News GeForce RTX 4070 vs Radeon RX 6950 XT: Which GPU Is Better?

[Admin]

Nvidia's new GPU takes on AMD's previous generation king.

GeForce RTX 4070 vs Radeon RX 6950 XT: Which GPU Is Better? : Read more

 

[oofdragon]

6950XT > 4070Ti, did you even need to ask? Just some TimeSpy facts :

6950XT = 22508
4070Ti = 22096
3090Ti = 21737
3080 = 17832
4070 = 17736

The 6950XT is 27% faster than a 4070 and any heads on benchmark will show you that. It also sports 16GB while the 4070 is just a overpriced rebranded 4060 at 12GB. Oh you not sure where to buy the 6950XT for $599? Just install a not well known app called "ebay" and buy it for even lower than that, new, you are welcome

 

[oofdragon]

LMAO call it bia$ed, this site just called a 4070>6950XT 😂😂😂

Just search 6950 XT vs 4070 on YouTube, here are real 1440p numbers:

10 game average 6950XT 16GB around 27% faster than the 4070 12GB

Shadow Tomb Raider 200 vs 166 (+20%)
Far Cry 6 156 vs 128 (+21%)
Death Stranding Ultra 218 vs 176 (+23%)
Witcher 3 Ultra 235 vs 201 (+%17)
Horizon Zero Dawn Ultra 153 vs 119 (+28%)
Borderlands 3 Ultra 129 vs 94 (+37%)
Cyberpunk High113 vs 89 (+27%)
Forza Horizon 5 Ultra 133 vs 96 (+34%)
Watch Dogs Legion Ultra 116 vs 86 (+36%)
Assassin's Creed Valhalla Ultra 116 vs 86 (+34%)

 

[mrv_co]

I wasn't encouraged by the 7700 XT rumors so I picked up a PNY 4070. Mainly because it's the only card that didn't require me replacing my SFF case and power supply while providing an appreciable upgrade over my 4yo 5700 XT. The PNY 4070 works as expected, but not having to spend the time and money to replace my case and power supply is the only thing that made it 'worth' $599 to me... definitely a <=$399 list price experience from the days of yore.

 

[irish_adam]

You guys are not even trying to be unbiased anymore are you? I mean who cares if the AMD card has far greater raster, DLSS is all that matters right? Reading your justification for giving the performance win to the 4070 was just pure cringe. I mean who isn't using their mid tier gaming GPU to run stable diffusion?

You used DLSS, ray tracing and AI as a justification for the performance win even though you had a whole category called Features and Technology where unsurprising you used the same argument!

If you would have given AMD the win in performance and Nvidia the win for pretty much everything else it would have been a fair article and the overall winner would have been the same. Twisting the performance win to fit your personal opinion which you openly admitted doing just makes you joke.

 

[The Historical Fidelity]

LMAO call it bia$ed, this site just called a 4070>6950XT 😂😂😂

Just search 6950 XT vs 4070 on YouTube, here are real 1440p numbers:

10 game average 6950XT 16GB around 27% faster than the 4070 12GB

Shadow Tomb Raider 200 vs 166 (+20%)
Far Cry 6 156 vs 128 (+21%)
Death Stranding Ultra 218 vs 176 (+23%)
Witcher 3 Ultra 235 vs 201 (+%17)
Horizon Zero Dawn Ultra 153 vs 119 (+28%)
Borderlands 3 Ultra 129 vs 94 (+37%)
Cyberpunk High113 vs 89 (+27%)
Forza Horizon 5 Ultra 133 vs 96 (+34%)
Watch Dogs Legion Ultra 116 vs 86 (+36%)
Assassin's Creed Valhalla Ultra 116 vs 86 (+34%)

Totally agree, with 16GB of vram and more powerful rasterization, the 6950xt wins every time. Ray tracing is still a gimmick and the article admits both cards cannot perform to an acceptable degree with ray tracing so why did the 4070 win the performance category? And DLSS is a tech/feature so idk why it was considered in the performance section either.

