Earlier this year, NVIDIA announced a new line 50 Series GPU Towed new features: “Multi -frame creation”. This new GPU can create multiple video frames based on a single frame based on the initial frame generation technology. But is it a good thing? Or is it a “fake frame”? Well, it is complicated.

At a very basic level, “frame creation” uses a deep learning AI model to create a frame between the two frames of the GPU rendered by the GPU. The graphics card performs more tasks to create “frame 1” and “frame 3” based on 3D models, lights, and textures, but frame generation tools take these two images and guess “frame 2”. Must be I see it.

Multi -frame generation develops this one step further. Instead of just creating one Create some extra frames, some things. This means that in the highest settings, three of the four frames can be created rather than directly rendering. But whether it is good depends on what type of gameplay and game experience you want.

What is the difference between upscaling and frame creation?

NVIDIA’s new multi -frame generation is provided as part of the DLSS 4 presentation. DLSS means deep learning super sampling, and as the name suggests, the previous repetition is not a frame creation, but rather a super sampling (or upscaling).

In this version, the GPU is rendered of the low resolution version 1080p of the frame and then high -quality at higher resolutions such as 1440p or 2160p (4K). DLSS’s “Deep Learning” refers to providing UPSCALER with a better idea of ​​a high resolution frame to train the machine learning model individually in each game.

Today, DLSS means the entire tool used by NVIDIA to provide better performance, and the above method is usually called super resolution. On the other hand, frame generation takes two full frames and generates a completely new frame from the beginning.

Of course, it is also possible to use all of this technology at the same time. The GPU can end in a technical rendering one low resolution frame for the latest GPU. If it sounds like a lot of exterior, it is. And it works incredibly well. In most cases.

When is frame creation useful?

In a relatively short time, I saw the demand for the GPU exploded. As mentioned above, 4K resolution includes the amount of pixel information in 1080p. In addition, media such as movies and TVs were trapped in a relatively consistent 24-30 frame per second, but gamers require at least 60fps on baseline, and often pushes 120fps or 240fps in high-end machines. And do not start the ridiculous display of Samsung that can be supported. Up to 500fps.

If the GPU needs to calculate all pixels of 4K images of 4K image every second, the results from the PC can be found in space. At least in a game with a detailed ray tracking graphic used in the AAA title, at least in the game, the game can be seen in the space at least.

From this point of view, frame creation is not helpful. In NVIDIA’s latest GPUs, multi -frame creation can increase the frame speed for the game. Hundreds of frames per second It still looks pretty good in 4K. It is not possible at resolution without industrial equipment.

When it works (and we will come back), frame creation can allow softer movements and tensions of eyes. If you want to taste the difference This little tool As long as the display supports, you can experiment with other frame speeds. Compare 30fps and 60fps or 120fps and follow each ball with your eyes. In many games, turning off the motion blur, which will be the default, will make it more effective.

In the case of a confusing game with a lot of movement, this additional frame can be a big advantage, even if it is not accurate. If you look closely at the image of each frame You can see some artifactsBut it may not be noticeable while playing. At least it must work theoretically.

What are the disadvantages of frame creation?

In fact, the operation of this technology can vary greatly for each game, and it may vary greatly depending on how powerful the machine is. For example, you go from 30fps to 60fps by frame generation. It may look like a jankier Rather than going from 60fps to 120fps. This is at least due to the fact that there is more time between reference frames when the frame speed is low. It leads to more noise and relics.

It is also very subjective whether the relics will bother you. For example, if you swing through the city Spider -Man 2And does the tree in the background look more strange than they have to do? On the other hand, the same as the slow game lakeIf the graphics details and set designs are more important for the atmosphere, ghost and bleeding can be more noticeable.

It should also be noted that the artifacts are not. certainly It is inherent in creating all frames. First of all, if the input frame is better, the frame creation can be improved. For example, NVIDIA is promoting a new model of super resolution and light reconstruction. different We simply work on technology to improve the results of light tracking that is not enough time- Improve the image passed In the frame creation of the pipeline.

You can think of it as a huge complicated version of the phone game. The only way to get the most accurate and detailed frame in the game is to render it yourself. The more you add extra pixels and frames, the more opportunities for mistakes are. But our tools are getting better in reducing these mistakes. Therefore, it depends on you to determine if more frames or more details are worth you.

Why is the frame creation about competing games (perhaps)

There is one major exception to this entire debate, which is related to competitive games. When playing online games Overwatch 2,,, Marbleor FortniteThen soft movements are not necessarily a major concern. You may be more interested in the waiting time. In other words, the delay when responding to something and the delay when the game registers the reaction.

Frame creation requires a frame in order, so it complicates the waiting time problem. Remember the previous example. The GPU generates frame 1, then creates frame 3, then presents the frame generator what frame 2 is. In this scenario, the game cannot actually play Show You frame 2 until you find out which frame 3 should be.

In most cases, this is not usually a problem. At 120fps, each frame is about 8.33 milliseconds on the screen. Your brain is unlikely to cause a big problem because the delay cannot be registered. In fact, human reaction time is generally Hundreds of Millycho. Continue to test to get completely non -scientific evidence. This reaction time test. Please let me know if it’s less than 10 milliseconds. I’ll wait.

But this is a problem in competing games because the delay is not the only waiting time you deal with. There is a waiting time between a keyboard and a computer, a computer and a server, a server and another player.

Most of the individual links of the chain can be very low, but it must be synchronized somewhere. That “somewhere” is in the game Tick ​​speed. This is how often you play games that are updated on the server. for example, Overwatch 2 The tick speed is 64. In other words, every second, the server is updated every 64 or 15.63 milliseconds.

For example, in your game, the enemy’s cashi shows the investigation frame 1 on the cross line, but it has not yet been updated to frame 3 when the server was able to tick before the screen was updated. It can mean registering as a miss even if you feel that your shot should hit. This is also one problem Actually obtained Worse By creating multiple frame.

For example, there is a way to relieve this hit. NVIDIA’s reflection technology This reduces the input waiting time in other areas, but it is not completely avoided. If you are playing a fierce online game, it is better to lower the graphics settings without using the frame creation.