NVIDIA GeForce RTX 5070 Founders Edition Review

With AMD lurking, did NVIDIA deliver?

It has been just over a year since the last xx70 card from NVIDIA, as the GeForce RTX 4070 Super was launched in January of 2024, and while that was a mid-cycle refresh it does seem to be the logical point of comparison with the latest in the RTX 50 Series. But that may not be the best point of reference, as we may discover here.

Wait a minute, you may ask, does this mean that we have reached the point where meaningful gains from an RTX 50 Series card over an RTX 40 Series card sometimes need to look past the “Super” variants of their precursors, and back instead to the vanilla launch versions? I don’t want to spoil anything, but…yes.

Enter the GeForce RTX 5070.

Forgetting all about architecture for a moment, and Blackwell definitely has improvements over Ada Lovelace (insert ROP joke here), a high-level look at the specs might make one concerned about the performance of an RTX 5070 over the RTX 4070 Super from last year (plus, we already told you that might be the best point of reference anyway):

On paper the RTX 5070 is closer to the original RTX 4070 than the RTX 4070 Super, though the disparity in CUDA count compared to the Super is mitigated somewhat by the higher bandwidth memory and higher GPU clocks from the RTX 5070 (which also has a higher TGP). It should be noted, however, that the RTX 4070 Super is not a viable alternative to this new GPU, as available cards are hopelessly marked up – with many at double the original MSRP. Let up hope that the RTX 5070 avoids that fate.

Here is a look at the Founders Edition card, which is the same size as the RTX 4070/4070 Super FE cards (but with smoother and more refined industrial design):

Performance Testing

Let’s get right into some benchmark results, with our Founders Edition sample running on our Ryzen 7 9800X3D-powered GPU test bed.

3DMark Results

It isn’t big, but the RTX 5070 does score a victory in all three 3DMark benchmarks over the RTX 4070 Super – and by a much more convincing margin over the original RTX 4070. Will this translate into game benchmark wins, as well?

2560×1440 Benchmark Results

An impressive showing in Cyberpunk 2077 for the RTX 5070, actually matching the performance of an RTX 4070 Ti Super (with better 1% lows to boot)! Then in The Talos Principle 2 it’s just a frame faster than the RTX 4070 Super, but maintains a sizable lead over the RTX 4070.

And now, the other side of the coin. Yes, there are instances where the new RTX 5070 is slower than the RTX 4070 Super. It’s always faster than the RTX 4070, of course, but the results to follow will not exactly endear the new card to enthusiasts who expect larger generational gains than this:

I don’t fully understand the performance I’m seeing from the RTX 5070 on the charts above, but I re-ran the tests and validated the results. The margins are small, but the fact remains that at these settings the older xx70S card can be faster than the new xx70 non-S offering. Perhaps it comes down to the RTX 4070 Super’s sizable CUDA core advantage – though it has quite a bit less memory bandwidth than the new RTX 5070…

Ultimately, the RTX 5070 is a solid upgrade over the RTX 4070, but it only trades blows with the RTX 4070 Super – depending on the test. This is not exactly compelling, particularly considering the competition may heat up considerably in this space as AMD releases their new RX 9070 offerings, but there it is.

Testing DLSS and Frame Generation

One thing NVIDIA always has is software. DLSS has evolved into something I have heard described as “better than native”, and indeed it looks very, very good at the “Quality” preset that I prefer. On the other hand, “Balanced” is a more, well, balanced take on the tech, so, switching back to the 3440×1440 monitor and running Cyberpunk at the extremely demanding RT Overdrive preset, here are some DLSS “Balanced” results (with and without FG):

Aside from the omission of cards like the RTX 4070 Ti Super on the chart above (I forgot to test that one in time), it is still interesting to see how evenly matched the RTX 4070 Super and new RTX 5070 are in this DLSS test – until things like multi-frame generation (and 1% lows) are taken into account, that is. The RTX 5070 has a massive potential FPS advantage with up to 4x Frame Generation, but the RTX 4070 Super had more consistent frame times with FG enabled.

I am not sure what is going on with the 1% lows when using Frame Generation with the RTX 50 Series so far, considering the excellent performance from RTX 40 Series cards in this regard. And when you go all the way up to 4x FG the disparity in the frame times is very large. It’s not as noticable as it looks on the chart, but it’s still an interesting data point whenever we compare the two most recent generations from NVIDIA. And no, this 1% low phenomenon is not limited to Cyberpunk 2077, that’s just all I had in time for this review.

Power Draw

Ok, we’ve seen the gains the RTX 5070 has over the RTX 4070, but how much additional power might be required to fuel that advantage?

The power draw is consistent with the rated TGP, though it does momentarily exceed 250 watts at times. The fact that this new GPU exceeds the power draw of the larger RTX 4070 Super on the same process node may be a bit puzzling, until the RTX 5070’s use of more powerful GDDR7 – and the higher GPU clocks – is considered.

Final Thoughts

Well, if you have read this far you know that the new GeForce RTX 5070 can be around 20% faster than the GeForce RTX 4070, and of course there is the new multi-frame generation available to RTX 50 Series buyers. However, against the more recent RTX 4070 Super we only see consistent wins in synthetics benchmarks with the RTX 5070, with game results varying per title. It really doesn’t matter, as NVIDIA doesn’t intend this new card to be the replacement for the RTX 4070 Super anyway.

Like it or not, the RTX 5070 is positioned as the successor to the RTX 4070, and carries that original version’s $549 USD MSRP, rather than the $599 USD MSRP of the RTX 4070 Super. And in any case, remaining stock of the RTX 4070 Super, when available, is marked up to such a laughable degree that it’s probably for the best if we ignore its continued existence altogether.

The retail market for previous-gen RTX cards looks bleak

The real question remains unanswered for at least one more day, when we will finally have independent benchmarks of AMD’s latest GPUs. NVIDIA remains untouchable at the top of the graphics world, with the GeForce RTX 4090 still out-classing everything but their own RTX 5090, but this $549 segment is about to get a lot more interesting. Stay tuned.