This article reeks of Nvidia shadow money!

 

[The Historical Fidelity]

I wasn't encouraged by the 7700 XT rumors so I picked up a PNY 4070. Mainly because it's the only card that didn't require me replacing my SFF case and power supply while providing an appreciable upgrade over my 4yo 5700 XT. The PNY 4070 works as expected, but not having to spend the time and money to replace my case and power supply is the only thing that made it 'worth' $599 to me... definitely a <=$399 list price experience from the days of yore.

I mean…sure…but…nah…mate. I hope you don’t get buyers remorse when the 7700xt comes out

 

[Elusive Ruse]

There's no competition here 6950 XT wins by a mile.

 

[kiniku]

Nvidia's new GPU takes on AMD's previous generation king.

GeForce RTX 4070 vs Radeon RX 6950 XT: Which GPU Is Better? : Read more

No brainer. One is a more powerful, far more efficient GPU with more features including widespread support for upscaling.

And on another note, the 7950XTX hype train is already running. Just like the 7900 pre-launch hysteria. If it was that great they would have launched that first as a actual competitor to the 4090. It's going to be another meg-over-clocked space heater.

 

[Vezza]

There's no competition here 6950 XT wins by a mile.

Agreed, this article seems to suggest you would be dumb to not go Nvidia.

 

[bitbucket]

I wasn't encouraged by the 7700 XT rumors so I picked up a PNY 4070. Mainly because it's the only card that didn't require me replacing my SFF case and power supply while providing an appreciable upgrade over my 4yo 5700 XT. The PNY 4070 works as expected, but not having to spend the time and money to replace my case and power supply is the only thing that made it 'worth' $599 to me... definitely a <=$399 list price experience from the days of yore.

Similar to my reasoning. I built a new PC that replaced one with an RX5700(non XT). Power requirements continue to go up. The cpu, the MB with its PCIe v5 and even the chipset. Also our regional power co is moving to variable rate pricing depending on demand at the time. New for me but I believe it's not for lots of others. The topper is that nearly everything in the basement is on the same circuit, the load is becoming a concern until I can have a new/separate circuit, or circuits, installed. The 4070 was a good upgrade for relatively little power increase. While I would have preferred to go with an AMD card, they just aren't competative in my desired performance range. But, ugh, that was a high price.

 

[Ravestein NL]

Why test a new gen card like the Nvidia RTX4070 vs a previous gen card like the RX6950XT?
Is that even a fair comparison?
Also imo the Nvidia and AMD video cards are so different it's like comparing an 🍎 with a 🍐.
The only thing you can actually compare is FPS and power (W) with the same settings for both cards (no RX or DLLS or FSR stuff).

 

[Vezza]

i only have a 3080, but i hate DLLS, who pays thousands for latency?

 

[motocros1]

generally i think jared does a pretty good job at being unbiased. in this particular comparison lumping the performance into such a general category does seem to be a bit vague. i mean who cares about ray tracing more than raster? who cares about dlss more than running native? the difference could've been stated more clearly. and i think the performance bracket should've had at least a few different sub categories.

 

[citral23]

How much did you get paid for this piece of cringe?

 

[The Historical Fidelity]

Similar to my reasoning. I built a new PC that replaced one with an RX5700(non XT). Power requirements continue to go up. The cpu, the MB with its PCIe v5 and even the chipset. Also our regional power co is moving to variable rate pricing depending on demand at the time. New for me but I believe it's not for lots of others. The topper is that nearly everything in the basement is on the same circuit, the load is becoming a concern until I can have a new/separate circuit, or circuits, installed. The 4070 was a good upgrade for relatively little power increase. While I would have preferred to go with an AMD card, they just aren't competative in my desired performance range. But, ugh, that was a high price.

$600 is a lot of money for a card that in the near future will require low texture settings due to lack of vram. Should have waited for the 7800 or 7800xt to come out.

