NVIDIA GeForce RTX 4090 Founders Edition
USD 1,599 USD and AUD 2,959
October 12, 2022
Hardware generations work differently in the PC space, with new architecture arriving every two years versus the five-plus in the console space. And with that, you might think that generational improvements of any serious eye-opening magnitude might happen every two generations. Traditionally, that’s been the case.
generation arrived on the scene in the latter part of 2020, with the NVIDIA GeForce RTX 3080
and GeForce RTX 3090
combo presenting the sort of generational leap in performance and cutting-edge features we rarely see. This was covered in our reviews at the time, where alongside the arrival of fresh consoles in the form of the PS5 and Xbox Series X|S, we got to see pure next-gen goodness on PC.
The GeForce RTX 3080 and GeForce RTX 3090 felt next-level, with uncompromised 4K rendering, high frame-rates, AI-powered DLSS, and the sort of ray-tracing that took Cyberpunk 2077
’s stunning Night City into the realm of neon futuristic bliss. 2020 wasn’t that long ago either, so the RTX 3080 and RTX 3090 are still extremely capable and powerful. They are GPU beasts.
And with that, you might think that the RTX 3080 and RTX 3090 follow-up would be a revision or a modest step forward.
One of the most forward-thinking bits of technology with this new Ada generation arrives in the form of DLSS 3. Ground-breaking frame generation that leverages advanced AI Tensor Core hardware in the 40 series to increase performance by a factor of 2X or even 4X.
Fast forward to 2022, and NVIDIA is debuting the latest in GeForce RTX technology - the GeForce RTX 40 series, powered by the new Ada architecture. One of the most forward-thinking bits of technology with this new Ada generation arrives in the form of DLSS 3. Ground-breaking frame generation that leverages advanced AI Tensor Core hardware in the 40 series to increase performance by a factor of 2X or even 4X. AI supercomputer stuff that delivers double the AI processing power of even the RTX 3090 Ti. Which until today, was the most impressive (and somewhat over-the-top) graphics card available.
So here we are with the flagship Ada release, the GeForce RTX 4090 Founders Edition. Which is not only large in stature, but like the GeForce RTX 3090 before it, mind-blowing in what it can achieve and deliver. The question “is there a need for a new generation when the RTX 3090 series was so good?” is answered with a series of exclamation marks. That’s a question that arose for us here at AusGamers, on account of gaming in 2022 with an RTX 3090 still being so impressive. A card that still outperforms the PS5 and Xbox Series X by magnitudes.
Beyond the sorcery that is DLSS 3 in action, getting to see a 30-40% increase in 4K performance over the previous generation would have been wonderful. Awesome even. Beastly. The tagline for the GeForce RTX 4090 is “Beyond Fast”, and even that can’t prepare you. In terms of 4K performance -- even without DLSS 3 -- the GeForce RTX 4090 can double the performance of the GeForce RTX 3090. Yeah, a remarkable and enormous engineering feat bolstered by the new Ada architecture. Architecture that is cutting-edge and efficient, so much so that the 450W power rating rarely got tapped into during our benchmarks.
The funny thing is the GeForce RTX 4090 can easily be overclocked to boost performance even further. It is something else. Beyond fast
. It’s the biggest generational leap we’ve ever seen. We’re talking triple-digit 4K frame-rates in every game we tested.
Introducing NVIDIA Ada
The below is a summary of the RTX 40 Series technology, applicable to all RTX 4090 models. An Ada Lovelace primer if you will.
Across PC and console, from indie games to competitive shooters to AAA blockbusters, we’re now in the age of visual fidelity plus
performance. This is a transition that has happened somewhat gradually over the past decade, but we’re now firmly at a point where CG-like visuals need to be backed up by 60 frames-per-second performance at a minimum.
With the GeForce RTX 30 series, we saw NVIDIA take the RTX name to new heights across the entire range - with visual fidelity plus
performance at the heart of it all. From second-generation RT Cores (NVIDIA’s dedicated ray-tracing hardware) to third-generation Tensor Cores (the AI stuff that makes DLSS the perfect match for ray-tracing and boosting performance), the Samsung 8nm process and chipset also redesigned how calculations and other things were handled. Doubling FP32 and INT32 operations... you know, technical stuff that you don’t
need to understand.
The tagline for the GeForce RTX 4090 is “Beyond Fast”, and even that can’t prepare you. In terms of 4K performance, even without DLSS 3, the GeForce RTX 4090 can double the performance of the GeForce RTX 3090.
