Nvidia RTX 4090 Vs. RTX 3090 [Gaming Benchmarks]

Colossal in size, extreme performance, and a power-hungry beast are some words you are likely to use when describing NVIDIA’s new GeForce RTX 4090. Since the card is here to dethrone the RTX 3090, it’s only fair to do a complete RTX 4090 vs. RTX 3090 guide.

Previously, we did the ultimate GeForce 4090 vs. GeForce 3090 Ti battle, and to no one’s surprise, the GeForce 4090 ran the games with significantly better FPS.

So, in this guide, we’ll do a thorough comparison of the RTX 4090 with the RTX 3090 to figure out how much better the new graphics card is.

Key Takeaways

• The GeForce RTX 4090 has 5888 more CUDA cores, 1.7x more transistors, up to 925 MHz more base, boost clock speeds, a smaller die, and a 4nm process node. It also supports DLSS 3.0 and SER.
• During our gaming benchmarks analysis, the RTX 4090 left the RTX 3090 in the dust with over 73% better performance on average. Not only that, but the Ada Lovelace GPU also ran ~0.7 cooler on average in the 10 games we tested. 
• The RTX 4090 also consumed about 18% more power on average than the RTX 3090.
• For now, the RTX 4090 is available on Amazon for over $2000, and the RTX 3090 is available for about $999.

Comparison Table

Detailed Specifications

To further understand how much of a power bump the RTX 4090 packs, let’s go over the detailed specifications of the two graphics cards first.

GeForce RTX 4090 Specifications

• Release Date: Sep 20th, 2022
• Launch Price: 1,599 USD
• GPU Name: AD102
• Architecture: Ada Lovelace
• Process Size: 4 nm
• Transistors: 76,300 million
• Die Size: 608 mm²
• Number of CUDA Cores: 16384
• Base Clock: 2235 MHz
• Boost Clock: 2520 MHz
• Memory Capacity: 24GB GDDR6X
• Memory Clock: 1313 MHz (21 Gbps effective)
• Shading Units: 16384
• TMUs: 512
• ROPs: 176
• SM Count: 128
• Tensor Cores: 512
• RT Cores: 128
• L1 Cache: 128 KB (per SM)
• L2 Cache: 72 MB
  

Features

• DirectX: 12 Ultimate (12_2)
• OpenGL: 4.6
• OpenCL: 3.0
• Vulkan: 1.3
• CUDA: 8.9
• Shader Model: 6.6

Design

• Slot Width: Triple-slot
• TDP: 450 W
• Suggested PSU: 850 W
• Outputs: 1x HDMI 2.1 and 3x DisplayPort 1.4a
• Power Connectors: 1x 16-pin
• Board Number: PG139 SKU 33

GeForce RTX 3090 Specifications

• Release Date: Sep 1st, 2020
• Launch Price: 1,499 USD
• GPU Name: GA102
• Architecture: Ampere
• Process Size: 8 nm
• Transistors: 28,300 million
• Die Size: 628 mm²
• Number of CUDA Cores: 10496
• Base Clock: 1395 MHz
• Boost Clock: 1695 MHz
• Memory Capacity: 24GB GDDR6X
• Memory Clock: 1219 MHz (19.5 Gbps effective)
• Shading Units: 10496
• TMUs: 328
• ROPs: 112
• SM Count: 82
• Tensor Cores: 328
• RT Cores: 82
• L1 Cache: 128 KB (per SM)
• L2 Cache: 6 MB
  

Features

• DirectX: 12 Ultimate (12_2)
• OpenGL: 4.6
• OpenCL: 3.0
• Vulkan: 1.3
• CUDA: 8.9
• Shader Model: 6.6

Design

• Slot Width: Triple-slot
• TDP: 350 W
• Suggested PSU: 750 W
• Outputs: 1x HDMI 2.1 and 3x DisplayPort 1.4a
• Power Connectors: 1x 12-pin
• Board Number: PG132 SKU 10

Right off the bat, it’s quite clear that the RTX 4090 packs a bunch of improvements over its predecessor in every aspect. The new Ada Lovelace architecture uses a much smaller 608 mm² die size with a 4 nm fabrication process, allowing NVIDIA to pack significantly more transistors into the new graphics card.

