The video game market has been moving steadily towards multiplayer titles. According to NVIDIA, 73% of users who buy their products play multiplayer or eSports titles. One of the main points that require optimization is latency, critical when we enjoy multiplayer games. This is why NVIDIA has developed NVIDIA Reflex technology.
NVIDIA Reflex technology is based on a set of elements such as the GPU, G-Sync monitors and software tools that measure and reduce the latency of systems in games. Reducing latency in competitive titles is critical, as an almost instantaneous and accurate response is sought. Mouse and keyboard inputs are intended to be reflected on the screen quickly to provide more precision.
No products found.
Table of Contents
What is latency?
We can define it as the amount of time that elapses since an action is executed until it is reflected on the screen. For example, latency is the time it would take for a pot of water to boil once we put it on the fire. In the case of computing, these latencies are much smaller, but sometimes noticeable. There are two types of latencies:
Network latency
Time that elapses in communication between a user and the multiplayer server. This type of latency is commonly known as Ping. This delay can affect our game as the game's network code is capable of handling latency. The following problems can occur:
- Delayed confirmations: Time that elapses, for example, between when we get a cancellation and it is confirmed on our screen. This can lead to wasting ammo or at a disadvantage to an approaching enemy.
- Delayed interaction: Delay when we interact with the environment, whether by opening doors, chests or others.
- Delay of the opponent's position: This is known as a 'peeker advantage', which we will see later.
We must know how to differentiate that latency with the server has nothing to do with network stability problems. Packet loss or packets out of service are another type of problem, which can be caused by load on the router or desynchronization. The following phenomena can occur in these cases:
- Rubberbanding: It is when we move and after a few seconds we are moved to a previous position. Basically what happens is that we return to the point where the server had us located.
- Desync: It is when a packet loss occurs, leading to a flickering in the network. It seems in this case that the enemies suffer a very brief freeze of time, to later make a small jump to the real position.
These two problems described are not problems with latency, but with the distance to the server. By having to travel a longer distance, there is a longer delay time between the action, the arrival at the server and the return of the response.
System latency
The delay between an input from a peripheral device and its reflection on the screen. Of course, there is a time lapse between a keyboard or mouse press, until it is seen on the screen. This amount of time is known as end-to-end system latency or click-to-screen. This latency has nothing to do with internet latency and affects only peripherals, PC and screen.
It can be appreciated in the following ways:
- Answer's capacity: For example, when we move the mouse too fast to shoot and miss the shot due to the movement of the mouse
- Delayed shots: When we shoot, but the bullet path, the bullet hole and the recoil of the gun are behind the actual mouse click
- Delayed position of the opponent: It is known as' advantage of peeker', which we will see later.
There are three main latencies in the system: the input peripheral such as the mouse, the PC and the screen. Such latency is difficult to describe due to the use of terms such as 'input latency' or 'input lag', which describe various parts of the system's latency.
The 'input latency' is the time it takes for the mouse to process the click. It can be found on a monitor, which is the time it takes for the screen to process the image.
End-to-end system latency
Before continuing it is necessary to define some terms for the 'input delay':
- Peripheral latency: Time it takes an input device to process an action and send it to the PC
- Game latency: The time it takes for the processor to process the input or changes and send them to the GPU for rendering on the screen.
- Rendering: Time from when the frame is queued to be rendered until the GPU displays it in the frame
- PC Latency: The time it takes for a frame to travel through the PC. Game and Render latencies are included
- Display time: The amount of time it takes for the screen to render an image after the GPU has finished rendering the frame
- System latency: Time including end-to-end measurement, ranging from onset of peripheral latency to screen latency
Difference between FPS and system latency
The first thing we need to do is differentiate between the FPS and the system latency. First, we have the number of images that are displayed on the screen per second. This parameter is a rate of return that is measured in FPS (Frames Per Second). The second parameter is the time it takes for an action to be reflected in one of these images, which is called system latency.
Having a computer capable of offering 1000 FPS but it takes a second to display a mouse click on the screen is a disastrous experience. If on the other hand the click is instantly displayed on the screen but we play at 5 FPS, we have not advanced much either
NVIDIA a year ago began to investigate which of the two parameters had more relevance for a good gaming experience. In short, the system latency affected the user in the task of aiming more than the amount of FPS of the monitor.
No products found.
Why is system latency more important?
