NVIDIA is Selling Lies | RTX 5070 Founders Edition Review & Benchmarks
jimmy_thang
March 13, 2025
We analyze the RTX 5070’s specs, pricing, gaming performance, ray tracing, and power efficiency
The Highlights
- The RTX 5070 has 6144 CUDA cores, a 192-bit bus, and 12GB of GDDR7 memory along with an advertised 2.51 GHz boost frequency
- NVIDIA’s claim that the 5070 is on par with a 4090 with MFG is simply not true
- NVIDIA’s marketing here is intentionally manipulative and a misrepresentation of reality
- Original MSRP: $550
- Release Date: March 5, 2025
Table of Contents
Grab a
GN15 Large Anti-Static Modmat to celebrate our 15th Anniversary and for a high-quality PC building work surface. The Modmat features useful PC building diagrams and is anti-static conductive. Purchases directly fund our work! (or consider a
direct donation or a
Patreon contribution!)
Intro
NVIDIA is selling lies and reviewers shouldn’t be afraid to mince words. That is precisely what the RTX 5070 is. It was marketed on the back of lies considering NVIDIA stated that the card, which has 12GB of VRAM, would offer 4090, which has 24GB of VRAM, performance for $549 with the help of AI.
Here’s an NVIDIA-compliant comparison of the RTX 5070 with MFG against the RTX 4090 without any frame generation at all.
Editor's note: This was originally published on March 4, 2025 as a video. This content has been adapted to written format for this article and is unchanged from the original publication.
Credits
Test Lead, Host, Writing
Steve Burke
Testing
Mike Gaglione
Video Editing
Vitalii Makhnovets
Camera
Tim Phetdara
Writing, Web Editing
Jimmy Thang
These are really abusive settings, stressing both cards heavily. But NVIDIA said the 5070 would be equal to the 4090 (watch our review), and we’re not even close. The 5070’s spiky behavior is exceeding 200 ms in this like-for-like comparison except that it has MFG. The 4090 isn’t perfect, but WITHOUT any frame generation at all, it’s hitting at worst about 50-60 ms, with an average closer to 26.
Here’s what that actually looks like. Using NVIDIA’s own FrameView tool, we also get a glimpse into another problem: Latency. The PC Latency on the 4090 was around 51 ms during a like-for-like comparison. 51 is a lot better than what we saw in the 5070 with MFG 4X after 5 minutes of being in the game. That was 500-720 ms. Now, to be fair, if you only played Cyberpunk with these settings for about 15-20 seconds at a time, things look a lot better…
The 5070 is getting absolutely clobbered for VRAM, which is just 12GB to the 4090’s 24GB. To call these the same is an absolute, flat-out lie. These cards are not the same. In situations where the card runs out of VRAM in particular, it can never dream to be a 4090. It is simply impossible.
Here’s a quick chart showing passes 1 through 3 of the 5070 as it gets progressively worse with each benchmark pass as the framebuffer fills. The 4090, meanwhile, maintains its performance the entire time.
NVIDIA has got to stop lying. This isn’t just marketing, this is an actual, verifiable lie. But let’s just pretend MFG 4X is somehow a fair comparison against the 4090 in a scenario that’s maybe not VRAM constrained. It’s an OK technology, but it is not a like-for-like comparison because the images themselves are not the same. When you turn it on, it’s like Clive here has 6 feet.
NVIDIA could have launched its RTX 5070 to mediocre or lukewarm reviews. Instead, it decided to openly lie on stage and manipulate its performance numbers to mislead consumers into thinking an otherwise lukewarm card is equal to the flagship of last generation. NVIDIA, this was avoidable. You are making all of the unforced errors right now. This was a completely unnecessary fumble. We could have just listed all the numbers and called it a day, but now have to start by writing about NVIDIA’s lies and marketing.
RTX 5070 Overview
The RTX 5070 is supposed to have an MSRP of $550. It’s hard to say what it’ll actually be for street pricing. But basing only off of MSRP, that makes this card $50 higher than the MSRP of the RTX 3070 at its launch in 2020 and $50 cheaper than the RTX 4070’s launch MSRP of $600.
