tl; dr: With the tech demo “Neon Noir”, Crytek demonstrated in March 2019 how raytracing effects can be reproduced without resorting to Microsoft DXR and thus dedicated raytracing hardware such as Nvidia's Turing GPUs. Now the Tech Demo is the benchmark and ComputerBase calls the community for testing.
Ray tracing based on voxels
Crytek uses a grid-based (voxel-based) global lighting, Voxel-Based Global Illumination (SVOGI) called. It does not require dedicated ray tracing units such as Nvidia's RT cores in Turing, but according to Crytek it also runs on conventional “midrange hardware” from AMD and Nvidia.
Part of the work is done by the CPU, which creates a voxel image in real time for each frame – incrementally, as CryTek explains. On the GPU then for each frame “Thousands of rays“Tracked through the grid and the shadow maps. Indirect light should be so static and “the most dynamic“Track objects. For dynamic, this only applies if they are close to the static geometry mapped by voxels. Indirect shadows, in turn, can only be considered by static objects.
Currently with restrictions
On particles or running water, the effects should not yet be applied. Other limitations include offsetting one to two frames in the display of indirect lighting, and it may take a few seconds for the system to work when the camera jumps to a completely different location – obviously, the initial computation of the voxel environment is needed much time. Another limitation: Raytracing is only used for objects that are close to the eye of the beholder.
CryTek, however, speaks of “current limitations“, Wants to tackle the problems so still. Contrary to the announcement in March, the benchmark initially relies on DIrectX 11 and not DirectX 12 or Vulkan.
Benchmark against Crytek account
The benchmark is available as a free asset in the Marketplace for the CryEngine. The nearly 4.4 GB download requires an account at Crytek, the registration of which can be completed by specifying an e-mail address. After registering, Neon Noir can be added to the shopping cart, the actual download also requires the CryEngine Launcher. With this tool, the benchmark can then be downloaded after logging in with the previously created user data. A download of the CryEngine is not required.
The settings for the benchmark
Apart from the resolution, the benchmark published by Crytek offers just one more setting option: the quality of the raytracing can be chosen between “Very high” and “Ultra”. In Ultra the ray tracing effects are calculated in the set resolution, in Very high it is reduced.
Readers who want to test with their system should use both ray tracing levels in the available resolutions. The benchmark is “full screen”.
However, there is still a “problem”: Although the benchmark shows the current FPS and frametimes by overlay, in the end it only delivers one score as the result. In order to determine the FPS, the external tool CapFrameX must still be used.
Tutorial: Determining FPS with CapFrameX
As a benchmark tool, only CapFrameX should be used for optimal comparability. The formerly popular FRAPS no longer works in all games and is limited in functionality. The download of CapFrameX can be done directly via ComputerBase. The developer Zero Trat is also active in the ComputerBase community, which is actively involved in the project.
During installation, please note that the hint can be ignored by Windows Defender. The tool does not have a valid signature, which is why the warning appears automatically – many programs that are not from major manufacturers have the problem. After installation, the program can be found as a shortcut on the desktop and in the Windows start menu or the executable EXE file in the directory “C: Program Files (x86) CapFrameX”.
At the top of the CapFrameX UI are five different tabs, with only the first two playing a role in the community test. All settings for the benchmark are made in the Capture tab. It is important to set the “Capture Hotkey” to start the benchmark according to your own wishes (“F12” uses the editors) and the “Capture time in seconds” on 90 seconds (The measurement then automatically stops after 90 seconds, Neon Noir runs a good four seconds longer). Pressing the capture hotkey in the game starts the benchmark. This must be done exactly when the road becomes visible (not already at the black screen in front of it).
When the benchmark is finished, CapFrameX displays the filename of the created file, the creation date and the corresponding time on the left side. If the program finds the entry, the benchmark has worked. If this is missing, the likelihood is high that in addition to the game an additional process unknown to the program has been started by Windows and CapFrameX therefore did not know which process should be measured. If audio indications are active, CapFrameX indicates this when starting the measurement. In this case, the “Process Ignore List” can be found on the right side. If there is a second process next to the game in “Running Processes”, this must be marked with the mouse and then moved to the ignore list by clicking on the arrow pointing to the right.
In the editorial office, both “GameLauncher” and “CRYENGINELauncher” were active when running the benchmark. “CRYENGINELauncher” must go to the Ignore List.
If the benchmark was successful, the graphical evaluation in CapFrameX is done by clicking on the tab “Analysis”. Then the benchmark run on the left side must be marked by mouse. Even if the tool specifies more data: For the community benchmark, only the average frames per second (FPS) are important. These can be found in the bar graphs under the entry “Average”. The numerical value stated there is the benchmark result. In addition, the tool displays many more series of measurements that can be used for a more detailed analysis or comparing multiple benchmark runs.
This is how results are reported
Results can and must also be submitted in this reader test via the following form. If one of the forms is submitted, the CMS will automatically design a post in the comment thread of this article. Participants can add more information here, however mandatory outside the preformatted area – that is automatically evaluated by our CMS. Only when the post is saved are the values for ComputerBase visible.
It is important to pay attention to the correct name, which must always contain information about the GPU, CPU and RAM. The information must be given in the way that the following list and the already available benchmarks demonstrate – this increases the clarity enormously.
- RTX 2080 FE, R7 2700, 32/2933
- RTX 2080 CD, R7 2700, 32/2933
- RTX 2080 CD OC, R7 2700 OC, 32/2933
- RTX 2080 CD, R7 2700 UV, 32/3200
For GPU and CPU it can be specified if the chip was overclocked (OC) and / or undervolted (UV). For GPUs, it should also be mentioned, whether it is a reference design (FE), a custom design (CD) or a mobile chip (MQ or M). The RAM is marked according to the pattern GB / CLOCK.
- FE = Founders Edition or reference design
- CD = custom design (partner card)
- OC = overclocked
- UV = undervolted
- MQ = Max-Q (mobile GPU from Nvidia)
- M = Mobile GPU (except Max-Q)
The form does not have to be filled in completely, also selected results can be reported. Additional readings can be added later by submitting the form again.
The diagrams below show the results for Full HD, WQHD and Ultra HD (through-switching) with the reduced requirements. The editorial results are in green and red. The results of the community bear the color orange.
Benchmarks in Full HD
Benchmarks in WQHD
WQHD, Raytracing “Ultra”
GTX 1080 Ti CD, R5 3600, 16/3200
WHQD, Raytracing “Very high”
GTX 1080 Ti CD, R5 3600, 16/3200
Benchmarks in Ultra HD
Ultra HD, Raytracing “Ultra”
Ultra HD, Raytracing “Very high”
This article was interesting, helpful or both? The editorial team is happy to receive any support in the form of deactivated ad blockers or a subscription from ComputerBase Pro. More about Ads on ComputerBase.