When talking about the number of CUDA cores, the RTX 4090 received a ~56% increase as it has 16384 CUDA cores compared to the RTX 3090’s 10496 cores, which ultimately means superior performance.

The sheer number of CUDA cores and the base and boost clock speeds have increased by up to 925 MHz. The base and boost clock speeds of the RTX 4090 are 2230 MHz and 2520 MHz, respectively, which is quite an upgrade over the RTX 3090’s 1395 MHz and 1695 MHz.

While both GPUs have the same 24GB GDDR6X memory capacity, the RTX 4090 has a higher memory clock speed of 21 Gbps, whereas the RTX 3090’s memory clocks at 19.5 Gbps.

All in all, no matter how you compare the two graphic cards, the RTX 4090 has better specifications, and this comes at a cost. The GeForce RTX 4090 is rated to consume 100W more power than the RTX 3090 at its full potential, as the GPU has a TDP of 450W.

Meaning that you’ll need at least an 850W PSU to power the monstrous 4090, whereas the 3090 could be powered by a 750W PSU. On top of that, the RTX 4090 also uses a new power connector, which comes with its own pros and cons.

Additionally, higher power consumption also means more heat, and so, to house the giant cooler, the RTX 4090 is colossal in size when put next to its predecessor.

Everything Improved With The Ada Lovelace Microarchitecture

NVIDIA has a simple timeline to release new architectures. Basically, the company releases an all-new architecture for its latest-gen graphics cards every two years. 

For instance, the company released the Turing architecture back in 2018 for the RTX 20-series and GTX 16-series graphics cards. Similarly, it released the Ampere architecture for the RTX 30-series GPUs back in 2020.

Well, in 2022, NVIDIA announced its latest and greatest Ada Lovelace microarchitecture, which has catapulted the performance figures of the 3rd-gen RTX GPUs.

Half The Processor Size

The most considerable improvement that the Ada Lovelace comes with is that it uses TSMC’s N4 (4nm) technology. For reference, the older RTX 30-series GPUs use Samsung’s 8nm RF process technology.

In essence, the Ada Lovelace microarchitecture has managed to reduce the processor size by half, allowing the RTX 40-series GPUs to achieve higher clock speeds. Consequently, the RTX 4090 is also equipped with about 1.7x more transistors. Yes, you read that correctly; the 4090 has 76 billion transistors, whereas the RTX 3090, which was based on the Ampere architecture, only has 28 billion transistors.

More Tensor Cores and DLSS 3.0

For quite some time, NVIDIA has been trying to achieve better performance figures with the help of AI computation. To further fill the gap in their goal, the Ada Lovelace microarchitecture not only uses the latest 4th-gen Tensor cores but also has them more in quantity than its predecessor.

The greater number of Tensor cores, combined with the improved technology, has also allowed NVIDIA to introduce DLSS 3.0. For context, Deep Learning Super Sampling (DLSS) is used to reduce GPU workload in higher resolutions by using AI to guess what the next pixel is going to be.

Well, with DLSS 3.0, the prediction part has moved on from just the pixels to entire frames. Now, without going through the unrendered image data, the 40-series GPUs can guess the next frame, which basically means less load on the GPU and the CPU.

According to NVIDIA, DLSS 3.0 brings up to 4 times better performance than brute-force rendering.

Shader Execution Reordering

The introduction of Shader Execution Reordering (SER) with Ada Lovelace microarchitecture is quite a big deal. To understand how this works, let’s go over how a GPU works in general.

GPUs have a lot of cores when compared to CPUs. That’s because, while a single core of the CPU is powerful enough to handle a task all by itself, making it suitable for multitasking, the many cores of a GPU work together in parallel to perform the same large task.

Now, what is ray tracing? Well, in short, it is a technology that is used to render different lights and shadows in a game. Since light can hit and bounce off any surface in a game, it requires different computations to render.

One interesting thing to note about GPUs is that they are extremely efficient in working on similar data over and over again. For instance, if you ask them to load the same shader 100 times and then ask them to load another shader 50 times, it will load the first shader 100 times quicker than the other shader 50 times.