The proof was on-screen though, and the added bonus of HDMI 2.1 support and a PCIe Gen 4 interface helped handle all of the extra data and stuff needed as high-end gaming moved beyond 1080p into the realm of 1440p and 4K resolutions. NVIDIA even took steps to improve system latency with the arrival of NVIDIA Reflex, which is now a key part of many action-focused games on top of being a competitive staple in the likes of Apex Legends
The new NVIDIA Ada Lovelace GPU architecture is named after mathematician Ada Lovelace, considered to be the world’s first computer programmer. Built on cutting-edge TSMC 4N process technology, the process shrinkage not only means more of everything in terms of NVIDIA CUDA cores and RT Cores, but improves efficiency and paves the way for advances in how real-time graphics are rendered on-screen. From a pure numbers perspective, the full NVIDIA Ada GPU features 76.3 billion transistors, up to 18,432 CUDA Cores (70% more than the previous gen), and delivers clock speeds in excess of 2.5 GHz whilst maintaining the same power requirements of the GeForce RTX 3090 Ti. Impressive.
Behind the scenes... well, inside
the silicon there’s the new Shader Execution Reordering which basically means that rendering tasks happen more efficiently and without bottlenecks. With ray-tracing becoming more and more prevalent with each passing year, it’s the sort of technology that boosts performance by simply Marie Kondo-ing workloads to be more organised, efficient, and joyous. Naturally, the joy comes once you see the improved performance.
Ray-tracing is the realistic rendering of light, covering shadows, reflections, and how the sun interacts with objects. Traditionally games bake (pre-render) lighting because the effect is costly to do in real-time, and more prominently found in the realm of film and Pixar animation. With NVIDIA being a pioneer when it comes to real-time ray-tracing hardware, thanks to the Turing-based RTX 20-Series, Ada RT represents the next evolution in the stuff that makes games like Cyberpunk 2077, Dying Light 2, Metro Exodus, Ghostwire: Tokyo, DOOM Eternal and more, look as impressive as they do when all of the detail settings are cranked.
Ada’s new AV1 encoder is 40% more efficient than the H.264 video encoder used in the GeForce RTX 30 Series. For streamers and content creators, the GeForce RTX 40 Series is also a game changer, delivering better quality video using the same bitrate.
Ada RT Cores are capable of doing all of the complex calculations to deliver real-time ray-tracing two times faster than the GeForce RTX 30 Series’ Ampere architecture. And with that CD Projekt Red will be releasing an update to Cyberpunk 2077 to significantly upgrade the level of ray-tracing in the game, which is even more impressive when you factor in that a high-end GeForce RTX 30 series graphics card is required to run the game in its current state with full ray-tracing.
This ‘Overdrive’ update will make full use of DLSS 3, a cutting-edge form of neural graphics exclusive to the GeForce RTX 40 series. DLSS 3 takes advantage of new Optical Flow Acceleration to generate entire frames. It’s all thanks to the Ada Fourth-Generation Tensor Core, hardware that turns the GeForce RTX 4090 into an AI supercomputer of sorts. On top of this, there are also advances in video encoding. Ada’s new AV1 encoder is 40% more efficient than the H.264 video encoder used in the GeForce RTX 30 Series. For streamers and content creators, the GeForce RTX 40 Series is also a game-changer, delivering better quality video using the same bitrate.
Ada, and the GeForce RTX 4090, represent a monumental leap forward for graphics technology.
Of Numbers and Heat Sinks
In terms of how it looks, the GeForce RTX 4090 Founders Edition is an extension of the striking FE design NVIDIA introduced with the RTX 3090 and RTX 3080. But even though the RTX 4090 shares visual similarities with the RTX 3090, there are significant cooling advances made that result in better airflow that makes the heatsink more effective. In practice, the RTX 4090 idles silently (drawing around 20W in power), and when pushing through benchmark after benchmark the temperature rarely rises above 65 degrees.
NVIDIA GeForce RTX 4090 Founders Edition
- Architecture (GPU): Ada
- CUDA Cores: 16384
- RT Cores: 128 (3rd Gen)
- Tensor Cores: 512 (4th Gen)
- GPU Boost Clock: 2520 MHz
- Memory Capacity: 24GB GDDR6X
- Memory Interface: 384-bit
- Memory Bandwidth: 1008 GB/s
- TDP: 450W
Of course, the differences between the fins and the stylish heatsink and form factor are massive. Here are the GeForce RTX 4090 stats compared to the GeForce RTX 3090 Ti - which features a similar power rating of 450W.
Here, the magic number is 52% which represents the pure increase in CUDA Core, RT Core, and Tensor Core counts. Of course, the new generation of Ada makes improvements beyond simply upping the numbers, as seen with DLSS 3 which leverages brand-new hardware inside the RTX 4090. Which is the reason why, as of now, DLSS 3 isn’t possible on the GeForce RTX 3090 or any other GeForce RTX 30 series card.
