Bang for buck? AMD Radeon RX 7800 XT vs Nvidia GeForce RTX 4070 in 2024

Sayan Sen contributed to this review, and also provided the benchmark graphics.

AMD reached out to us asking if we were interested in testing the Radeon RX 7800 XT. They ended up shipping a Gigabyte Radeon RX 7800 XT GAMING OC 16G to us, which we will be pitching against the MSI GeForce RTX 4070 VENTUS 2X 12G seeing as they are both in a similar price class.

First up the specs of the 7800 XT which was given to us by AMD:

and our 4070, which was privately purchased for our test system and this review:

The list price on Amazon is given as $599, however with the release of the 4070 SUPER, its predecessor dropped $50 in price, which means it now sits between the $439.99 ASUS TUF 4060Ti OC Edition and the newer $599.99 MSI 4070 SUPER Ventus 2x.

Test system

Our test system consists of the following

• Cooler Master MasterBox NR200P MAX
• ASRock Z790 PG-ITX/TB4
• Intel Core i7-14700K with Thermal Grizzly Carbonaut Pad
• T-FORCE Delta RGB DDR5 (2x16GB) 7600MT/s CL36 (XMP Profile)
• 2TB Kingston Fury Renegade SSD
• Windows 11 23H2 (build 22631.3880)

Drivers used for the 7800 XT were Adrenaline v24.6.1, and for the 4070 Geforce v560.70 was used.

Benchmarks

For our benchmarks, hwinfo.com provided a commercial license of HWiNFO, and UL Solutions provided us with Professional (commercial use) licenses for 3DMark, VRMark, and Procyon.

We start with synthetics and take a look at the productivity outputs of the cards. First up we have Geekbench OpenCL compute performance where, a bit surprisingly perhaps, the 4070 comes out quite a bit ahead of the 7800 XT. There was a time when AMD cards would generally be better than Nvidia at similar price levels but such is not the case any more it seems. The latter has clearly made strides in OpenCL even though CUDA is its main API.

Up next we move to AI performance assessment using UL Procyon's benchmarks. We test inferencing by utilizing DirectML on both the cards and they appear neck and neck.

For AI image generation, we again use DirectML for AMD but for Nvidia we utilize TensorRT. The RTX 4070 is 70% faster than the 7800 XT in the Stable Diffusion XL test.

We conclude the productivity testing with Cinebench 2024 which uses Redshift rendering engine. The Nvidia card, again, is 70% faster.

Therefore it can be said that productivity-wise, you will be better off with Nvidia.

Continuing with synthetics, we now start our game testing.

On 3DMark Fire Strike Ultra, which measures DirectX 11 performance, the 7800 XT is almost 30% faster.

On Time Spy Extreme, which is DirectX 12-based, the 7800 XT wins again though with a bit lower margin of 16%.

Steel Nomad is also based on DirectX 12 but is more intense than Time Spy Extreme and this time the 7800 XT is about 10% ahead.

Therefore the trend is clear, the 7800 XT does better and better the older we go. This could be one of two things: either the 7800 XT is better suited for older technologies that games employ or the driver overhead of the AMD card is lower than the Nvidia. Either way, it's a win for the RX 7800 XT in rasterization-based synthetic benchmarks.

We also test ray tracing performance using 3Dmark's Port Royal and Speed Way.

In the less intense Port Royal test, the 7800 XT does well against the 4070. This is representative of games where ray-traced visual effects are modest.

Speed Way is more intense than Port Royal and is more representative of most modern titles with ray tracing, and here the Nvidia GPU is just over 11% faster.

With that, we wrap up our synthetic testing and move to real games. The RX 7800 XT is advertised as a 1440p GPU and thus 3440 x 1440 was our resolution of choice. If you are wondering why we went even higher than 2560 x 1440, we wanted to ensure the CPU was not in any way holding the GPU back. This is essentially the opposite of CPU game testing, something Sayan covered in great detail in his editorial.

AMD requested us to have a look at HYPR-RX which now combines with the power of AMD Fluid Motion Frames (AFMF). While HYPR-RX works on all games, we specifically decided to test it in two titles from our test suite.

We picked titles that do not have either FSR and/or DLSS. That is because DLSS and FSR typically show differences in the quality of upscaling they offer. Generally, FSR lags slightly compared to DLSS so it would not be an apples-to-apples comparison. Thus we stuck to native rendering testing only in the rest of the games except Deus Ex Mankind Divided and Assassin's Creed Odyssey.

We also did not play around with the HYPR-RX settings since AMD advertises it as a one-click performance improvement feature and thus we wanted to see how well HYPR-RX did just by enabling it.