So, since different lights require different computations to render, GPUs are relatively less efficient in ray tracing, and that’s where Shader Execution Reordering (SER) comes in.

SER basically reorders those different shaders in terms of similarity, making sure that the GPU processes those similar shades together. Consequently, SER significantly improves the efficiency as the multiprocessors of the GPU now work simultaneously on the same data.

RTX 4090 Vs. RTX 3090 Gaming Benchmarks

So far, we have seen that the GeForce RTX 4090 is miles ahead of its predecessor. No matter how you look at it, the GPU simply packs more power.

To further understand how much of an upgrade the RTX 4090 is, we will be analyzing the different gaming benchmarks of the RTX 4090 and the RTX 3090.

Luckily, Game Tests has already done the RTX 4090 vs. RTX 3090 battle, where the two GPUs were tested against each other in 10 games while running at 4K resolution.

The following rig is used in the gaming benchmarks:

• Processor: Core i9-13900K
• Motherboard: ROG MAXIMUS Z690 EXTREME
• RAM: Corsair Vengance 2X16 GB 5200mhz
• GPU: GeForce RTX 4090 and GeForce RTX 3090
• CPU Cooler: ARCTIC Liquid Freezer II 360
• Storage: 2x Corsair Force Series MP600 1TB Gen4 PCIe X4 NVMe M.2 SSD
• Resolution: 4K

Forza Horizon 5

Let’s start off our benchmark analysis with Forza Horizon 5 — a beautiful racing game with gorgeous graphics and realistic physics.

Well, to no one’s surprise, the RTX 4090 completely obliterates the RTX 3090 by running the game with ~90.6% better FPS. On average, the RTX 4090 is running Forza Horizon 5 at 164 FPS. On the other hand, the RTX 3090 is running at 86 FPS.

Even in the 1% low, the RTX 4090 is only dipping to about 132 FPS, which is still ~94% better than the 1% lows of RTX 3080 as it is going down to 68 FPS.

Looking at the thermals, the RTX 4090 is running at 65°C, and the RTX 3090 is running at 63°C. Overall, that’s a difference of about 3.1%, which is basically negligible.

After going through the specs of the two GPUs, we already knew that the RTX 4090 has a 100W higher TDP. So, at 98% usage, the graphics card is consuming 420.8W, which is about 14.8% more than the RTX 3090 as it is consuming 366.3W at 98% usage.

All in all, a ~90.6% performance upgrade at the cost of ~14.8% higher power consumption seems like a fair deal.

Watch Dogs: Legion 

Next up, we have Watch Dogs: Legion, an open-world game that relies heavily on the GPU to render high-end graphics.

The RTX 4090 again leaves the RTX 3090 in the dust as it is running the game at 142 average FPS. Comparatively, the RTX 3090 is running at 79 average FPS, giving its successor a ~79.7% lead.

The performance difference increases to about 81.6% in the 1% lows as the RTX 4090 has 1% lows of 109 FPS and the RTX 3090 of 60 FPS. 

What’s most surprising is that while running Watch Dogs: Legion, the RTX 4090 actually runs cooler than its predecessor. The GPU has a temperature of about 60°C while the RTX 3090 is running at about 66°C. Seeing that the latest and greatest RTX 4090 is ~10% cooler, there is no doubt that the Ada Lovelace microarchitecture is quite thermally efficient.

The power consumption of the two GPUs is exactly what you expect it to be. The RTX 4090 is consuming 429.6W, and the RTX 3090 is consuming ~19.2% less power at 360.4W. This is completely normal, as the RTX 4090 has a higher TDP.

Shadow Of The Tomb Raider

Even though Shadow of the Tomb Raider was released over 4 years ago, the game still has gorgeous graphics to this date. Which is why it is a good place to benchmark the two GPUs.

Keeping up with the above two games, the RTX 4090 does not go easy on its predecessor, as the GPU is running the game at an average of 160 FPS. To put things into perspective, the RTX 3090 is performing about 88% inferior as it is running at 85 average FPS.