An added bonus of Ada architecture being so efficient (compared to Ampere) on top of being powerful, is that the RTX 4090 can draw far less power than the 350W-rated RTX 3090 and deliver better and more stable performance. In fact, in more than a few benchmarks, the GeForce RTX 4090 drew less power than equivalent GeForce RTX 3090 tests.
Speaking of benchmarks, which tell the performance story through a series of bar charts and numbers, let's get to it.
Calling the GeForce RTX 4090 “something else” in the introduction might have read like a throwaway line - something fun to say to build up a little hype. It wasn’t. And you’re about to see why, triple-digit 1440p and 4K performance numbers (that is, average frames-per-second) that represent a massive leap forward over the previous Ampere launch flagship GPU - the GeForce RTX 3090.
The RTX 4090 is… something else. 260+ frames-per-second DOOM Eternal in 4K with ray tracing is simply incredible to witness in action. A 1.7X increase over the GeForce RTX 3090.
When testing in 3D Mark, which offers a benchmark score, we decided to see how the GeForce RTX 4090 Founders Edition stacked up. Basically, the results broke single-GPU world records. That is GeForce RTX 3090 Ti graphics cards overclocked, running at insane power levels, and cooled with liquid nitrogen.
But first though, a quick look at our new 2022 testing rig.
- GPU: NVIDIA GeForce RTX 4090 Founders Edition
- CPU: AMD Ryzen 9 7900X
- Motherboard: MSI MAG X670E Carbon WiFi
- Memory: Corsair DOMINATOR PLATINUM RGB DDR5 DRAM 5200MHz
Here’s the first batch of games tested in 1440p and 4K resolutions with all detail settings set to Ultra-equivalent levels and DLSS and ray-tracing enabled where possible.
One of the more interesting stories in the lead-up to the RTX 4090 launch was Blizzard allowing Overwatch 2 to render at over 500 fps in 1440p simply because the GeForce RTX 4090 could run the game that fast. The reason why the GeForce RTX 4090 sees the biggest gains in 4K isn’t purely because that’s the resolution that demands the most out of a GPU. We’re talking four times the pixel count of 1080p. But, it shows that the RTX 4090 is potentially too much for 1440p. 1080p results aren’t included in this review because that’s the resolution where the RTX 4090 essentially runs a game as fast as a game engine allows. With Death Stranding topping out at 240 fps, the result is 1440p and 4K numbers that are similar.
This review will focus on 4K performance which, as hinted at, sits in the triple-digit range no matter if a game features ray-tracing or DLSS. Assassin’s Creed Valhalla running at 123 fps in 4K represents a whopping 81% or 1.8X increase over the GeForce RTX 3090, whilst Gear Tactics’ 177.7 fps result represents a 118% or 2.2X increase. Both titles do not feature DLSS, nor does Forza Horizon 5, which runs 1.8X faster in 4K than the RTX 3090. There are of course exceptions, like Horizon Zero Dawn and Guardians of the Galaxy, the latter featuring ray-tracing and DLSS, but even those results represent a 1.6X increase over the GeForce RTX 3090.
The RTX 4090 is… something else. 260+ frames-per-second DOOM Eternal in 4K with ray-tracing is simply incredible to witness in action. A 1.7X increase over the GeForce RTX 3090.
One thing that stood out during all RTX 4090 tests was how consistent everything was, from 1% lows being close to average, through to the power usage and temperatures rarely -- if ever -- spiking. As mentioned above, the RTX 4090 can draw less power than the 350W-rated RTX 3090 and deliver improved and more stable performance. It’s pretty incredible, and with the second batch of games above The Division 2 in 4K represents a 2X increase in performance over the GeForce RTX 3090.
One thing that stood out during all RTX 4090 tests was how consistent everything was, from 1% lows being close to average, through to the power usage and temperatures rarely - if ever - spiking.
Even the intensive Total War: Warhammer III runs at over 100 fps, alongside the visually stunning Red Dead Redemption 2 running at over 150 fps. Metro Exodus Enhanced, which features a number of ray-tracing effects as standard and requires a high-end GPU even to run in 1080p, also hits triple-digits. The GeForce RTX 4090 is a 4K monster.
Click Here to View GeForce RTX 4090 vs GeForce RTX 3090 Results Across 18 Games.
To provide a clearer picture of the 3D Mark results we’ve included GeForce RTX 3090 results captured on the same system with the same software and drivers. The key to understanding 3D Mark results is to look at the various sets of tests targeting different resolutions and different APIs.
