Wednesday, November 13th 2019

Crytek Releases Hardware-Agnostic Raytracing Benchmark "Neon Noir"

Crytek today released the final build for their hardware-agnostic raytracing benchmark. Dubbed Neon Noir, the benchmark had already been showcased in video form back in March 2019, but now it's finally available for download for all interested parties from the Crytek Marketplace. The benchmark currently doesn't support any low-level API such as Vulkan or DX 12, but support for those - and the expected performance improvements - will be implemented in the future.

Neon Noir has its raytracing chops added via an extension of CRYENGINE's SVOGI rendering tool that currently Crytek's games use, including Hunt: Showdown, which will make it easier for developers to explore raytracing implementations that don't require a particular hardware implementation (such as RTX). However, the developer has added that they will add hardware acceleration support in the future, which should only improve performance, and will not add any additional rendering features compared to those that can be achieved already. What are you waiting for? Just follow the link below.
System requirements:
  • AMD Ryzen 5 2500X CPU/Core i7-8700
  • AMD Vega 56 8 GB VRAM/NVIDIA GTX 1070 8 GB VRAM
  • 16 GB System Ram
  • Win 10 x64
  • DX11
Sources: Cryengine Marketplace Neon Noir, via Neon Noir
Add your own comment

55 Comments on Crytek Releases Hardware-Agnostic Raytracing Benchmark "Neon Noir"

#26
bug
IceShroom
Then why you guys were complaining about Vega prices. HBM was not cheap and was more expensive than TU104 die.
Probably because Vega didn't offer a unique feature in return? (And no, HBM is not a feature.)
Posted on Reply
#27
biffzinker
RTX 2060 with a 75 MHz bump over the MSI factory overclock, and 875 MHz+ for the GDDR6
Same clock speed for the GPU/GDDR6 at 1440P

Ultra
1440P: 4781
1080P: 7853
[ATTACH alt="Untitled.jpg"]136397[/ATTACH]

Nothing seems off with the bullet shell casings rasterized rendering or RTRT. Maybe it shows up at the Very High setting?
Posted on Reply
#28
QUANTUMPHYSICS
CRYTEK needs only to release a refresh of CRYSIS.

NANOSUIT vs. North Korea.

Take the stupid aliens out of it and make the game solely about a government issued supersoldier team going up against a bunch of technologically inferior North Koreans.

Melt my computer if you can.
Posted on Reply
#30
xkm1948
With Unigine and Unreal engine gobbling up the dev market Crytek is desprately looking for ways to attract game developers:

"Hey look at us! We got Ray Tracing on DX11! Buy our engine PLEASE!"
Posted on Reply
#31
lordmartin
i7 4790@4GHz, GTX980@1400MHz, ----> score 4434 (1080p ultra fullscreen). so old computer doing very well. i'm waiting for first games for ps5 with RT.
Posted on Reply
#32
biffzinker
xkm1948
With Unigine and Unreal engine gobbling up the dev market
Shouldn't that be Unity instead of Unigine?
Posted on Reply
#33
Lionheart


- 1080p
- Ultra

Disappointing........ :wtf:
Posted on Reply
#34
biffzinker
Lionheart
Disappointing........ :wtf:
I wouldn't call the score disappointing.
Posted on Reply
#35
Lionheart
biffzinker
I wouldn't call the score disappointing.
Well I figured this benchmark was all about GPU horsepower, no RTX involved but it seems your RTX 2060 is edging out my RX 5700 XT which gives me the impression that those RT/tensor cores are kicking in? Idk but I was just expecting a bit more but then again this benchmark is new, don't really have a reference point of performance yet. :toast:
Posted on Reply
#36
Prime2515102
biffzinker
The benchmark seems to run without the additional launcher.
That's good then. You still have to use the launcher to get it in the first place though.
Posted on Reply
#37
biffzinker
Lionheart
Well I figured this benchmark was all about GPU horsepower, no RTX involved but it seems your RTX 2060 is edging out my RX 5700 XT which gives me the impression that those RT/tensor cores are kicking in?
It most likely is the Turing architecture than a hardware assist from RTX hardware on the die. Crytek is suppose to patch in DXR hardware support at a later time.