First of all, we must be clear about the concept of 'hit record'. This term is used by gamers when talking about how the game records the impacts on the opponent. Many of us have been about to smash the mouse against the wall when a shot that we know has been accurate, has not been registered. But has the game really not registered the impact?
NVIDIA has posted a video showing this situation. We can see how at the moment of pressing the crosshair is on the objective, but the success is not registered. The latency of the system and the movement of the opponent, as well as the game engine have the answer. The game has actually interpreted the crosshairs to be behind the target due to system latency and opponent's movement.
All the action time comes into play as well as the information processing, the rendering of the frame and its representation on the screen. For Shooter games, every millisecond matters and a lag of 30-40 ms means a loss of victory.
Peeker Advantage
The defending player is static watching the exit of an obstacle that prevents the attacker from knowing the position of the defender. When we come out from behind a wall or go through a door, curiously, the attacker has the advantage.
The attacker has a two-second advantage to look around in front of the defender holding the position. The positional information of the attacker has a delay with respect to the defender due to the network. Thus the attacker has a small advantage. It allows the attacker to lean out a little and know the position of the rival, due to these two seconds. An effect that usually occurs due to the game's network code or network latency. But, system latency can play a big role in the peeker advantage.
We can see in the video how in the shot the two are at an equal angle and the pings are equal. The difference in both cases is in the latency of the system.
High latency delays the vision of what is actually happening. This gives an important advantage to the rival, who can see us before we see him. When our latency is lower than that of the opponent, the peeker advantage is significantly mitigated. We still have to count the latency of the peer-to-peer network, but with reduced system latency, the peeker advantage is mitigated and we can better defend and improve in attack.
Target accuracy
One wrong shot can make a big difference between surviving and winning or dying and losing. The flickering of the image in the change of frames at the time of shooting is critical. One of the most important workouts we can do is a blink shooting test in the Counter Strike: Global Offensive or Valorant games. We must see the target, move to have it in the crosshairs and shoot accurately in a fraction of a second.
Surely you have ever noticed an inconsistency in the shots, even if you try to perfect the action. This is because aiming is based on a set of small movements, some almost impossible to perceive. When latency is high, the feedback loop time increases, resulting in poor accuracy.
The greater the system latency, the greater the variation in targeting. This means that it is more difficult for us to predict movement and adapt to improve precision. NVIDIA has done a study on this and shows that the lower the latency the higher the precision.
Impact on competitive
Competitive games, high FPS and refresh rates (Hz) offer lower latency, improving the chance that your shooting actions will hit the mark. Small reductions in latency have a high impact on performance.
NVIDIA Research has found that a 12 ms versus 20 ms latency difference can be very significant in targeting performance. The difference between the time it takes to acquire target and shoot with a system latency of 12 ms and 20 ms, a network latency of 182 ms was measured. This means that the network latency is 22 times the system latency.
Let's put the data in perspective. This means that with the same objective difficulty, 128 ticks are given on a Valorant or CS: GO server. Shots fired hit the target an average of 23 ticks earlier on a 12nm latency system. The problem is that most users play on systems with latencies between 50-100 ms.
That we improve the latency of the system (and that of the network) is not indicative that we will be the best in the world. To be good at mechanical shooters requires more than low latency and mechanical skill. You must read the game and have a robust and clear strategy, elements that make the difference between victory or defeat. Despite this, NVIDIA highlights that it has measured data in PUBG and Fortnite that show an important correlation between high FPS rates (low latency) and the K / D ratio (low / deaths).
Reduce Latency Using NVIDIA Reflex
For this technology NVIDIA has proposed to develop a rendering system for latency that combines an SDK and optimized drivers. Some of the technical solutions can dramatically reduce latency, while others will have a more modest impact, depending on the situation. But the ultimate goal is to provide gamers and developers with the optimal tools to reduce system latency.
NVIDIA Reflex SDK
It allows developers to implement modes that offer low latency that adjust the work of the graphics engine to complete the rendering 'just-in-time'. The GPU render queue is eliminated and dependency on the CPU is reduced in situations where the GPU is heavily impacted.
We can see how the work queue is full of elements. The CPU is processing frames faster than the GPU's ability to render them by generating a backup, resulting in an increase in processing latency. NVIDIA Reflex SDK has some similarities to the driver's ultra-low latency mode. However, when integrated directly into the game you can manage the amount of work the CPU receives from the render queue and downstream.