Similarly, there is a ~93.8% performance difference in the 1% lows where the RTX 4090 is running at 157 FPS and the RTX 3090 at 81 FPS.

For a graphics card with such performance figures, the RTX 4090 is fairly tame in thermal efficiency. Even in Shadow of the Tomb Raider, the GPU is just ~1.5% hotter at 64°C than the RTX 3090 at 63°C.

With about 96% usage, the RTX 4090 is below its TDP (450W) at 415.5W. On the other hand, the RTX 3090 is above its TDP (350W) at 369.9W with 97% usage. Overall, the RTX 4090 is consuming ~12.3% more power.

Red Dead Redemption 2

There is no doubt that Red Dead Redemption 2 is one of the most beautiful games out there. Its attention to detail, physics, and high-end graphics stole the show right after its release.

Red Dead Redemption 2 is the first game in our RTX 4090 vs. RTX 3090 guide so far, where the difference in performance has taken a hit. But, even then, the RTX 4090 still outperforms its predecessor with a ~65.7% lead. On average, the RTX 4090 is running the game at 126 FPS, and the RTX 3090 is running at 76 FPS.

The 1% lows also follow the same pattern; the framerates of the RTX 4090 are dropping to about 100 FPS and the RTX 3090 to about 61 FPS, giving the RTX 4090 a ~63.9% edge.

As usual, there is not much of a temperature difference between the two GPUs. The RTX 4090 is at 65°, and the RTX 3090 is a smidge cooler at 63°C. Overall, that’s a difference of about 3.1%, which can be neglected.

The power consumption figures are no different either, the RTX 4090 is consuming 420.9W, and the RTX 3090 is consuming 366.3W. While the former is consuming about 14.8% more power, it is still below its TDP.

Horizon Zero Dawn

Coming up next is Horizon Zero Dawn, yet another game with beautiful graphics that make it a worthy game to test our two GPUs on.

Here the RTX 4090 performs exactly how you would expect it, with ~71% better average FPS. On average, the RTX 4090 is running the game at 158 FPS. In comparison, the RTX 3090 cannot even surpass the 100 FPS mark as it is running at 92 FPS.

The 1% lows tell no different story either, the RTX 4090 is going as low as only 122 FPS. On the other hand, the RTX 3090 stayed behind by about 62.6% as it is going down to 75 FPS.

Horizon Zero Dawn is the second game so far where the RTX 4090 is more thermally efficient than its predecessor. Here, the GPU is running at 62°C, whereas the RTX 3090 is running at 65°C. Even though the ~4.8% difference is not a big deal, it speaks volumes of how thermally efficient the RTX 4090 is.

The power-hungry RTX 4090 is consuming 427.7W of power when running Horizon Zero Dawn, which is about 18.2% more than the RTX 3090 as it is consuming 361.7W.

Hitman 3

When it comes to some stealth action, nothing beats Hitman 3.

Similar to what we have analyzed so far, the RTX 4090 outperforms its predecessor by an enormous margin. The RTX 4090 is running Hitman 3 at 184 average FPS while the RTX 3090 is running it at 100 average FPS. That’s a difference of a solid 84%, giving the latest RTX 4090 yet another huge win.

While the 84% average FPS margin is a big deal, the difference in performance only increases in the 1% lows. That’s because the RTX 4090’s framerates are dipping to only 150 FPS. On the other hand, the RTX 3090’s framerates are going down by about 87.5% to 80 FPS.

No doubt that the performance difference between the two GPUs is anything but negligible. However, what’s definitely negligible is the difference in thermals as it is only 1°C where the RTX 4090 is running at 63°C, and the RTX 3090 is running at 64°C.

As one would expect, the RTX 4090 is consuming about 15.9% more power at 429.1W, and the RTX 3090 is taking in around 360.8W.

Far Cry 6

When it comes to Far Cry 6, the RTX 4090 yet again slays its predecessor by a considerable margin.

On average, the RTX 4090 is running the game at 165 average FPS. On the other hand, the RTX 3090 hardly manages to make it to the 100s with 103 FPS. While the difference in performance is over 60%, it is the lowest margin we have seen so far.

