GeForce RTX 4070 vs. Radeon RX 7800 XT: 45 Game Benchmark

It's all about the pricing. At $100 less the 7800XT is a no brainer. At near price parity I lean towards Nvidia for various well worn and much argued issues about documented ray tracing superiority, DLSS superiority, drivers, thermals and so forth.

Overclocking is similar for budget models, the Radeon looks better if you buy a higher end model but nothing too spectacular. Nvidia's better thermals and power mean you can probably get away with a lower end 4070. Feels like a common theme playing out.

16GB v 12GB of VRAM is much less of an issue than the previous generation 16GB v 8GB because we look to be over that major bump new console hardware always creates. I would say 12GB will be sufficient (albeit only just) for the next few years at the common resolution and settings you would employ with this GPU tier.

In many regions availability of the Radeon is poor but I hope it'll improve given a few months. If you need a card now you buy what best models are available.
 
Radeon is cheaper, faster, even ray tracing performance is neck and neck, the worst case being Cyberpunk, rest is pretty good, FSR3 on the way. I really don't know which one to pick...
 
When titles run better on AMD hardware, they can really run a lot better. Hogwarts, Vahalla, and now Starfield are scandalous in their disparity between Nvidia and AMD GPUs that typically average only a few percentage points in performance different. Each of those titles were AMD sponsored. So, what gives? I mean, if I were a PC news journalist I would at least be asking the question, how does this happen? Perhaps it is just AMD optimizations, then again, perhaps it is a lack of Nvidia optimizations due to console first and then ported to PC. Either way, it seems more titles are coming along that just run better on AMD. I don't think anything nefarious is going on, but logic would suggest there is at least something amiss.
 
Perhaps it is just AMD optimizations, then again, perhaps it is a lack of Nvidia optimizations due to console first and then ported to PC.
Hogwarts, Valhalla, and Starfield all run on three very different engines but the first two of them have one thing in common -- they're designed to work on as broad a range of platforms as possible. I don't just mean consoles vs PCs, I also mean PCs vs PCs. Unreal Engine and Ubisoft's Anvil are very similar in that they're not specifically focused on one architecture, whereas Creation Engine 2 almost certainly is these days.

Developers optimizing their engines for various architectures have to do this at the fundamental stage of writing the engine code -- e.g. taking into account differences in thread batch management isn't something that can just be done by tweaking a few different lines of code, once the whole thing is done. Optimizations by hardware vendors, though, typically come in the form of adjusting the compiler to account for what the game is sending to the drivers and the use of shader injection to better suit the specific nature of a given piece of hardware.

Anyway, with regard to the two cards in this article, the RX 7800 XT has notably higher fill rates (texel and pixel), more last-level cache, and more memory bandwidth than the RTX 4070. Even if one ignores dual issuing shader rates, for both GPUs, the 7800 XT is still ahead. In short, AMD's product should be better than the 4070, and depending on what aspect of the GPU is limiting the performance, it has the potential to be consistently 20 to 30% faster.
 
No real shocks here. 7800XT is the faster card in pure rasterization, and at $100 less than the 4070. Factor in the free copy of Starfield premium and save an additional $99 as well as better performance in this specific game.
Also, something not mentioned in this article is the way the 7800XT scales when overclocked. I am able to average 2800mhz when undervolted to 980mV and get performance close to my stock 6950XT (without the heat and noise)... The instability in certain games is likely just immature drivers, I expect the drivers to get better with time.

I purchased the Hellhound version which was the same price as the stock AMD card at $499. This is the first card in years that I have purchased that does not require a waterblock to run quietly under load....
 
So,
After checking lots of reviews. 4070 is the clear winner.

4070 12GB vs 7800 XT 16 GB
Price- Amd in US but Equal in WorldWide
Raster - 5% Amd
RT + Raster - 5% Nv
DLSS - Nvidia
Watt- 60 Watts Less Nv
Heat- Nv
Frame Gen- Nv
Driver- Nv
Features - Nv
Streaming - Nv
Vr - Nv
ReSale Value - Nv
NonGaming Work- Nv
Vram- 4GB More Amd
Winner Nv

What I and many other would like to see in say 2026 if 12GB vram Will screw 4070. Hopefully not.
 