This looks like the patterned NVIDIA stutter-stepping of price slowly upwards, where they overshoot, then bring it back down. We saw this from 10 to 20 to 30 series.
The RTX 5070 GPU has an advertised 6144 CUDA cores, a 192-bit bus, and 12GB of GDDR7 memory. It might even have all of its ROPs. The 5070 Ti has an MSRP at $750 and carries 16 GB of GDDR7 memory on a 256-bit bus, so there’s more bandwidth, a little more capacity, and an increase to 8960 CUDA cores, and it may even have all of the ROPs that you’re paying for.
Clocks on the 5070 are advertised at 2.51 GHz boost.
RTX 5070 Pricing
A quick glance at pricing breaks it down like this:
There are 5 MSRP models of the 5070 Ti (read our review) on Newegg, which is just enough to technically claim the MSRP exists. The rest of them are higher, often $900 and over. Board partners have now directly repriced their cards, with VideoCardz noticing that MSI had eliminated all of its $750 RTX 5070 Ti cards just before the 5070 launch.
As we write this review, following the VideoCardz news post, MSI has now reintroduced two $750 models, dropping from $820 and $840 to $750. NVIDIA has historically applied pressure to partners to make MSRP units available, so it wouldn’t surprise us if the VideoCardz story triggered a reaction to keep two models available. They can also achieve this with rebates.
As for things available right now, we found the RX 7800 XT Pulse for $530 and... that’s it. When we sorted Newegg to $450 to $650 for “in stock” and “sold by Newegg,” meaning no third-party sellers and no unavailable models, all we got were a bunch of refurbished units and the 7800 XT Pulse.
It really is crazy right now.
We’re going to get into the benchmarks next. This review contains most of the games that we benchmark since we’re preparing for the 9070 and 9070 XT launch. We currently have around 50-60 GPUs tested in our lineup, so to fit the charts, we need to remove a few from the chart.
Here’s what we’ve added and removed:
Cards Removed & Added
- We’ve added the RX 7900 GRE (read our review). We won’t be spending time talking about it today since it’ll be replaced tomorrow and will likely be a key comparison for the 9070 series, but it’s on the charts.
- We’re removing either the RTX 4080 (watch our review) or Super (read our review) in every chart, but keeping the other. We’ll favor the Super where we have data for it. These GPUs are within 1-3% of each other and are not worth the chart space. We have more interesting stuff to show.
- We’re removing the 2080 Ti (watch our review), 2080 Super (watch our review), and 2080 (watch our review). Most of these results are available in our previous charts for the 5070 Ti review and are directly comparable.
- We’re removing the 5080 with missing ROPs. You can find that data in a standalone article.
- And, of course, we’re removing the 9070 and XT -- we can’t show that until tomorrow at the time of this writing.
RTX 5070 Benchmarks
FFXIV 4K
Final Fantasy 14: Dawntrail is up first for a traditional raster title at 4K, without any bulls*** to artificially inflate numbers.
This game is actually playable on most modern hardware at this resolution. The RTX 5070 ran at 78 FPS AVG, landing it right between the 4070 Ti (watch our review) and the 4070 Ti Super. Maybe we should call it the 4070 Ti V3 instead. The 5070 Ti outperforms the 5070 by 25% here, the 4070 Ti Super is ahead by 11%, and the 7900 XT by 6%.
Going by name only, the generational lead over the 4070 (watch our review) is 30%, or 48% over the 3070 (watch our review) and 118% over the 2070 (watch our review), followed by 244% over the 1070.
Unfortunately, the 5070 is not better than the 4090 -- but if we were to be artificially intelligent and actually intellectually dishonest, it might be!
