Tech

Why should I buy a NVIDIA GeForce RTX 40 Series GPU? – DLSS 3 and DLSS Frame Generation Explained

4K at 120Hz is no match for the new RTX 4090. But there’s more to a graphics card than raw rasterised performance. From real-time raytracing performance to AI-powered frame generation with DLSS 3, here’s why you should buy a NVIDIA RTX 40 Series GPU.

This blog was updated in October 2022.

  • Nowadays, there’s more to a graphics card than raw rasterised performance. You’ve got to look at real-time raytracing and AI performance as well.
  • Image reconstruction is an essential component of modern rendering, and NVIDIA DLSS is arguably the best we’ve seen.
  • DLSS taps into Tensor Cores, exclusive to RTX GPUs.
  • Exclusive to the RTX 40 Series is DLSS 3 and DLSS Frame Generation, which generates entirely new frames.
  • DLSS 3 combines DLSS Super Resolution and DLSS Frame Generation to multiply performance by up to 4x over native rendering.
  • RTX 40 Series GPUs are extremely performant at real-time raytracing – the ‘holy grail’ and future of graphics.
  • DLSS 3 frees up valuable GPU resources so you can enable truly transformative technologies like raytracing.
  • RTX technologies have seen huge adoption among developers, with 216 games and apps and counting supported.

When it comes to raw rasterised performance, it’s no surprise the NVIDIA GeForce RTX 40 Series graphics – like the RTX 4090 and RTX 4080 – are blisteringly fast. There’s no doubt that they’ll achieve well over 60 FPS at a native 4K in many of the latest releases.

Nowadays, however, there’s more to a graphics card purchase than just looking at benchmarks and bar charts. With the advent and rise of real-time raytracing, AI-powered graphics, and image reconstruction, it’s just as important to look at how a given GPU handles these technologies.

Compared to AMD and Intel – who have just entered the dGPU game with Intel Arc – NVIDIA is leading the pack in this regard. And their lead is about to extend even further with the game-changing DLSS 3 and DLSS Frame Generation, exclusive to RTX 40 Series GPUs.

NVIDIA Deep Learning Super Sampling

Image by: NVIDIA

As image reconstruction technologies such as NVIDIA Deep Learning Super Sampling (DLSS), AMD FidelityFX Super Resolution (FSR) and now Intel Xe Super Sampling (XeSS) continue to advance and gain adoption among developers and their engines, it’s safe to say we’re in a post-resolution era.

If you want to run real-time raytracing – the future of graphics – at an acceptable frame rate, reconstruction is practically essential. While each reconstruction technology has its advantages and disadvantages, DLSS is arguably the best we’ve seen. And with the RTX 40 Series, it’s about to get even better.

What is image reconstruction?

It started with 2016’s mid-generation console refresh cycle, which introduced the ‘4K-capable’ PlayStation 4 Pro and, later on, the Xbox One X. Targeting such an absurdly high resolution has forced developers to adapt their approach to rendering graphics. Compared to 1080p (1,920 pixels on the horizontal axis, 1,080 vertical) which was common at the time, 4K (3,840 pixels on the horizontal axis, 2,160 on the vertical) is an enormous increase in resolution. That’s 4x as many pixels, tasking hardware to churn out over 8 million pixels every frame. When gamers also expect greater and greater graphical fidelity and frame rates, exhausting so much horsepower on all those pixels just wasn’t worth it.

So, to make their games presentable on increasingly popular 4K TVs and monitors, developers turned to image reconstruction – transforming a lower-resolution input up to a higher-resolution output.

For instance, a game could be producing a 4K-like image on your display, but underneath it’s only rendering at an internal resolution of 1080p. This gives a massive boost to performance, which can be used to run future technologies like real-time raytracing in the here and now. Early on, you might’ve heard of technologies such as checkerboard rendering and temporal injection, but they’ve advance considerably since then.

Importantly, reconstruction isn’t the same as your basic spatial upscale. It’s not just stretching the image out. Technologies such as DLSS are integrated at an engine level, using motion vectors and other temporal information to generate the highest quality frame possible.

