The latest batch of updates for Cyberpunk added, in addition to content, significant improvements to the graphics using ray tracing with Ray Reconstruction technology. We’ll see how the improved Cyberpunk 2077 fares with performance and responsiveness, and what effect it has on input lag if you increase performance on GeForce with the help of DLSS frame generation.
Nvidia Reflex and frame generation
The first thing we look at today when measuring performance is frame rate. It gives you an idea of the performance of the system and how smooth the game is. However, you will not be able to tell from the average frame rate how quickly the entire assembly reacts to mouse movement. In modern games with demanding graphics that use post processing, such as Cyberpunk 2077, the game can achieve fairly smooth frame rates, but even with them, playing may not be comfortable when the game reacts to mouse movements with a delay.
System responsiveness is largely related to frame rate. The more often the image is redrawn, the smaller the changes and the delay between individual frames, and thus the delay with which the game responds to the player’s input decreases.
While at 60 frames per second it can take almost 17 ms to render the next frame, at 120 frames per second the delay is reduced to a maximum of 8.3 ms. However, this does not mean that the game will respond to your movement within eight milliseconds – the path from moving the mouse to rendering the image on the monitor is very long, and there are many other delays after it.
Delays accumulate on devices throughout the image processing chain from the sensor in the mouse or keyboard input to the delays caused by lazy crystals on the monitor, the necessary calculations take some time in between, and it does not help that the image processing queues working for more fluidity are used simultaneously on multiple frames at the same time and different buffers that games use to make the rendering of the frames even and the image does not stutter.
If you want the game to respond to the player’s movement as quickly as possible, it is desirable to have as few delays as possible along the way, and this is what Nvidia focused on when creating the Reflex application interface.
Nvidia Reflex SDK is a set of APIs for game developers, which allows not only to reduce latencies, i.e. to speed up the system’s response to controller input, but also to measure response time. It works on all newer GeForces starting with the GTX 900 series. The whole ecosystem and the issue around responsiveness is discussed on several pages on nvidia.com, at this point we will just remind you that integrating the API directly into the game allows you to tune the game engine so that the rendering takes place just-in-time (in the supply it is called just in time), i.e. so that the image does not lie around unnecessarily somewhere in the buffer , but to start rendering with the last data from the input as late as possible – just in time, so that the rendering is completed just before rendering to the monitor.
Thanks to the application interface, which allows more efficient use of the graphics card and shortens the graphics rendering queue directly in the game, the response time can be reduced more significantly than is possible using existing techniques through drivers. The graphics card’s frame timing can be adjusted via the controllers, but it is no longer possible to influence, for example, at what moment the game will read the mouse movement before rendering the frame and when the processor will start working with the scene. However, this can be influenced when integrating the Reflex SDK. The result is sometimes a third shorter response time.
Along with Reflex, you will usually find the Low Latency Boost function in the game settings with the implemented SDK, which allows you to adjust the power management of the card. Under normal circumstances, if the card is not sufficiently busy, it can reduce clocks. This may not be a problem with framerates, because if the chip is not busy, the card manages to render all the frames even at lower clocks, but it can be a problem if you want to render the frame as fast as possible from the moment the game receives input from the player. Boost will set the graphics chip’s power saving functions to stay at maximum clocks even when running at half throttle due to the CPU not having enough time to supply the graphics card. The resulting image is then rendered faster at the maximum clock rate.
Ray Reconstruction: a step further on the way to high-end graphics
Cyberpunk was one of the first games to support the generation of cutscenes. And today it is one of the first titles that, in addition to image generation, also supports Ray Reconstruction technology, which Nvidia introduced this August. With Ray Reconstruction, Nvidia managed to combine two steps from rendering – denoising the image and its reconstruction during DLSS super sampling – into a single step, thereby achieving higher image quality and faster image response to changes in the environment.
A more detailed description of this and an explanation of the following video, which deals with how traditional rendering with ray tracing and DLSS was carried out and where the use of ray reconstruction is moving it, we devoted several pages in the article after the launch of DLSS 3.5, so if you missed it, tap it through.
The entire GeForce RTX 40 model series is already out, so Nvidia prepared a batch of technological improvements at Gamescom. DLSS…
Image generation in DLSS 3.x and system response
With the latest generation of GeForce, DLSS Frame Generation technology has also been added, which helps to increase the frame rate to make the image smoother by adding another intermediate frame between the frames rendered by the graphics card with the help of artificial intelligence, significantly increasing the frame rate.
We have already covered the description of the generation of intermediate frames under DLSS on our website.
Nvidia presented the first three models from the new generation of GeForce. The launch of the RTX 4090 is planned for October, the RTX 4080 awaits us…
One of the things that many people criticize about image generation is the fact that it increases the response time of the system. The intermediate frame is generated between two already rendered frames. Because of this, the graphics card cannot send the latest and most recent rendered frame directly to the monitor, but must put it in a buffer and display the generated frame before it. In the ideal case, the delay of the rendered image will therefore be extended by roughly half the time required for rendering the image – for example, at 60 frames per second, it will be at least one hundred and twentieth of a second, by 8 ms.
It is for this reason that Nvidia has combined image generation with Reflex technology, which shortens the response time. During active image generation, Reflex is also always active at the same time and cannot be turned off. So with frame generation and Reflex, the response will still be shorter than without it. But it will not be shorter than if you turn on Reflex and not generate images – in that case, the system will react to mouse movement a little faster.
Today’s test will focus on how the response time differs when rendering without Reflex technology, with Reflex technology and with image generation, and how much image generation lengthens the response time.
2023-12-28 09:09:19
#Cyberpunk #update #2.x #Overdrive #DLSS #frame #generation #responsiveness