 

[newtechldtech]

... another "just buy it" crap

 

[Nexus52085]

I love Toms, and I've read many articles from you guys rarely feeling the need to post. However, this is one of those times. I must have seen dozens of comparisons anywhere from reputable YouTube channels to amateur gaming benchmark channels, and none have put the 4070 in the same performance class as the 6950XT. This article is beyond BS.

 

[CharlesOCT]

The 6800XT gives higher performance than the 4070 (ref toms' GPU heirarchy chart), costs $100 less, and gives 16Gb.

Bad logic in this article all around.

 

[JarredWaltonGPU]

6950XT > 4070Ti, did you even need to ask? Just some TimeSpy facts :

6950XT = 22508
4070Ti = 22096
3090Ti = 21737
3080 = 17832
4070 = 17736

The 6950XT is 27% faster than a 4070 and any heads on benchmark will show you that. It also sports 16GB while the 4070 is just a overpriced rebranded 4060 at 12GB. Oh you not sure where to buy the 6950XT for $599? Just install a not well known app called "ebay" and buy it for even lower than that, new, you are welcome

Time Spy is literally the stupidest way of comparing GPUs. It's not even a game. And your referenced benchmarks don't jive with mine at all. Where did you get those from (other than ensuring there's not a single DXR title represented, nor are there details on what hardware was used for testing). Your quoted numbers, where I ran the same benchmarks:

Far Cry 6: 156 vs 128 (+21%)
Mine @ Ultra+HD: 149 vs 134 (+11%)

Horizon Zero Dawn Ultra: 153 vs 119 (+28%)
Mine @ Ultimate: 166 vs 140 (+19%)

Borderlands 3 Ultra: 129 vs 94 (+37%)
Mine @ Badass: 140 vs 103 (+36%)

Forza Horizon 5 Ultra: 133 vs 96 (+34%)
Mine @ Extreme: 137 vs 119 (%15%)

Watch Dogs Legion Ultra: 116 vs 86 (+36%)
Mine @ Ultra: 117 vs 95 (+23%)

Every one of the numbers you quoted looks highly suspect to me, given I ran those same tests. Unless they're not using the built-in benchmarks? But clearly my test PC must have been at least 10% faster in several cases (HZD, BL3, FH5).

Did they use an AMD CPU? Perhaps that would skew things in the 6950's favor more. But in BL3 (a highly pro-AMD game), using Badass, I scored ~10 fps higher with both GPUs. I also scored higher with Forza Horizon 5 at the Extreme preset. But Forza has had some major changes over time, so if one GPU was tested last year (6950) versus the other in 2023, that would explain things.

Even then, the difference between my rasterization results (6950 is ~14% faster at 1440p) and what you gave isn't really that huge. Bias the testing in favor of AMD and you'd get there easily enough.

Now what about all the results you're ignoring? I guess anything with ray tracing doesn't matter, where the 4070 wins by anywhere from 3% (Metro) to 65% (Minecraft), with an average advantage of 21%. And of course DLSS isn't a thing either, probably, because that would just boost the 4070 by 30%. And AI applications aren't a thing!

But yeah, go ahead and buy a used card off eBay. You can score some great deals... and also potentially a serious lemon.

 

[JarredWaltonGPU]

i only have a 3080, but i hate DLLS, who pays thousands for latency?

DLSS reduces latency. Only DLSS 3 Frame Generation increases latency.

 

[JarredWaltonGPU]

The 6800XT gives higher performance than the 4070 (ref toms' GPU heirarchy chart), costs $100 less, and gives 16Gb.

Bad logic in this article all around.

No, that's cherry picking results. The 6800 XT barely beats the RTX 4070 in the rasterization hierarchy (127.3fps vs 123.2fps at 1080p ultra, 103.0fps vs 98.9fps at 1440p ultra). In the DXR hierarchy, it's not even close, and as noted, DLSS isn't factored in at all (69.4fps vs 48.5fps at 1080p ultra, 45.2fps vs 31.1fps at 1440p ultra).