As such, DLSS isn’t a global toggle that applies to every game. A game must be updated to support it, but the list of supported games is ever growing. Many of 2022’s biggest releases, like Call of Duty: Modern Warfare II, will ship with DLSS support at launch.

The AI-powered DLSS Super Resolution is the best in the business

DLSS Super Resolution is NVIDIA’s answer to image reconstruction. In regard to pure image quality, it’s the best in the business. It produces a result that’s not only comparable to native rendering, often it’s even better.  

DLSS has one key advantage that places it above, say, FSR 2.0: it’s powered by AI co-processors called Tensor Cores found exclusively on NVIDIA GeForce RTX graphics cards.

DLSS taps into an RTX GPU’s Tensor Cores to accelerate machine learning algorithms, helping to enhance and reduce artifacting in the image reconstruction. These algorithms are trained against thousands of absurdly high-res images of games run on a supercomputer. So, even if you’re running a game at a low resolution, DLSS can use these images as reference to infer what it’s supposed to look like at a higher resolution with AI.

This AI-driven approach can spot and correct for artifacting with temporal anti-aliasing, such as ghosting behind fast-moving objects. As it extrapolates information from a much higher resolution, DLSS can resolve detail that’s not even there at a lower resolution. If you’re annoyed by how distant and finer details tend to shimmer and ‘pop’ in games – telephone pole wires and hair are notorious examples – DLSS can help clean up this sub-pixel detail.

Sometimes the AI makes a slight error, sure, but all in all DLSS can produce a pristine-looking image that’s arguably better than native rendering. This ‘work smarter not harder’ approach allows DLSS to deliver up to 2x faster performance compared to native rendering. With a NVIDIA GeForce RTX 40 Series graphics card and DLSS enabled, you’ll not only get a crisper image, but faster performance too – a win-win for gaming.

DLSS Frame Generation

NVIDIA’s innovations don’t stop there. DLSS 3 takes it to a whole new level. It’s comprised of three technologies: DLSS Super Resolution (also known as DLSS 2.0), NVIDIA Reflex, and the new DLSS Frame Generation.

DLSS Frame Generation is a breakthrough technology, leveraging AI to analyse sequential frames and create additional high-quality frames. It’s calculated on the RTX 40 Series’ high-speed Optical Flow Sensor – boasting 2x faster performance than the previous generation – which then executes on the forth-gen Tensor Cores.

So, not only does DLSS 3 reconstruct frames up to a higher resolution, it then also creates new ones. When you break it down, DLSS 3 generates 7 out of the 8 pixels you see on your monitor! With both of these technologies combined, DLSS can multiply performance by up to 4x over native rendering.

While DLSS Super Resolution can run on RTX 20 Series and newer, and NVIDIA Reflex on GTX 900 Series and newer, DLSS Frame Generation is as of now exclusive to RTX 40 Series. In theory this technology could run on older RTX architectures, though they’re likely too slow to be worthwhile. NVIDIA says it’d require “further innovation and optimisation for the optical flow algorithm and AI model” to support these architectures.

DLSS Frame Generation Gaming Performance Boost

To give you an idea of how much DLSS Super Resolution and DLSS Frame Generation impact performance, let’s look at a benchmark provided by NVIDIA. In Cyberpunk 2077’s new, incredibly taxing Overdrive Mode at 4K, the RTX 4090 ran at 20 FPS. Turning on DLSS Super Resolution bumps it up to a playable 60 FPS, and with DLSS Frame Generation on top it reaches a mind-blowing 90 FPS.

DLSS can boost performance even when you’re CPU-bound. Say you’re playing a game with a highly detailed physics simulation and an expansive open world – Microsoft Flight Simulator 2020, for instance. This can place a heavy load on the CPU and ‘bottleneck’ the performance of even the fastest RTX 40 Series GPUs. As DLSS Frame Generation is generated as a post-process on the GPU, separate from the normal rendering, it can increase frame rate irrespective of the rest of a system’s hardware.

Using the same integration points as DLSS 2.0 and NVIDIA Reflex, DLSS 3 is just as easy for developers to integrate. It’s already coming to Unreal Engine and Unity, two of the world’s most popular game engines, as well as in-house engines such as EA’s Frostbite. At the RTX 40 Series’ launch, NVIDIA says over 35 games and apps will have DLSS 3 support.

