Crytek releases Neon Noir Ray Tracing Benchmark

Idk, that depends on whether that sorting is fast enough right? There's a reason they increase the amount of RT cores per GPU, clearly they are in contention as well.

 

This is more proof the 1080TI was a DX11 powerhouse, but in DX12 it falls short

 

Pretty sure that's a dx11 benchmark.

 

It falls short on compute (which we knew compared to AMD for instance). Since this benchmark doesn't require dx12 / DXR, I doubt it shows much about dx12 performance of Pascal, just compute?

 

If i get this right, the demo features a mix of reflections solution seamlessly switching from different techics at runtime to optimize performance?

 

The RT solution here is only used for sharp single-bounce reflections (high ray coherency) with very few dynamic objects, all placed in a narrow/closed low-polygon environment with aggressive LOD scaling. The goal is to minimize the traversal structure re-building (for each frame update) and avoid ray scattering. In other words, the demo itself is very closely designed around the tech's performance limitations.

 

DX12 and Vulkan support is planned at next year, only DX11 version is available for now.

 

ZOTAC RTX 2070 AMP
i5 2500k @4.8GHz
P8Z68-V PRO
16GB DDR3 2200MHz

1080P - ULTRA - 9321 points

Benchmark use ONLY DirectX 11 (You can run it on Windows7 and ANY DX11 graphics card)
Not use hardware Nvidia RT Cores (Nvidia 16/20 series cards)
Not use DXRT on Windows10 (Windows10 1803 or higher)
Not use VulkanRT on Windows7/10

 

I'm worried that my RTX 2080Ti at 4K runs about the same as a GTX 1060 at 1080p. We really need a ray scaler in the future, I'd feel more comfortable with slightly noisier/less accurate raytracing effects with full rasterized shaders. Right now it's all or nothing of sorts.
Like in this demo, the reflections are rendered at 1/4 of the screen resolution. At 4K I wouldn't mind having it at 720p like quality instead of 1080p, with temporal reconstruction the visual hit is likely a lot less compared to the performance gain.

 

3440x1440
Fullscreen
Ultra
i9- 9900x @ 5.0GHz / 2080ti

8004 OC +130 core / +950 Mem
8054 OC +145 core / +1100 Mem
8097 OC +150 core / +1150 Mem
8102 OC +155 core / +1200 Mem
8133 OC +160 core/ +1300 Mem
+165 locked up the benchmark

1920 x 1080
Fullscreen
Ultra

14284 OC +130 core / +950 Mem
14981 OC +160 core/ +1300 Mem

 

Mantle was always open source the reason nvidia acted like it was not existant was because amd's name was so tightly connected with it.

 

And, under dx11, Nvidia saw a smaller gains than AMD due to different driver architecture, iirc.

 

nvidia acted like it didn't exist because they were already doing in their drivers what amd was wanting to achieve with mantle from the games development end.

 

Without Nvidia pushing RTX out, this demo wouldn't even exist. All I'm seeing is everyone jumping on the RT bandwagon. The whole industry has either gone this way or are deep into R&D to implement it into their products within 1-2yrs.

 

The dx11 drivers on nvidia where for sure superior to amds on taking advantage of their cards while amd cards where to a point held back by their own dx11 drivers , now since your reasoning is that they ignored it cause they where doing so well on dx11 then why they supported vulkan on the get go ? Tech savy people already know that a huge portion of the work in it is a resault of direct contribution from the mantle code... Again the reason they supported vulkan so fast was because vulkan is not AMD's Vulkan . But this was not even the reason i did that post , if you see the part i quoted he made it appear like amd was the reason nvidia chose to ignore mantle because amd locked it down or something.

Edit: i wanted to add that again amds approach on the graphics api's with mantle seems that it was the way to go after all the 2 dominant api's went full on with what they wanted to do with mantle... Mantle might failed to stay alive but sure changed the 2 most dominant api's to follow it's ways so in a sense it did what it was supposed to do.

 

DirectX11 hit a wall and wasn't getting any better.

 

Hi guys


I have tried this Crytek Neon Noir Benchmark and for love of god my scores are same like with 1080p or 1440p or even in Ultrawide 3440x1440, scores are around 3200-3300 points

Looks like I'm CPU limited or my CPU does bottleneck very badly in this benchmark

Running 5960x with 4.6Ghz currently and Asus RTX 2080Ti Strix with 2160MHz OC

Although I'm running RAM at 2113MHz which can make a bit difference but still my scores are very low if I'm comparing against the 4690k or 6700k etc and GTX1080Ti

GPU usage its very low in this benchmark, literally sits in 20-40% during the whole benchmark and in only two instances seen full boost clocks at 2145-2160MHz


Thanks, Jura

 

AMD said it would go open source "eventually" but never got that far. It was clearly a PR push. They instead moved their efforts to Vulkan, as a means of effectively making it open source.

But make no mistake, Mantle's code under the Mantle name was never open source.

 

Hi,

something else must be limiting it, like a fps cap, try to check its cfg file..

I get 6K with my OC 980Ti @ 1080p ultra and cpu didn't really have a hard time, usually it was 20-40% in total.

 

oh i see i was almost convinced that mantle was open source ... i stand corrected !