FFXIV 1440p
At 1440p, the RTX 5070 Ti’s lead over the RTX 5070 is reduced to 22% from 25% at 4K. The 5070 is also reduced in its advantage over the 4070 Ti, now just 1% from around 6% at 4K. From 150 FPS to 172 FPS, we now have 3 GPUs from NVIDIA. If you want to make NVIDIA look as ridiculous as possible, you could maximize the size of the clown car by drawing a box from the 4080 Super at 202 FPS to the 4070 FE at 117 FPS. The result is 7 modern NVIDIA GPUs spanning an 85 FPS range. Divided evenly, that would be one GPU for every 12 FPS. In this situation though, 4 of them land within a 36 FPS range: The 5070 Ti, 4070 Ti Super (read our review), 5070, and 4070 Ti mean you can dial it into the individual decimals if you were so picky about your performance.
The 7900 XT (read our revisit) lands in the middle of all of these and has a 12% lead over the RTX 5070, assuming all ROPs are present on the 5070.
The 5070 is, again, not better than the 4090.
FFXIV 1080p
Remarkably, we still have OK scaling at 1080p for most of these GPUs. The fact that the 9800X3D allows the 5090 (read our review) and 4090 to still have a slight gap at 1080p really speaks to how exciting AMD’s 9800X3D is.
Anyway, the 5070 ran this workload at 225 FPS AVG and fell below the 4070 Ti. No wonder NVIDIA wants everyone using MFG. The 4090, predictably, outperforms the 5070 massively. It’s a 67% gap here.
The 7900 XT now has a reduced lead of 9%.
Black Myth: Wukong - 4K
Black Myth: Wukong at 4K and rasterized had the 5070 at 40 FPS AVG, giving the 5070 Ti a 27% lead. The 5070 really struggled in this one and ended up tied with the 4070 Ti and, to its benefit, the 7900 XT. We’ll be curious to see how the 9070 does in this one… Surely we don’t already know as we’re writing this and definitely haven’t looked at the results and in no way would we ever encourage you to just wait and see what the results are…
If we were to randomly multiply the frames, it’d be better than the 4090 -- until you randomly multiplied the 4090 with lossless scaling. Jensen’s claim is like a schoolyard argument: The 5070 is infinity times better.
Black Myth: Wukong - 1440p
At 1440p, we re-introduce other 70-class cards of the past. The 5070 FE ran at 72 FPS AVG, with frametime pacing unremarkable. The 5070 is tied with the 4070 Ti and 7900 XT. Generationally by name only, the 5070 improves on the 4070 by 22%, the 3070 by 58%, and the 2070 non-Super by 137%.
The 5070 Ti leads the 5070 by 21% here.
Black Myth: Wukong - 1080p
1080p has the RTX 5070 down at 98 FPS AVG, allowing the 4070 Ti to gain a slight 2% lead. The 5070 Ti is ahead by 17.6% here, assuming all ROPs are present.
Against the prior cards containing 70 naming, the 5070 is ahead of the 4070 by 18%, the 3070’s 63 FPS AVG by 56%, and the 2070’s 43 FPS by 126%.
Starfield - 4K
Starfield at 4K is up now. The RTX 5070 ran at 54 FPS AVG here, giving the 4070 Ti’s 59 FPS a 9% lead. It’d be interesting if AMD’s 9070 performed similarly to the 7900 XT here since the Hellhound is 20% ahead of the 5070. We’ll find out tomorrow.
The 5070 Ti has a larger lead over the 5070 in this test than the last 1080p benchmark, running a 68 FPS AVG for a 27% advantage over the 5070 FE. Performance of the 5070 over the 2070’s 22 FPS is about 145% ahead.
Starfield - 1440p
1440p significantly reduces the 5070 Ti’s uplift over the 5070, bringing it down to 22% from the prior 27%. The 7900 XT, which is an important comparison in this test for... reasons, runs at 98 FPS AVG with comparable lows to its neighbors. That has it 18% ahead of the RTX 5070’s 83 FPS AVG.
As for the RTX 4070, the 5070 leads by a paltry 10.6%, then 48% over the 3070’s 56 FPS.