NVIDIA Reflex

NVIDIA’s been working on DLSS Frame Generation for nearly as long as DLSS Super Resolution, and now it’s ready for prime time with the RTX 40 Series. But what’s the catch?

Well, AI’s doing a lot of heaving lifting with DLSS Frame Generation and there’s only so much information it can gleam from the data it’s fed. It can’t predict a fast-moving object suddenly appearing in frame or what’s behind an object, resulting in artificing. But when DLSS 3 is making the frame rate so high, this artificing will persist on screen for only a split second. You’re not going to be able to pick it out with the naked eye.

As DLSS Frame Generation’s frames are generated and presented in-between already rendered frames, the technology introduces a slight hit to latency. This is counteracted somewhat by NVIDIA Reflex, which is now a part of the DLSS 3 package.

Just like DLSS Super Resolution, NVIDIA Reflex is integrated at an engine level, helping to synchronise the CPU and GPU for lower system latency. In fast-paced games that demand the fastest reflexes, you might not want to enable DLSS Frame Generation.

Anyway, these types of games already run at ludicrously fast frame rates. DLSS Frame Generation is more suited to slower, graphically intensive games, in which the advantages of high frame rates are more the smoother animation and not so much the lower input latency.

DLSS Quality, Balanced, Performance & Ultra Performance

What’s more, the latest version of DLSS Super Resolution comes in several pre-sets: Quality, Balanced, Performance, and Ultra Performance. Each one will target your monitor’s output resolution, but the input resolution multiplier will be different. On a 4K monitor DLSS Quality reconstructs from 1440p, whereas DLSS Performance reconstructs from 1080p, giving you even more performance.

You’ll lose some clarity, but depending on your performance target, it may be worth it. Dropping from DLSS Quality down to DLSS Performance might give you the edge to saturate your high refresh rate gaming monitor or enable all those real-time raytracing effects.

Designed for 8K gaming, DLSS Ultra Performance takes this to the extreme with a 9x resolution multiplier that holds up shockingly well. Instead of rendering at a native resolution, DLSS frees up valuable GPU resources so you can enable truly transformative technologies like raytracing.

Why DLSS is so important for raytracing

Raytracing is often hailed as the ‘holy grail’ of computer graphics. Compared to conventional rasterised rendering, however, real-time raytracing is an order of magnitude more challenging to run. To run at an acceptable frame rate, games have long relied on cheap (as in fast to compute) but inaccurate techniques. You don’t even have to look that closely at, say, screen-space reflections to notice their limitations.

On the other hand, raytracing simulates the physical properties of light and how it bounces off surfaces and disperses in an environment, producing accurate reflections, shadows, and lighting in the process.

NVIDIA’s cutting-edge RT Cores on RTX 40 Series graphics cards help speed up ray-traced graphics tremendously. Nonetheless, it’s still extremely taxing. Raytracing is a true generational leap in graphics – the biggest we’ve seen in over a decade – but it’ll take a hit to your frame rate.

That’s why the performance boost DLSS affords is so necessary for raytracing in games. RTX GPUs make real-time raytracing a reality, as you can mitigate the performance penalty with DLSS and bring your games back up to an acceptable frame rate. It’s the secret ingredient that makes raytracing viable, even in games that make heavy use of it, like Cyberpunk 2077’s new ‘Overdrive Mode’.

RTX 40 Series GPUs are a worthy upgrade for owners of old GPUs

2016’s GTX 10 Series graphics cards (such as the GTX 1060 and GTX 1050 Ti) endure to this day, safely securing top spots in the Steam Hardware Survey. The GTX 1060 alone accounts for a 7% share of all reported graphics cards – that’s a lot of gamers who are missing out on DLSS and raytracing.

Between the chip shortages and crypto boom, you’ve likely used the same old hardware for longer than you ever anticipated. But as supply gradually returns to normal levels, now’s the time to upgrade your graphics card.