Ultra-low latency mode can typically reduce the render queue, but cannot correct for increased CPU and game consideration. The latency benefits of the SDK are better than the ultra-low latency mode of the drivers themselves.
Using the SDK, developers can effectively delay input sampling and game simulation in-game by dynamically adjusting the render job submission time to the GPU for just-in-time processing.
No products found.
Low Latency Boost
This feature overrides all GPU power saving specifications to deliver high graphics frequencies when heavily dependent on the processor. Even when the game is under CPU load, high render times add latency. Raising the frequency consumes more power, but at the same time can reduce latency when the GPU is being underutilized and the CPU commands the final rendering work in a large batch.
Note that competitive players tend to use low resolutions when playing first-person shooters. The higher the resolution, the higher the rendering load and latency. Using NVIDIA Reflex you can get low latency at high resolutions. This allows your watch competitive gamers to be able to play with better image quality.
Shooter titles tend to vary the load dynamically between the GPU and the CPU. When there is an explosion with a lot of particles, the game loads the GPU and SKD Reflex will keep the latency low by not letting the job queue stop at the GPU. If it is a simple rendering and is bound to the CPU, SKD Reflex will keep the latency low by increasing the GPU clock rates. Regardless, the Reflex SDK keeps latency low with smart balancing. This allows gamers to have a sweet spot of latency without having to reduce graphic quality.
Reflex SDK also supports GeForce GTX graphics from 2014 to the present. The Boost system to reduce latency is only compatible with the RTX 3000.
Games with support for NVIDIA Reflex
Games already supported by NVIDIA Reflex:
- Valorant (Support with Game Ready drivers released on September 17, 2020)
- Fortnite (Support with Game Ready drivers released on September 17, 2020)
Additionally, the following games have announced support for NVIDIA Reflex, although they do not support it yet.
- Apex Legends
- Call of Duty: Black Ops Cold War
- Call of Duty: War zone
- Destiny 2
- Cuisine royale
- E
- Kovaak 2.0
- mordhau
- The Plague Tale: Requiem
- Atomic Heart
- Battlefield 2042
- Bright Memory Infinite
- Call of Duty: Modern Warfare
- Call of Duty: Modern Warfare II
- Call of Duty: Vanguard
- Counter-Strike 2
- CRSED FOAD
- Cyberpunk 2077
- Cylce Frontier
- Deathloop
- Deathverse: Let it Die
- Deep Rock Galactic
- Deliver Us Mars
- disorder
- Destrroy All Humans 2
- Diablo IV
- Dota 2
- Dying Light 2: Stay Human
- Escape from Tarkov
- FIST Forged in Shadow Torch
- Fortnite
- Ghost runner
- God of War
- hifi rush
- Hitman: Word of Assassination
- Hogwarts legacy
- icarus
- iRacing
- Jurassic world evolution 2
- LEAP
- loopmancer
- Metal: Hellsinger
- Microsoft Flight Simulator
- Midnight ghost hunt
- My Time at Sandrock
- Naraka: Bladepoint
- Nine to Five
- Overwatch 2
- Party Animals
- Perish
- RTX-gateway
- Quakes: Champions
- Rainbox six siege
- Rainbox Six Extraction
- Ready or Not
- Red Dead Redemption 2
- redfall
- return
- Peace
- Sackboy
- scathe
- ShadowWarrior 3
- soda crisis
- Spider-Man Miles Morales
- Spiderman Remastered
- splitgate
- Super people
- The Finals
- the First Descendant
- The Lord of the Rings: Gollum
- the Swordsmen X: Survival
- Valorant
- War Thunder
- Warface
- Warhammer 40.000: Darktide
- warhaven
- Warstride Challenges
- The Witcher III
- witch fire
- World of Warcraft
- WRC Generations
No products found.
Dashboard improvements for latency management
NVIDIA Reflex Low Latency Mode
For games that do not have support for the Reflex SDK, from andl NVIDIA Control Panel the mode can be enabled NVIDIA Ultra Low Latency. It's as simple as opening the control panel and going to Manage 3D Settings. Here we select the Low Latency Mode and we select the option Ultra.
In case the game is compatible with NVIDIA Reflex Low Latency It is recommended to use the controller mode. But if we have both activated, the Reflex Low Latency mode will have higher priority in the system.