Also, amazing review steve.
Great to see RT+Raster average aka Mixed. 😇
Now, many of us need not to look in other sites.
Hub/techspot is enough.
 
I think the only good news with these new gens GPU is the price. There is no big improvement from gen to gen so I'll just wait, maybe for the 8800xt as my final GPU xD
 
@Steve, check it. RT numbers are fully contradictory vs other tech oulets. There might have been a mistake.
hat Control RT result is 100% wrong.

In no way the 7800xt and 4070 are the same performance.

For Eurogamer at 1440p https://www.eurogamer.net/digitalfoundry-2023-amd-rx-7800-xt-7700-xt-review?page=2

7800xt - 49.64 fps vs 4070 - 57.18

From Kitguru

https://www.kitguru.net/components/graphic-cards/dominic-moass/amd-rx-7800-xt-review/all/1/

7800xt - 44.3 fps vs 4070 - 52.8

From Tom's Hardware

https://www.tomshardware.com/reviews/amd-radeon-rx-7800-xt-review/3

7800xt - 49 fps vs 4070 - 59.4 fps



No way these two perform the same in Control RT. The 4070 should be about 20% faster



Edit - F123 is also. From AMD's own numbers the 4070 should be 16% ahead (the 7800xt is 14% slower)

Calisto Protocol is another one where the 4070 is 10% faster (7800xt is 9% slower)

https://wccftech.com/amd-radeon-rx-...ster-than-4070-7700-xt-9-faster-than-4060-ti/



I have serious questions to what RT settings were used to get this card to a point that even AMD couldnt
 
As @SalmanIsHere I have similar problem with relative high RT performance only 2% worse in 1440p than 4070
I can add another example -14% techpowerup 1440p:
relative-performance-rt-2560-1440.png
 
Last edited:
@Steve, check it. RT numbers are fully contradictory vs other tech oulets. There might have been a mistake.
hat Control RT result is 100% wrong.

In no way the 7800xt and 4070 are the same performance.

For Eurogamer at 1440p https://www.eurogamer.net/digitalfoundry-2023-amd-rx-7800-xt-7700-xt-review?page=2

7800xt - 49.64 fps vs 4070 - 57.18

From Kitguru

https://www.kitguru.net/components/graphic-cards/dominic-moass/amd-rx-7800-xt-review/all/1/

7800xt - 44.3 fps vs 4070 - 52.8

From Tom's Hardware

https://www.tomshardware.com/reviews/amd-radeon-rx-7800-xt-review/3

7800xt - 49 fps vs 4070 - 59.4 fps



No way these two perform the same in Control RT. The 4070 should be about 20% faster



Edit - F123 is also. From AMD's own numbers the 4070 should be 16% ahead (the 7800xt is 14% slower)

Calisto Protocol is another one where the 4070 is 10% faster (7800xt is 9% slower)

https://wccftech.com/amd-radeon-rx-...ster-than-4070-7700-xt-9-faster-than-4060-ti/



I have serious questions to what RT settings were used to get this card to a point that even AMD couldnt
I can confirm the Control RT results are accurate, having just retested. I'm benchmarking at the start of the game, both GPUs deliver similar results. The F1 23 data is also accurate, I'm using the built-in benchmark there so the data should be easy to validate.
 
I have an idea for benchmark with RT in some games and different settings and show are any relative difference in fps for this settings between rx7000 and rtx4000. Maybe some rt setting on/off causes significant drop on AMD hardware.

I know that benchmarking GPUs today are very hard because we have Rt/DLSS/FSR and sponsored titles. And one game and in game settings or location can be 30% better for AMD or NVidia cards.
 