The 1% lows of the Ada Lovelace GPU dropped to about 124 FPS, and the previous-gen Ampere GPU went down to as low as 81 FPS. Overall, that’s a ~53% difference in 1% lows which is another lowest difference we have analyzed.

The thermal efficiency of the RTX 4090 and the RTX 3090 while running Far Cry 6 is quite similar to what it was when running Hitman 3. The RTX 4090 is running a degree cooler at 63°C as the RTX 3090 is at 64°C.

As we saw in the detailed specifications, the suggested PSU for the RTX 4090 is 850W and 750W for the RTX 3090. Well, the RTX 4090 is sucking up about 421.3W of power, which is ~15.1% more than the RTX 3090, as it is consuming 365.9W. 

Dying Light 2

The RTX 4090 continues to stay ahead of the RTX 3090 with huge performance gains in Dying Light 2.

Throughout the entire benchmark, the RTX 4090, on average, ran the game at 4k with about 104 FPS. In comparison, its predecessor, the RTX 3090, could only make it to 60 FPS. Overall, there’s a difference of about 73.3% in the FPS figures, giving the RTX 4090 another huge win.

In the 1% lows, the RTX 3090 dropped below 60 FPS to about 51 FPS. On the other hand, the RTX 4090 performs considerably well as it took a dip to 87 FPS.

When looking at the temperature figures, we see no difference from the above two games as the RTX 3090, yet again, is running a degree hotter at 64°C. The RTX 4090, while running the game at a significantly better average FPS, is running at 63°C.

The biggest power consumption difference we have analyzed so far was ~19.2% in Watch Dogs: Legion. Well, now that title has been given to Dying Light since the difference in power consumption has bumped up to ~21.9%. That’s because while the RTX 3090 is consuming about 356.7W, its successor, the RTX 4090, is consuming a whopping 435.1W of power.

Cyberpunk 2077

When you hear of Cyberpunk 2077, its disastrous release will come to your mind. But, with constant updates, the game has improved a lot.

While the RTX 3090 cannot even achieve the 60 average FPS mark, the RTX 4090 is comfortably running Cyberpunk 2077 on 4K with 84 average FPS. The RTX 3090 is about 71.4% behind with 49 average FPS.

Similarly, the 1% low of the RTX 4090 is showing 71 FPS, which is still ~65% more than the RTX 3090 as it is at 43 FPS.

Even though the temperature difference remains the same in Cyberpunk 2077, both GPUs have switched them up, as now the RTX 4090 is running a degree hotter at 64°C while the RTX 3090 is at 63°C. Overall, there is no doubt that the RTX 4090 is just as thermally efficient as its predecessor, with little to no difference in temperature.

It looks like the RTX 4090 is firing up at all cylinders to achieve such high performance. Previously, we noted that in Dying Light 2, the difference in power consumption was the highest. But, the gap increases further in Cyberpunk 2077 as the RTX 4090 is consuming 440.2W. Comparatively, the RTX 3090 is more power efficient as it is taking in ~24.5% less power at 353.4W.

Assassin’s Creed Valhalla

To bring our RTX 4090 vs. RTX 3090 benchmarks analysis to an end, we will be reviewing Assassin’s Creed Valhalla.

Throughout the entire runtime, the RTX 4090 is running the game at an average of 122 FPS. In contrast, the RTX 3090 is about 46.9% behind at 83 average FPS. While it is another win for the Ada Lovelace GPU, the RTX 3090 performs commendably well as it has closed the gap by a lot.

The margin further decreases in the 1% low as the RTX 4090 seems to be dropping to 79 FPS, whereas the RTX 3090 goes down to 62 FPS. Overall, the RTX 4090 has about 27.4% better than 1% lows.

Similar to what we’ve seen in the other 9 games, there is not much of a thermal difference between the two GPUs. While the RTX 4090 is running ~3.1% hotter at 65°C and the RTX 3090 at 63°C, the difference is negligible.

Coincidentally, the power consumption figures are exactly the same as they were in Cyberpunk 2077. With the RTX 3090 consuming 353.4W, the RTX 4090 is sucking in about 24.5% more power at 440.2W.