Starfield - 1080p
At 1080p, the 5070’s 104 FPS AVG planted it between the 4070 and 4070 Ti again, with an unimpressive 9% improvement on the 4070. The gap shrinks as the resolution decreases in this game. The 5070 Ti is now ahead by only 19%, down from 27% at 4K. As for AMD, its RX 7900 XT is 15% ahead here, down slightly from the 1440p advantage.
Dragon’s Dogma 2 - 4K
Dragon’s Dogma 2 at 4K is next. This game is relatively heavy on the GPU in our test area, but is unique for its ability to also produce a heavy CPU load in cities.
The RTX 5070 ran at 56 FPS AVG here and had good frametime pacing represented in the lows, but then again, so did everything around it. It’s proportional.
The 5070 very slightly leads the 4070 Ti.
As for the 4090: We’re still somehow not matching the RTX 4090, which is up at 98 FPS. We must have mistyped the =5070*4 formula that Jensen prescribed reviewers.
The 5070 Ti leads the 5070 by a much larger 31% in this benchmark, producing one of the most notable jumps yet. The 3090 Ti ran at 64 FPS for a 14% lead over the 5070, which might seem totally non-sequitur and unrelated, but it isn’t.
Anyway, the 7900 XT held a 61 FPS AVG in this one, with the XTX just past the 5070 Ti.
Dragon’s Dogma 2 - 1440p
At 1440p, the RTX 5070 basically ties the RTX 4070 Ti again. The 7900 XT’s 104 FPS AVG is just a 9% lead here, with the 4070 Ti Super a slight step above that. The 5070 Ti has a 25% lead over the 5070 here, posting one of its better 1440p comparative results.
Somehow, and we don’t know how this happened, the RTX 4090 just seems to be better than the RTX 5070. That’s just so weird. Maybe we should run NVIDIA Multi-Fraction Generation to divide the 4090 down to a 5070.
Dragon’s Dogma 2 - 1080p
At 1080p, the 5070’s 126 FPS AVG has it just behind the 4070 Ti. The 7900 XT leads by only 7% here, with the 5070 Ti knocked down to a 21% lead from 25% at 1440p. The 5070 Ti’s 151 FPS AVG is also about the same as the 7900 XTX, which has its own slight advantage.
As for the 4090, well, ours really must be broken since it’s not making the 5070 look good enough.
Cyberpunk 2077: Phantom Liberty - 4K
Cyberpunk is up now. We have a more limited data set with this since we updated the game to v2.21 and have been working on re-running everything.
At 4K, the 5070 ran at 41 FPS AVG. That put it 6% ahead of the 4070 Ti and 144% ahead of the 2070, or 68% ahead of the 24 FPS result of the 3070. As for the 5070 Ti, its 50 FPS AVG gives it a 22% lead here, with the 7900 XTX (watch our review) ahead of that at 57 FPS AVG. The 7900 XT has a noteworthy 13% lead against the 5070 here.
Cyberpunk 2077: Phantom Liberty - 1440p
At 1440p, the 5070 ran at 89 FPS AVG and had lows at 75 FPS 1% and 71 FPS 0.1%. The 5070 Ti at 108 FPS is about 21% ahead, not changing much from 4K. The 7900 XT is ahead of the 5070 in AVG, 1%, and 0.1%, holding a 12% lead in average framerate.
Once again, the 4070 Ti basically ties the 5070, with the latter slightly ahead this time.
Cyberpunk 2077: Phantom Liberty - 1080p
1080p reintroduces other 70-class cards: The 5070 Ti ran at 167 FPS AVG, with the 5070 at 138. That’s still 21%. The 4070 hasn’t been rerun here yet, but the 3070 is present at 85 FPS, yielding a 62% lead to the 5070, followed by the 2070 at 60 FPS for 130%, then the GTX 1070 at 35 FPS AVG. That’s about a 293% improvement to the 5070.
AMD’s RX 7900 XT seems like it’ll be particularly relevant soon. This one ran at 150 FPS AVG, with lows expected at 116 and 104. That’s an 8.8% improvement over the 5070.
Dying Light 2 - 4K
Dying Light 2 is up now. This is another of the heavier games, especially with RT later.
