Time for another of our mega-game benchmarks covering 45 titles comparing the Radeon RX 6700 XT head-to-head against the GeForce RTX 3070, the nearest competitor in terms of pricing, at least based on the mythical MSRP.
When we reviewed the Radeon RX 6700 XT on day one, we found it to be about 4% slower than the RTX 3070 in a 14-game sample. We collected that data using our older Ryzen 9 3950X test system. But now we have an even more robust set of 45 games running on top of a Ryzen 9 5950X along with the latest display drivers which have brought some nice gains for the RTX 3070 in titles such as Godfall, Assassin’s Creed Valhalla, and Dirt 5.
Unfortunately, while performance in some games has improved, availability has not, with both GPUs nearly always out of stock. With stock clearly out of everyone’s control, this comparison serves as a guide for those researching what they might purchase once supply is able to meet or exceed demand.
There’s always a chance, if you’re patient enough, you’ll eventually get your hands on one of these GPUs. At some point, the situation will improve and when it does, this comparison review will help answer the question, which of the two GPUs should you buy? So, with that, let’s go over the test system specs and jump into the benchmarks.
Our Ryzen 9 5950X test system was equipped with 32GB of DDR4-3200 CL14 dual-rank memory. Representing both the GeForce and Radeon GPUs are Gigabyte Gaming OC graphics cards, but do note that in both cases we’ve clocked them to match the RTX 3070 and RX 6700 XT reference spec with no overclocking. We’re using GeForce Game Ready Driver 465.89 and Adrenalin 2020 Edition 21.3.2 drivers. Smart Access Memory was not enabled and this is always the case unless specified otherwise.
First, we’ll individually take a look at around half a dozen of the more interesting games in terms of test results and then jump into breakdown graphs to summarize all the benchmark data.
Starting with Apex Legends we find that at 1080p, the RTX 3070 is up to 22% faster which is quite a significant margin. It’s worth noting that the 6700 XT did average 162 fps using high quality settings, so for competitive quality settings frame rates are going to be more than sufficient with either product, but if you’re after maximum performance in Apex, you’ll be better off with the GeForce.
The RTX 3070 was also 19% faster at 1440p and for those using the high-quality settings, the difference between 113 and 135 fps can be noticed with a high refresh rate monitor.
Performance was more evenly matched in CoD Warzone as the RTX 3070 was ~3% faster at 1080p and 8% faster at 1440p. The margin does open up to 12% in favor of the GeForce at 4K, though I suspect few gamers will be targeting 4K with these GPUs.
Fortnite has always heavily favored Nvidia, though AMD has made some inroads here. Still, when compared to the RTX 3070, the 6700 XT was quite a bit slower, trailing by a 14% margin at 1080p, 17% at 1440p, and 20% at 4K.
Assuming you have a relatively modern CPU that isn’t being heavily utilized when playing Fortnite, the RTX 3070 is going to deliver considerably better performance than the 6700 XT in this title (DLSS/RT conversation aside).
Performance in Hitman 2 is very similar on either GPU. The RTX 3070 was once again faster at all three tested resolutions, though we’re looking at a relatively small 6% performance advantage.
Microsoft Flight Simulator 2020 is heavily CPU bound at 1080p and as a result, both the 6700 XT and RTX 3070 are seen delivering roughly the same performance.
However, this changes at 1440p, the 1% low result is still CPU limited, but the average frame rate is now much higher using the RTX 3070. A 23% performance uplift is significant. That margin is extended at 4K to 26%, though neither GPU is capable of delivering playable performance at 4K.
Borderlands 3 is the first game that we come across where the 6700 XT is faster than the RTX 3070 by a reasonable margin. At 1080p the Radeon GPU was 12% faster, then 13% faster at 1440p before dropping back to a 7% lead at 4K. A win here for AMD and the performance uplift at 1440p will be noticed for those playing Borderlands 3.
Frame rates are also competitive in Cyberpunk 2077 (before enabling DLSS). The RTX 3070 and 6700 XT deliver virtually identical performance at 1080p and 1440p, with the GeForce GPU running away at 4K with a 17% advantage, though as was the case in Flight Simulator 2020, performance at 4K was not playable with these settings.
Control is a horrible scenario for AMD and that’s largely due to it being an Nvidia sponsored game that was used to showcase RTX features. AMD hasn’t been able to optimize for this one and as a result the RTX 3070 was clearly faster at 1080p with a massive 28% margin, then 31% faster at 1440p and 38% faster at 4K.
We find performance in Red Dead Redemption 2 to be very competitive with nothing separating these two GPUs at 1080p and 1440p. As usual, the RTX 3070 enjoys a performance advantage at 4K, pulling away by a 9% margin to average 51 fps.
The Outer Worlds is based on Unreal Engine 4 and as is often the case with games using this highly popular game engine, the results swing heavily in Nvidia’s favor. Here the RTX 3070 was 17% faster at 1080p, 29% faster at 1440p and 28% faster at 4K. Like Control this is another wipeout for AMD.
Although performance for the RTX 3070 has improved in Assassin’s Creed Valhalla since our last test, the 6700 XT still enjoys a performance advantage here. We’re looking at an 18% win at 1080p for AMD and 10% at 1440p. At 4K, the RTX 3070 does get a slight upper hand though.
Moving on to F1 2020, we find fairly competitive performance across the three tested resolutions. The RTX 3070 was 4% faster at 1080p, 8% faster at 1440p and 9% faster at 4K. Perhaps most importantly, we are looking at 15-20% greater 1% low performance with the GeForce GPU.
Forza Horizon 4 has made it back into our benchmark suite thanks to its migration from the Windows Store to Steam. For a long time, this game heavily favored AMD GPUs, but Nvidia has addressed performance and now we’re looking at a situation where the RTX 3070 is a good bit faster than the 6700 XT, winning by a 15% margin at 1080p, 14% at 1440p, and 15% at 4K, so pretty consistent scaling there.