Power Management Mode
NVIDIA drivers have always incorporated the option Power management mode. This option allows the user to choose how the GPU works when linked to the CPU. When the GPU is saturated, it works at maximum frequency to give maximum performance. If the load is low, the graph can reduce the frequencies, reducing the consumption and maintaining the FPS.
Like the SKD Reflex's low latency Boost feature, the Maximum Performance Mode it eliminates the power-saving features of the GPU and allows the GPU to work at the highest possible frequency for longer. Higher frequency for a longer time reduces latency for processor-bound instances with a higher balance of power consumption. This is intended for the most demanding users who want to minimize latency regardless of power.
The RTX 3000 allows you to set a higher frequency, allowing the GPU to have a very low rendering latency when linked to the CPU. Those who have a previous generation graphics will be able to choose the mode Prefer Maximum Performance and maintain base working frequencies.
Automatic syncing in GeForce Experience
Within the GeForce Experience software, which will be updated in September, a beta feature will be added within the Performance Dashboard Overlay in-game that allows you to set GPU latency to a minimum with just one click without exiting the game.
This system performs a GPU analysis to maximize the frequency at each point of the stress curve. After finding and applying the perfect settings for our GPU, run a test for a stable period checking the stability.
No products found.
Latency measurement
The reason the system latency had not been adjusted was because it was very difficult to measure accurately. Measuring latency requires high precision measurement instruments that can read the start and end time of the measurement.
System latency measurement could only be measured using high-speed cameras, which are expensive and difficult to handle. In addition to this, an engineering team and a modified mouse with LEDs capable of recording the mouse click were needed.
Using a high speed 1000 FPS camera, the minimum measurable latency time is 1ms. The problem is that this type of equipment has a minimum cost of 7000 dollars. Even so, when you have the configuration it takes about 3 minutes to perform all the measurements, something unfeasible for gaming.
Latency Analyzer
Monitors with G-Sync technology that support 360Hz that will be launched in Q4 2020, will arrive with NVIDIA Reflex Latency Analyzer technology. This solution allows users to read the response time of the system. In turn, you can understand and adjust the performance of your computer before starting the game.
This function requires connecting the mouse to the USB port assigned for the latency analyzer on G-Sync monitors with a 360Hz refresh rate. Our monitor's USB Reflex port is a simple middle ground that analyzes mouse clicks without adding latency.
Reflex Latency Analyzer can detect mouse clicks by measuring the time it takes for a pixel change on the screen, giving you a complete measurement of system latency. The new GeForce Experience feature displays latency metrics in real time. This option is as simple as enabling “latency metrics” in the “performance overlay” settings.
Here the system latency metrics are broken down into:
- Mouse
- PC + screen
- System
We can put any mouse for this function, except for mice that go via Bluetooth. NVIDIA recommends ASUS, Logitech, Razer, or SteelSeries mice. These mice will be able to measure peripheral latency and obtain a complete end-to-end latency measurement.
NVIDIA is also committed to launching an open database with average mouse latencies. In addition, users will be able to add their values to the database in order to enrich the experience as much as possible.
Software metrics
We can measure system latencies without the need to buy a G-Sync monitor with a 360Hz refresh rate. Those games that integrate the NVIDIA Reflex SDK have the ability to display game latency and render latency metrics. It will not offer us a complete measurement of latency, but it does allow you to create a latency optimization.
GeForce Experience additionally adds in the overlay the ability to track rendering latency in any game. Render latency is measured by tracking the render queue and rendering by the GPU. Because it is the final rendering of a frame, the rendering latency will be slightly less accurate than the average rendering latency measured with the NVIDIA Reflex SDK, still a good figure.
To obtain these metrics we need the latest GeForce Game Ready drivers and GeForce Experience software. We must select in the 'Performance' menu and the parameters 'Latency metrics' and enable the 'Performance overlay'
AIM Training with NVIDIA Reflex Lower Latency
NVIDIA has also partnered with The Meta, developers of KovaaK 2.0 to introduce an NVIDIA Experiments mode. This mode will allow users to improve performance and hone their skills.
Within this mode we can choose the type of experiment that interests us the most. The NVIDIA Reflex SDK has been integrated into KovaaK 2.0, among other technologies that allow gamers to feel the experience between high and low latency on the system.
Participating in Experiments will not only help improve your aim, but also an important investigation in the world of eSports. One of the early experiments, for example, seeks to add science to target color choice based on competitive players' discussions around the environment colors in Valorant. Other experiments will test different latency ranges while adding challenging tasks to complete.