At 4K, the 5070 ran at 56 FPS AVG and struggled in this title, allowing the 5070 Ti an advantage of 25%. The 7900 XT doesn’t look great by comparison here, at 55 FPS AVG itself.
As for the 4070 Ti, it’s again roughly equal.
Dying Light 2 - 1440p
1440p has the 5070 at 106 FPS AVG, now trading places with the 7900 XT. The 5070 Ti’s lead is reduced to 22%, with the 4070 Ti now falling slightly behind the 5070.
By generational naming, the 4070 ran at 78 FPS AVG (so the 5070 is 36% better) and the 3070 was at 67 FPS AVG (or 60% better on the 5070).
Resident Evil 4 - 4K
Resident Evil 4 is up last for raster testing. We’re almost through these.
At 4K, the 5070’s 78 FPS AVG has it just behind the 4070 Ti that we’ve been tracking. It’s also about 6 FPS ahead of the 7800 XT, so measurably different but functionally equal. The 5070 Ti’s 107 FPS AVG positions it 36% ahead in this one, which is a huge gain and among the largest we’ve seen. The 7900 XT seems like a good AMD comparison, landing at 100 FPS AVG and leading the 5070 by 28%.
Resident Evil 4 - 1440p
At 1440p, the 5070 held a 152 FPS AVG and trailed the 4070 Ti by 8 FPS. The 5070 Ti is 30% higher framerate here, down from 36% at 4K. As we’ve seen, the gap tends to close at lower resolutions, although it’s still a huge gap in this game.
The 7900 XT is now up at 186 FPS AVG, with a reduction in the percentage advantage to 22% from 28% at 4K.
Generationally, the 5070 runs 23% faster than the 124 FPS on the 4070, 54% faster than the 91 FPS on the 3070, and 157% ahead of the 59 FPS for the 2070.
Resident Evil 4 - 1080p
Finally for raster, we’re now at 1080p for Resident Evil 4. This one is interesting for the further closing of the gap between the 5070 and 5070 Ti, which now ranges from 282 FPS to 224 FPS for a 26% improvement on the Ti. The 7900 XT is about 18% higher framerate here.
RTX 5070 Ray Tracing Benchmarks
Grab a
GN15 Large Anti-Static Modmat to celebrate our 15th Anniversary and for a high-quality PC building work surface. The Modmat features useful PC building diagrams and is anti-static conductive. Purchases directly fund our work! (or consider a
direct donation or a
Patreon contribution!)
We’re moving on to ray tracing now. NVIDIA has historically held significant advantages in some games for ray tracing.
AMD says it has significantly improved its ray tracing performance. We explained why the company is claiming this in our news piece covering the 9070 announcements, so this will become highly relevant in tomorrow’s reviews.
Ray Tracing - Black Myth: Wukong 4K
Black Myth: Wukong is up first. This is one of the two titles in this RT test suite that heavily favors NVIDIA. We’re testing with upscaling here.
At 4K, the RTX 5070 ran at 40 FPS AVG. That has the 5070 Ti at 30% ahead, with the 4070 Ti about tied with the 5070. AMD doesn’t appear until the 7900 XTX, down at 20 FPS AVG. That’s a massive lead of 99% over the 7900 XTX. The 5070 doubles the 7900 XTX’s performance.
That’s not good for AMD’s last generation. We’ll see if that lead can be halved with the new generation. The 5070 leads the RTX 3080 by 44%. Based on math from AMD’s claims in its presentation, that’s about where it should land here. Too bad no one knows if that’s a good reference point yet.
Ray Tracing - Black Myth: Wukong 1440p
At 1440p, the 5070 Ti ran at 88 FPS AVG, the 5070 at 73 FPS AVG (between the 4070 Ti and 4070 Ti Super), and the original RTX 70-class card ran at 24 FPS AVG. AMD’s best here, as of today and not tomorrow, is the 7900 XTX at 37 FPS AVG.