Overall Gaming Performance

Now that we have gone through 10 games and analyzed their benchmark performance in 4K resolution, we can finally say that the GeForce RTX 4090 is a card in its own league.

The GPU started off with a gigantic ~90.6% lead, which is a lot seeing that it was put against the beast RTX 3090. The performance difference was the least in Assassin’s Creed Valhalla, and even then, it was a whopping ~46.9%.

If we take an average of all the gaming benchmarks we analyzed, the RTX 4090 ran them at 4K with well over 73% more FPS. Such tremendous performance jumps in all games prove that NVIDIA has made quite a lot of remarkable changes to the core structure of their GPUs with the Ada Lovelace microarchitecture.

It looks like those 1.7x more transistors and 5888 more CUDA cores with a smaller fabrication process have paid off in terms of sheer performance.

Power And Thermal Efficiency

After going through the RTX 4090’s specifications, we know that the card will be consuming more power. Even its suggested PSU is rated for 850W, which is 100W more than the RTX 3090.

So, without any surprises, the RTX 4090 is definitely a power-hungry beast. On average, the card consumed about 18% more power than the RTX 3090. The difference in power consumption was the least in Shadow of the Tomb Raider, with ~12.3%, and it went all the way up to ~24.5% in Cyberpunk 2077 and Assassin’s Creed Valhalla.

Such a high power consumption points to one thing; you should only power the RTX 4090 with its suggested PSU or more. Anything below and things might get risky. While power efficiency is not the RTX 4090’s best suit, it is the only aspect where the RTX 3090 came out on top. 

When talking about thermal efficiency, it is quite clear why the RTX 4090 is colossal in size. Its huge cooler works wonders in bringing down the temperatures. In fact, in all the 8 games we tested, the RTX 4090 was actually about 0.7% cooler than the RTX 3090.

Yes, the card produced ~73 more FPS while staying ~0.7% cooler. To put things into perspective, the RTX 4090 was running a whopping 10% cooler in Watch Dogs: Legion. On top of that, the hottest it performed in comparison with the RTX 3090 was in 3 of the 10 games, where the temperature difference reached ~3.1%.

Price And Availability

Unfortunately for the RTX 3090, the GPU was released during the COVID-19 pandemic. The chip shortage, combined with lots of lockdowns, drove its price from $1499 to north of $2000.

However, now that things have settled down, you can get an RTX 3090 from Amazon for about $999. On the other hand, the RTX 4090 has an MSRP of $1599, and the GPU is still a bit difficult to find on shelves. In fact, it’s highly unlikely that you can get an RTX 4090 for its MSRP; the GPU is currently available on Amazon for upwards of $2299.

While the NVIDIA 40-series GPU is certainly more expensive, that’s not it. If you plan to purchase the RTX 4090 over the RTX 3090, you’ll likely have to spend more on a better PSU, which won’t come cheap.

Overall, there is no doubt that RTX 4090 performs significantly better than any other gaming GPU to date, but it will definitely hit your wallet.

RTX 4090 Vs. RTX 3090: Which One Is Better?

Coming back to the final question of our RTX 4090 vs. RTX 3090 guide, how much better is the GeForce RTX 4090 than the GeForce RTX 3090? Well, after doing a complete RTX 4090 vs. RTX 3090 battle, we can say that the RTX 4090 is miles ahead of its predecessor.

We put the two graphics cards against each other in all categories. We went through their specifications and gaming benchmarks and saw how much better the new microarchitecture is. And in a nutshell, the RTX 4090 is superior to the RTX 3090 in most categories, except, of course, the power consumption and price.

So, if you’re planning to build a high-end rig with the latest and greatest components, then you must pair the i9-13900K with the RTX 4090; the combo is a no-brainer. However, if you’re on a budget and you can’t spend the extra $1000 on the RTX 4090, then you can go with the RTX 3090.

Even then, we would recommend you hold on to your horses and wait for the rest of the 40-series GPU to hit the market. Their arrival will not only decrease the RTX 3090’s prices, but you might get one of the latest Ada Lovelace GPUs in your budget.