for RT with the r.Lumen.Reflections.HardwareRayTracing.LightingMode variable, so RT shading can be anything from simplistic lighting with diffuse textures and unified materials to full blown lighting with complex direct pass BRDFs and complex shaders.
Complex shading can put any GPU to knees (shading lighting in reflections for every pixel would be not cheaper than shading primary lighting in rasterization pass). Also, the game applies reflections to a wide range of rough materials, so this can introduce even more perf cost due to divergent shading.
A good denoiser would also be costly and BVH without async overlapping with g-buffer fill pass would not be free either (don't know whether UE4 hides BVH construction cost with async). Other than this, the game looks gorgeous on a 4K display with RT and DLSS Quality.
I honestly can't tell the difference from the screenshots... Maybe a reflection here and there?
D
Deleted member 13524
Guest
Also, are raytraced reflections that much better than screen-space reflections in a game with an isometric view?
I do understand the limitations of SSR in a 1st and 3rd-person view, but would there be much of a difference in a game where the player already has a bird's eye view?
Perhaps this applies to other cheaper rasterization techniques like SSAO in isometric viewpoints, which is why we're seeing so little difference between RT and non-RT despite the massive performance loss.
I do understand the limitations of SSR in a 1st and 3rd-person view, but would there be much of a difference in a game where the player already has a bird's eye view?
Perhaps this applies to other cheaper rasterization techniques like SSAO in isometric viewpoints, which is why we're seeing so little difference between RT and non-RT despite the massive performance loss.
Reflections are of higher fidelity on the RT mode, that much can be seen IMO. I don't know if the game is using SSR (or at what quality settings) for non-RT reflections. Whether that makes a significant contribution to general image quality perception and immersion is up to anyone's tastes, I guess.
All you have to do is look at the Ascent's values for its RT in the dev console with Unreal Engine Unlocker to understand why its RT reflections are so expensive in comparison to other Unreal Engine 4 titles. It is using Full resolution reflections (most titles on the market are at best checkerboard usually (Fortnite), console games are quarter or 1/8th res. It has a very low roughness cut off (0.45 where other UE4 titles are often 0.1) and it includes transparency in reflections (particles). It is using shadow rays to get lighting info in reflections (instead of shadow maps). It is not using the exerimental deffered path which was deved for consoles that cheapens ray tracing by making it more smooth biased among other things.
Its cost makes a lot of sense IMO.
It probably also has an inefficient BVH given how it is indie and has this greebly world. I bet a lot of those tiles they make the levels out of are not even mesh merged.
Its cost makes a lot of sense IMO.
It probably also has an inefficient BVH given how it is indie and has this greebly world. I bet a lot of those tiles they make the levels out of are not even mesh merged.
Because it's hard to spot the difference in screenshots in general.
When you play the game, it's hard to miss reflections of animated billboards, neon signs, mega constructions on metallic floors, in puddles and on walls, as well as reflections of characters in mirrors, puddles, on glossy walls, on metall barrels, etc.
The top side down view makes it only worse for screen space technics since they can't basically reflect anything out of view, such stuff as tall buildings, billboards, etc.
I've almost finished the game and RT reflections make a dramatic impact.
That's fair, and as I said I get the reflections, but I literally see no difference with the shadows and AO.
Yah, when I was playing around with the settings I felt the same way. There may be cutscenes or particular levels where it's noticeable, but just from my playing around with the settings I couldn't notice these two. Even the non RT reflections look good in the game, but I can at least notice the RT reflections.
Maybe UE4 does not do SSR first and uses RT only for failure cases?
In that regard i wonder why this practice is a win at all. Maybe using RT always but only fetch the shaded material from screen where possible would be interesting to try.
That's surely a point, thinking of Path of Exile which does the whole lighting in SS only, even shadows.
Though Ascent has scenes with bigger depth complexity, also at high frequencies. Depends on scene as well.