Fire Strike is a DirectX 11 test, with the three benchmarks covering 1080p, 1440p, and 4K resolutions. Comparing the RTX 4090 numbers to the RTX 3090 the increase across the three Fire Strike tests, and resolutions, is 1.4X, 1.7X, and 2.0X. Time Spy Extreme, which is a 4K DirectX 12 test, sees a 1.7X increase. Finally, Port Royal - which tests ray-tracing, sees a 2.0X increase over the RTX 3090.
Results that are not only in line with the raw game performance seen above, but a testament to the sheer power of the RTX 4090. And then there’s DLSS 3.
You might have noticed that Cyberpunk 2077 wasn’t included in the tests above, and that’s because NVIDIA provided reviewers with a special branch to be able to play and test the game using the new DLSS 3 rendering - amongst other titles and tech demos. Instead of being the next version of the impressive AI-rendering tech - the 3 is actually in reference to the combination of DLSS Super Resolution (which is what DLSS 2 is today), DLSS Frame Generation (the new magic), and NVIDIA Reflex. DLSS Frame Generation is limited to Ada-based GPUs like the RTX 4090 on account of it using brand-new AI hardware to generate frames.
Cyberpunk 2077 sees a 3.4X increase in performance, F1 22, a 2.4X increase, and even the CPU-bound Microsoft Flight Simulator sees a 2.1X increase in performance. Yeah, and we flew over the dense Sydney city to capture that result.
One of the best things about it, apart from the fact that it works (and looks great), is how it goes beyond brute force to deliver performance. A 2X to 4X increase over the already formidable “brute force” of the GeForce RTX 4090. It takes the most visually intensive games that the RTX 4090 can run in 4K and makes them run at frame-rates that shouldn’t be possible for at least another four years.
How does it do it? Well, we’d be lying if we said we fully understood the tech - but it's a blend of looking at current and prior game frames, taking into account motion vectors and depth, with the secret sauce being an “optical flow field” generated by the Optical Flow Accelerator inside the Ada architecture. What’s an “optical flow field”? Something pretty advanced, as it calculates particle movement, reflections, shadows, lighting and more, to figure out what each pixel is doing. Naturally, the addition of NVIDIA Reflex ensures that system latency doesn’t become a problem.
Is it perfect? No. But based on initial tests, artifacts and issues are just about impossible to spot unless you’re zooming in and comparing frames. As per above the results are insane, incredible, and unbelievable. Cyberpunk 2077 sees a 3.4X increase in performance, F1 22, a 2.4X increase, and even the CPU-bound Microsoft Flight Simulator sees a 2.1X increase in performance. Yeah, and we flew over the dense Sydney city to capture that result. The Enemies tech demo saw the biggest increase with 3.3X, which was remarkable because the demo effectively renders and animates a realistic-looking human being.
DLSS 3 is another game changer in the making, with NIVIDA already confirming that over 30 titles will see DLSS 3 integrated in the coming months.
Beyond Fast… And Then Some
The GeForce RTX 4090 and the new Ada architecture represents the biggest gen-on-gen increase in performance we’ve seen to date, with the result delivering 4K performance that’s truly impressive. A GPU tailor-made for those that want nothing but the best, or those that do their PC gaming on a 4K 120Hz LG-OLED. And even with all of the numbers above, there’s plenty of headroom for overclocking. We were able to push clock speeds on the GeForce RTX 4090 Founders Edition close to 3 GHz - with performance, temperatures, and power levels all remarkably stable.
The GeForce RTX 4090 and the new Ada architecture represents the biggest gen-on-gen increase in performance we’ve seen to date, with the result delivering 4K performance that’s truly impressive.
Two years ago Ampere and the GeForce RTX 3080 launched and it was one of the most impressive NVIDIA launches in the company’s long history. The fact that Ada and the RTX 4090 push things even further, into brand-new territory with the likes of DLSS 3, it’s a little strange to wrap your head around. A generational leap like this, so soon, might seem unnecessary - and you wouldn’t be out of place thinking that. That is until you’re blown away by the results. Of course, the high price point for the GeForce RTX 4090 (especially in Australia) makes it an enthusiast-level opening salvo for Ada.
But what a beginning. And really, it makes the prospect of the upcoming GeForce RTX 4080, RTX 4070 and eventual RTX 4060, something to look forward to. From DLSS 3 to new video encoding that will create a new era for streamers, content creators, and their audiences, to simply being able to deliver visual fidelity plus
performance in the games we play. The GeForce RTX 4090 is beyond fast, and then some.