Nearly identical frame rates in Doom Eternal and this time it was the Radeon GPU that pulled ahead at 4K, albeit by just a 5% margin. Typically, it’s the other way around, but I believe the 6700 XT has a slight performance advantage here due to the larger VRAM buffer. Doom Eternal uses up to 9GB of VRAM at 4K, so the spill over into system memory does result in a slight performance hit for the RTX 3070.
At 1080p and 1440p frame rates are excessively high, so the margins there don’t really matter, but as luck would have it both delivered the same experience.
Speaking of VRAM, I wanted to include these Ghost Recon Breakpoint results using the ultra-quality preset, which is not the maximum preset which is called ‘ultimate’. These results might look a bit off at first glance because at 1080p they’re neck and neck, then at 1440p the GeForce GPU pulls ahead, but finally running at 4K it’s the 6700 XT delivering more than twice the performance.
This is a great example of what happens when you run out of VRAM. The 8GB buffer on the RTX 3070 is overwhelmed as the game calls for 9.5 GB of memory at this setting. As a result, game assets start flowing through the PCI Express bus into system memory at a significantly slower rate than local VRAM.
Of course, the simple fix here is to lower the demand on VRAM by reducing texture size, for example. The game probably requires only minor tweaks to solve the problem, but you will be reducing visuals to get the game running at a satisfactory level. Which does make us wonder how long it will be before this is standard operating procedure for those with 8GB graphics cards.
Last up we have Kingdom Come Deliverance. We wanted to go over these results because when testing the RX 6800 series we noticed how AMD Big Navi GPUs were suffering from performance issues in games using the CryEngine, like this one.
The RX 6800, for example, was capped at 71 fps at 1080p and 65 fps at 1440p. Thankfully, it looks like AMD has addressed those issues because the 6700 XT is now able to render 109 fps on average at 1080p and 84 fps at 1440p. It’s still slower than the RTX 3070, but performance is more where you’d expect it to be.
That’s how the Radeon 6700 XT and GeForce RTX 3070 match up in 16 of the 45 games we tested. Instead of having you spend an enormous amount of your time going over the other 30 graphs, here’s a summary of the results at 1080p and 1440p.
At 1080p, across the 45 games tested, we see that the 6700 XT is on average 4% slower than the RTX 3070, which is exactly what we found in our day-one review using a select 14 game sample. Of the 45 games, there were about a dozen where the 6700 XT was slower by a 10% margin or greater. Control and Apex Legends could be considered the outliers in this data.
For those wondering though, if we removed Control and Apex Legends results, the 6700 XT is still 4% slower on average. That’s still close enough to call a technical draw, albeit if you’re an avid Fortnite or Apex Legends player, you might want to go with the RTX 3070, but if you’re into Battlefield V, the 6700 XT might be a better choice.
Now at 1440p the gap does widen. It’s far from significant overall, but here the 6700 XT is 8% slower on average. It was also slower by a 10% or greater margin in 17 of the games tested. Basically, for 1440p gaming the RTX 3070 is not only the faster product but also the better value product, so not an ideal situation for AMD.
What We Learned
There you have it. For those of you wanting to see how the Radeon RX 6700 XT and GeForce RTX 3070 compare in a massive range of games, you have your answer. Things haven’t changed much from our day-one reviews. The 6700 XT fared a little worse at 1440p, but ultimately it doesn’t change our conclusion.
If you had the luxury of buying either GPU at or near the MSRP, which one should you get? Between the two, our pick would be the GeForce RTX 3070.
It just happens to be the case I’m a Fortnite player and the GeForce GPU is already faster before you even enable DLSS, making it the obvious choice for me. I also only play multiplayer games and typically opt for competitive-type quality settings, so memory capacity is not an issue.
We also like the higher recording quality when using NVENC. It’s a small improvement when recording, but it is better and if you stream it’s a lot better.
The Radeon RX 6700 XT remains a solid graphics card and should hold up well into the future, largely thanks to that 12GB VRAM buffer. We’re starting to see a few examples where you can run out of memory with 8GB of VRAM, though they’re still few and far between. Right now, you can easily get away with an 8GB frame buffer, but whether or not that will be true in a year or two remains to be seen. I’m a little doubtful, at least for those seeking maximum in-game visuals.
As we’ve discussed before, the Radeon 6700 XT could be made a more compelling choice by undercutting the RTX 3070’s pricing, but because of the current situation and GPU demand, it simply doesn’t make sense for AMD to take any such steps. In a scenario where we’re back to MSRP pricing across the board, the 6700 XT would need to be priced no higher than $400 — the same price as the 5700 XT — to become my top choice in this shootout.
Ray tracing support keeps maturing, but as you likely know by now, that’s of zero interest to me. It’s nonetheless an additional feature of the GeForce RTX series which you may care about. DLSS support is also rather limited, but if you can take advantage of it in the games you play, the performance benefits can be quite substantial.
One last thing to note. While I do prefer the RTX 3070 over the 6700 XT, the RX 6800 becomes an interesting proposition at $580. That’s 16% more than the 3070, but those extra $80 get you twice the VRAM and for those that like to hold onto their graphics card for more than 2 years, I feel the extra VRAM will come into play, assuming you play the latest titles with the visuals maxed out. It’s also 11% faster at 1440p according to our testing, and even without DLSS, you’re looking at well over 144 fps in a game like Fortnite.
But back to our two-way comparison, while the 6700 XT offers more VRAM and is supposed to be slightly less expensive, it’s not compelling enough given the inferior encoding support, weaker overall performance, and current lack of a DLSS competitor. And that’s really all I have to say on this one.