DLSS really is a thing that people use, and frankly probably should use more. It's in most demanding ray tracing games, and even Quality upscaling is enough to get the 4070 to 60+ fps at 1440p in most games. I'm not talking about Frame Generation here, just regular upscaling.

 

[irish_adam]

Time Spy is literally the stupidest way of comparing GPUs. It's not even a game. And your referenced benchmarks don't jive with mine at all. Where did you get those from (other than ensuring there's not a single DXR title represented, nor are there details on what hardware was used for testing). Your quoted numbers, where I ran the same benchmarks:

Far Cry 6: 156 vs 128 (+21%)
Mine @ Ultra+HD: 149 vs 134 (+11%)

Horizon Zero Dawn Ultra: 153 vs 119 (+28%)
Mine @ Ultimate: 166 vs 140 (+19%)

Borderlands 3 Ultra: 129 vs 94 (+37%)
Mine @ Badass: 140 vs 103 (+36%)

Forza Horizon 5 Ultra: 133 vs 96 (+34%)
Mine @ Extreme: 137 vs 119 (%15%)

Watch Dogs Legion Ultra: 116 vs 86 (+36%)
Mine @ Ultra: 117 vs 95 (+23%)

Every one of the numbers you quoted looks highly suspect to me, given I ran those same tests. Unless they're not using the built-in benchmarks? But clearly my test PC must have been at least 10% faster in several cases (HZD, BL3, FH5).

Did they use an AMD CPU? Perhaps that would skew things in the 6950's favor more. But in BL3 (a highly pro-AMD game), using Badass, I scored ~10 fps higher with both GPUs. I also scored higher with Forza Horizon 5 at the Extreme preset. But Forza has had some major changes over time, so if one GPU was tested last year (6950) versus the other in 2023, that would explain things.

Even then, the difference between my rasterization results (6950 is ~14% faster at 1440p) and what you gave isn't really that huge. Bias the testing in favor of AMD and you'd get there easily enough.

Now what about all the results you're ignoring? I guess anything with ray tracing doesn't matter, where the 4070 wins by anywhere from 3% (Metro) to 65% (Minecraft), with an average advantage of 21%. And of course DLSS isn't a thing either, probably, because that would just boost the 4070 by 30%. And AI applications aren't a thing!

But yeah, go ahead and buy a used card off eBay. You can score some great deals... and also potentially a serious lemon.

You are ignoring the fact that you had a whole section dedicated to features which is where ray tracing and DLSS belonged. In the performance section you gave the win to Nvidia because..

However, we have to pick a winner, and so Nvidia ends up with the overall performance lead thanks in large part to DLSS tech, but also due to its significantly faster ray tracing hardware. Factor in the AI performance and RTX 4070 easily comes out on top.

If you were going to have a whole section to raytracing and DLSS why was the performance section not based of the raster performance? which btw is still very much relevant in ALL games not just a few which DLSS and even ray tracing are in.

But you are right, we are all wrong. DLSS is all that matters, we are all stupid for playing games native. Your opinion is all that matters and who cares what actual performance numbers say.

 

[Amdlova]

In darktimes where VRAM need it... You might choose AMD or pay more for a 4070ti

 

[CharlesOCT]

The 6800 XT barely beats the RTX 4070 in the rasterization hierarchy

yes the 6800XT barely wins in rasterization.

In the DXR hierarchy, it's not even close,

...except neither GPU really provides a great gaming experience at 1440p native with ray tracing.

Yes, the 4070 wins in a metric that results in unplayable framerate (good luck with 45fps in 1440 DXR).

DLSS really is a thing that people use

Is your source for this claim NVIDIA's blog article NVIDIA BLOG DXR/DLSS? I think they're "cherry picking" statistics there: it doesn't state if people turned features on and leave them on, or just turn around and turn it back off later..

I personally prefer to just run natively at a lower resolution when I can't get enough fps to go to a bigger resolution. I find it very hard to distinguish when ray tracing effects are on or off (if severely reduced fps isn't already a dead giveaway).