Link to all results : Techpowerup

For example F1 23.
rt-f1-2023-2560-1440.png


I think that differences are caused by different settings and choosed games.
But I cannot find summary in reputable portals with 7800xt as good in Rt as in Your test.
I believe that the results are correct, but I don't know why they are so on positive side for 7800xt in compare to others

You can also ask Steve from Gamers Nexus:
1440p f1 - gamers nexus
What's really odd is that my frame rates for both GPUs are much higher, using what looks to be the same settings. I'm wondering if this is an issue with the 13900K. We've seen some odd scaling behavior in some games comparing the 13900K and 7800X3D. I'm going to re-test these two GPUs using the 13900K in Control and F1 23 tomorrow to see if I can replicate these results.
 
Yeah, HUB's numbers are super odd.

pcgh have rtx4070 beat 7800xt in a 10 game rt roundup by +22%

purepc show +27%, but those are all very demanding locations in very heavy rt games, with multiple effects enabled, all pc-level rt, not console-level rt.

same as THG, also showing about 27% avg difference in 6 games (though their MetroEE run included hairworks), and 4070 winning in Control RT by 20%

Guru3D shows 4070 winning Callisto RT

I read multiple reviews and only on HUB I've seen 7800xt comparable to 4070 on average in RT, and not losing by 15% or more. Sure, there are cases where it gets close or wins, but take +10 games and no one showed it close.
the F1 23 result I've seen from THG shows 4070 win by 9% on avg. fps and 25% on min fps
 
Last edited:
Yeah, HUB's numbers are super odd.

pcgh have rtx4070 beat 7800xt in a 10 game rt roundup by +22%

purepc show +27%, but those are all very demanding locations in very heavy rt games, with multiple effects enabled, all pc-level rt, not console-level rt.

same as THG, also showing about 27% avg difference in 6 games (though their MetroEE run included hairworks), and 4070 winning in Control RT by 20%

Guru3D shows 4070 winning Callisto RT

I read multiple reviews and only on HUB I've seen 7800xt comparable to 4070 on average in RT, and not losing by 15% or more. Sure, there are cases where it gets close or wins, but take +10 games and no one showed it close.
the F1 23 result I've seen from THG shows 4070 win by 9% on avg. fps and 25% on min fps
Doesn't Guru3D have them neck and neck in the 3 games they tested with RT?
Obviously it comes down to the games tested and the number of games.
 
Doesn't Guru3D have them neck and neck in the 3 games they tested with RT?
Obviously it comes down to the games tested and the number of games.
I agree 3 games is quite anecdotal evidence, therefore I only used callisto result as it's different form Yours. And none of them are actually that rt-heavy looking at fps they're getting at native.
Plus I think all 3 are amd sponsored anyway, kinda one-sided, but that's what g3d community is nowadays frankly. Hilbert used to include heavy rt titles, but not anymore as literally all he has to hear now is complaining from amd owners in comments when their cards lose.
I think the pcgh result is the most objective one, 10 games, 4 resolutions, ranging from light console level rt to heavy pc rt.
If you wanna compare pc ultra-level rt only, look at purepc, but I think that's too punishing for a mid-range gpu . No one runs Ultra RT on 4070 in every game. I personally pick a maximum of two rt effects that I like most on 3080, the rest is not rt. And the rt ones I usually put at med/high, rarely (or maybe never) on ultra.
 
Last edited:
I agree 3 games is quite anecdotal evidence, therefore I only used callisto result as it's different form Yours. And none of them are actually that rt-heavy looking at fps they're getting at native.
Plus I think all 3 are amd sponsored anyway, kinda one-sided.
I think the pcgh result is the most objective one, 10 games, 4 resolutions, ranging from light console level rt to heavy pc rt.
Sure, there's lots of bad titles there for AMD like The Riftbreaker, Dying Light 2, Metro Exodus and CP2077. In the cross over games like Dying Light 2 and CP2077 we also found the 7800 XT to be very weak.
 
Sure, there's lots of bad titles there for AMD like The Riftbreaker, Dying Light 2, Metro Exodus and CP2077. In the cross over games like Dying Light 2 and CP2077 we also found the 7800 XT to be very weak.
It must be hairworks or some other setting in Metro, I found 6800 never had problems with running +60fps rtgi on at 1440p native.
 
Back