Raevenlord
However, the developer has added that they will add hardware acceleration support in the future, which should only improve performance,
Posted on Reply
#38
xkm1948
The benchmarks need its own separate thread in the benchmark section of the forum. @T4C Fantasy what you say buddy? pal? comrade?
Posted on Reply
#39
biffzinker
xkm1948
@T4C Fantasy what you say buddy? pal? comrade?
I think he checked out on doing benchmark threads. I haven't seen any activity on the prior threads.
Posted on Reply
#40
Lionheart
xkm1948
The benchmarks need its own separate thread in the benchmark section of the forum. @T4C Fantasy what you say buddy? pal? comrade?
This x100 ;)
Posted on Reply
#41
nguyen
On 2080Ti oced I get:
RT Ultra
1080p : 15400
1440p : 9900
162fps avg and 122fps 1% low at 1080P ultra.

Posted on Reply
#42
Mamya3084
This is with a bios modded 5700 on water. No idea if this score means anything.
This was 1440P with Ultra Raytracing.


Posted on Reply
#43
renz496
Lionheart
Well I figured this benchmark was all about GPU horsepower, no RTX involved but it seems your RTX 2060 is edging out my RX 5700 XT which gives me the impression that those RT/tensor cores are kicking in? Idk but I was just expecting a bit more but then again this benchmark is new, don't really have a reference point of performance yet. :toast:
so far this the only comparison i can find on youtube. maybe we will see more in a few days:

Posted on Reply
#44
nguyen
Seems like the reflections when using AMD gpu is less detailed and has some rendering errors there


with Nvidia


Wccftech clip


My Clip with 2080Ti
Posted on Reply
#45
bug
nguyen
Seems like the reflections when using AMD gpu is less detailed and has some rendering errors there


with Nvidia


Wccftech clip


My Clip with 2080Ti

It's been reported before (and acknowledged by devs) that in the absence of hardware acceleration, this implementation cuts many corners.
It may not be that big of a deal if you won't notice the difference in motion.
Posted on Reply
#46
nguyen
bug
It's been reported before (and acknowledged by devs) that in the absence of hardware acceleration, this implementation cuts many corners.
It may not be that big of a deal if you won't notice the difference in motion.
Kinda defeat the purpose of vendor agnostic ray tracing when one vendor can render ray traced images better than the other ain't it :D. Also could you kindly point me to where the developers admit they were cutting corners ?
Also the torchlight reflections from the drone with AMD is quite jarring.
Posted on Reply
#47
cucker tarlson


2070 super is getting about 40% higher framerate than 5700XT,and that's a hardware-agnostic benchmark that's no using any of turing's additional RT-specific cores.
Radeon VII loses to 5700XT by 10% despite being 10% faster otherwise.
GTX 1080 is 11% faster than Vega 64.
Posted on Reply
#48
spectatorx
If anyone is curious here is a video showing 3800x paired with rx580 nitro+, 1080p ultra settings:


I just recorded this video and finished uploading, right after i finished uploading i see new driver got released... Well, i assume it wouldn't make much of a difference in this particular case.
Posted on Reply
#50
bug
nguyen
Kinda defeat the purpose of vendor agnostic ray tracing when one vendor can render ray traced images better than the other ain't it :D. Also could you kindly point me to where the developers admit they were cutting corners ?
Also the torchlight reflections from the drone with AMD is quite jarring.
You're in luck, I was able to dig this up: https://www.cryengine.com/news/view/how-we-made-neon-noir-ray-traced-reflections-in-cryengine-and-more
Look for the section where the guy explains their use of voxels which will artifact in some (known) circumstances.
Posted on Reply
Add your own comment