RTX 20 Series was cutting-edge but needed refinement

While 2018’s RTX 20 Series introduced RTX technologies to the world, for many they weren’t fast enough to justify an upgrade. As the world’s first real-time raytracing graphics cards, in retrospect the RTX 20 Series was way ahead of its time. AMD wouldn’t release raytracing-capable graphics cards until late 2020.

But as tech enthusiasts will know, the first generation of a new product is rarely a home run. It’s as if the RTX 20 Series released too early, before NVIDIA could iterate upon and perfect DLSS’s quality and raytracing’s performance. Raytracing completely tanked frame rates, and DLSS couldn’t save the day as its ‘1.0’ version lacked a temporal element. It didn’t help that developers were blind-sided by the introduction of ray-traced graphics either. Games just weren’t ready to support these emerging technologies.

The RTX 20 Series was cutting-edge but needed refinement. Which the RTX 30 Series brought, but it’s release has such unfortunate timing. Throughout the RTX 30 Series’ lifespan, so many gamers couldn’t get their hands on these cards for a multitude of reasons.

RTX 40 Series: An Unprecedented Generational Leap

Image by: NVIDIA

That’s where the RTX 40 Series comes in, fully capitalising on the RTX concept. Powered by NVIDIA’s third-generation RTX architecture – codename Ada Lovelace – RTX 40 Series GPUs bring an unprecedented generational leap in real-time raytracing and AI performance.

Compared to the previous generation, this architecture introduces new streaming multi-processor for up to 2x performance and power efficiency, forth-gen Tensor Cores for up to 2x AI performance, and third-gen RT Cores for up to 2x raytracing performance. Paired with DLSS 3 and DLSS Frame Generation, run on the new Optical Flow Accelerator exclusive to RTX 40 Series GPUs, gamers will experience incredibly details visuals and massive performance boosts.

RTX has seen huge adoption

In the time between the RTX 20 and 40 Series, DLSS and real-time raytracing support has gained huge traction. So much so that raytracing-capable hardware was built into the PlayStation 5 and Xbox Series X/S. This commitment to raytracing from the major parties makes it clear as day that it’s the future of graphics. Soon these consoles will represent the baseline hardware for development. If you don’t have a raytracing-capable GPU, like one from the RTX 40 Series, you’ll be left behind.

According to NVIDIA, 216 games and apps have incorporated RTX technologies since their release, with more and more releases always on the horizon. You can check out the ever-growing list of games, engines and applications that support RTX here.

Cyberpunk 2077 and Metro Exodus: Enhanced Edition are the stand-out examples, incorporating a kitchen sink’s worth of RTX features. Metro Exodus: Enhanced Edition requires raytracing-capable hardware to run, one of the only games on the market to do so.

While Cyberpunk 2077 is getting updated with an ‘Overdrive Mode’, making its raytracing even more comprehensive.

Now with RTX Direct Illumination, every emissive light source – from neon signs to streetlamps and car headlights – will give off ray-tracing lighting and shadows, bathing surrounding objects in accurate, coloured lighting.

What’s more, NVIDIA has streamlined the integration of DLSS, to the point it’s now a plug-in for popular game engines such as Unreal Engine and Unity. Big-name publishers have also integrated it into their engines, such as EA’s Frostbite, Activision Next-Gen Call of Duty Engine, and more. Now, there’s a very high chance you’ll find RTX technologies in upcoming games.

NVIDIA GeForce RTX 40 Series at Ebuyer

Although impressive, it’s not just the raw speed of the RTX 40 Series that makes it such a compelling product. It’s the technologies it supports that takes it to the next level.

These technologies are where GPU wars are fought nowadays. You wouldn’t see claims like “4x the performance” otherwise. NVIDIA’s the only one out of the big three who’s got a technology like DLSS Frame Generation.

How long is it going to take for their opposition to catch up? At present, AMD doesn’t have an AI-powered image reconstruction technology, never mind frame generation. This might change soon, however, as AMD’s RDNA 3 GPU architecture and RX 7000 Series GPUs are to be unveiled in November.  

So, if you’re rocking an old GTX card or simply want the best technologies, AI, and raytracing performance, upgrade to an RTX 40 Series GPU at Ebuyer.

Most Popular

To Top