We start with Deus Ex Mankind Divided powered by the Dawn Engine. The AMD 7800 XT does better here in terms of averages but the minimum FPS was lower.

With HYPR-RX (AFMF) enabled, the average frame rate for the 7800 XT did not go up but the minimums improved tremendously, by ~20%.

The Dawn Engine makes an appearance again later.

Up next, in Assassin's Creed Odyssey running on AnvilNext 2.0, we saw that the Radeon RX 7800 XT was quite a bit ahead of the RTX counterpart, and again, with HYPR-RX (AFMF) enabled, we saw better minimum FPS. Oddly though, we lost performance leading to the lower average FPS, as shown below.

Therefore it may be said that HYPR-RX (AFMF), at least in some of these older titles can be a hit or miss.

Next up, we have Shadow of the Tomb Raider based on the Foundation Engine, and in this game, it was a wash. Note that the game was tested on DirectX 11.

Hitman 3 was next and here, the RX 7800 XT showcased its strongest win as it was almost 29% faster.

In Far Cry 6, the 7800 XT was the faster card. But you are unlikely to notice the difference without an FPS counter.

Marvel's Guardians of the Galaxy was the only game in our rasterization test suite in which the 4070 won. However, just like in Far Cry 6, you are unlikely to notice much of a difference without an FPS counter.

Like Deus Ex Mankind Divided, this one too is powered by the Dawn Engine.

For ray tracing, we ran the standalone Unreal Engine 4-based Bright Memory: Infinite free benchmark and as expected, based on our synthetic ray tracing tests (above), the 4070 wins by an 8.33% margin.

Since this is a 2019 title, some of the newer games will likely favor the RTX 4070 even more.

We now highlight the VRAM usages of the two cards where the games used more than 4 GB of dedicated GPU memory. Keep in mind, these are the actual used VRAM amounts not the allocated amounts.

Typically, it can be seen that the titles used more VRAM on the 7800 XT indicating the better compression capability of the RTX 4070. However, that is not a flex in this case as Nvidia certainly needed that.

As you can see in the next chart indicating the percentage of VRAM usage, it can be seen that in two out of five instances, more than two-thirds (66.67%) of the 4070's memory was being consumed by the games.

Hence, in terms of VRAM, the 7800 XT has significantly more headroom. In more modern titles, you may find the 4070 struggling as video memory requirements keep going up.

While VRAM is the strong point for the AMD RX 7800 XT, power consumption is not. The Radeon GPU takes up more power. You will ideally need a 650-watt PSU to be comfortable with this, and it's also AMD's recommendation. The 4070 can be run on a 500-watt unit too (depending on the TDP of the CPU), although Nvidia also recommends a 650W power-supply.

Adding to the power consumption data, we also add performance per watt figures of the two cards to highlight their power efficiencies. Expectedly, the Nvidia GPU wins in this category and with the biggest deltas in Guardians of the Galaxy (M GOTG) and Bright Memory Infinite (BM: I) at 59% and 56% respectively.

We also ran HwINFO64 and observed the following Hot Spot temps (Max) for each card:

• RTX 4070: 91 C
• RX 7800 XT: 92 C

Conversely, the AMD Radeon RX 7800 XT wins in the value-for-money (VFM) category as can be seen in the performance per dollar chart below. If you do not care about power efficiency and are after value, the XT looks to be the superior choice. The biggest gap in VFM can be seen in Hitman 3 (HM 3) where the AMD GPU is 43% ahead. Even in Bright Memory Infinite (BM: I), Team Red's GPU manages a marginal win indicating that it provides similar value for moderate levels of ray-traced games too.

Finally, we tested virtual reality performance with VRMark in which we can see that the 7800 XT was better at both light (Orange Room) and heavy (Blue Room) VR tests.

Conclusion

It would seem that the winner here is the consumer, as you are spoiled for choice in this category. If you want bang for buck, the obvious choice is the 7800 XT with its 16GB of GDDR6 VRAM, better raster performance, and DisplayPort 2.1. On the other hand, the slightly more expensive 4070 is better in terms of ray tracing and productivity, and is also more energy efficient.

I think the 7800 XT wins, in terms of value, and what you're getting. But as I said before, there really are no losers in this category. If you prefer Nvidia, stick with Nvidia, but if you like AMD, then this is a great card for its price point. (If you go Nvidia though, make sure to avoid the 4060 Ti 16 GB which also is priced similar to the 7800 XT.)

As an Amazon Associate we earn from qualifying purchases.