Ray Tracing - Black Myth: Wukong 1080p
Now for 1080p. The 5070 held a 97 FPS AVG here, encroaching on the 4090 but still not beating it. The 5070 Ti leads the 5070 by just 15% in this situation, with the 5070 now ahead of the 4070 Ti and Ti Super cards after slow gains in the other resolutions. The 7900 XTX ran at 49 FPS AVG, closer to a 4060 and behind the 3070. AMD named the 3080 in its slideshow announcing the card. If it lands near that mark, it’d be around or ahead of the 66 FPS AVG result for the FTW3.
Speaking of: EVGA really knew what was coming when it left the GPU market.
Ray Tracing - Dragon’s Dogma 2 4K
Dragon’s Dogma 2 is back with RT now. This one is more balanced between the vendors.
At 4K, the 5070 ran at 49 FPS AVG, establishing a 9% lead for the 7900 XT. That’s a similar gap to what we saw without RT. The 5070 Ti leads by 30% again here, with the 7900 XTX leading that.
Ray Tracing - Dragon’s Dogma 2 1440p
At 1440p with RT, Dragon’s Dogma 2 puts the 5070 at 83 FPS AVG, reducing the 5070 Ti’s lead to 24%. The 7900 XT sits between both, with the 7900 XTX ahead of the 5070 Ti.
Based on the charts AMD has released, the math would position the 9070 and 9070 XT as flanking the 5070 Ti, but we’ll find out soon enough.
Generationally, the 5070 leads the 4070 by 23%, the 3070’s 51 FPS by 62%, and the 2070 non-Super by 145%.
Ray Tracing - Dragon’s Dogma 2 1080p
Down to 1080p, the 5070 Ti’s lead over the 5070 is now 22%, with the 7900 XTX still leading the Ti and 7900 XT still leading the 5070. The 4070 is relatively close to the 5070 here, now with an 18% advantage to the 5070.
But maybe the 4090 can breathe some excitement into it. The 4090 ran at 169.2 FPS AVG. If we multiply the 5070 by a billion, that’d put it at 107.4 billion FPS AVG, which is an uplift of 6,347,516.73%. This is clearly the card to get. And as we all know, it’s not possible to multiply the 4090 by arbitrary numbers with lossless scaling because then that would hurt 50-series marketing.
Ray Tracing - Dying Light 2 4K
Dying Light 2 with RT is next. This first one is 4K upscaled. The RTX 5070 ran at 44 FPS AVG here, just below the 4070 Ti. The 7900 XTX held 46 FPS AVG in this one, again showing AMD’s prior deficit in RT performance. It’s not as bad as in Black Myth, but considering the original pricing of the 4070 Ti and 7900 XTX, AMD was in a position that was hard to fight from. We’ll see how the 9070 series compares tomorrow.
Ray Tracing - Dying Light 2 1440p
At 1440p, the 5070 beats the 7900 XT by 13%, falls behind the 4070 Ti, and allows the 5070 Ti, which is basically an RTX 4080 v3 or v4, a lead of 27%.
Ray Tracing - Dying Light 2 1080p
At 1080p upscaled, the 5070 produced 116 FPS AVG and again sat just below the 4070 Ti, with the 7900 XTX and 7900 XT flanking the 5070. The 5070 Ti ran at 141 FPS AVG, which will be the number for AMD’s 9070 XT to target.
Generationally, the 5070 leads the 4070 by 25% and 3070 by 58%.
Ray Tracing - Resident Evil 4 4K
Resident Evil at 4K upscaled is next. This one has the 5070 at 91 FPS AVG, roughly tying the 4070 Ti once again. The 7900 XT leads the 5070 by 18%, with the 5070 Ti leading by 29%. The 7900 XTX sits ahead of that at 134 FPS AVG.
Ray Tracing - Resident Evil 4 1440p
At 1440p, the 5070’s 149 FPS AVG put it 23% ahead of the 4070 and 70% ahead of the 3070. The 7900 XT leads the 4070 Ti and 5070.
Ray Tracing - Cyberpunk 4K RT Ultra
Finally for RT, our new Cyberpunk results with the updated game version.
First, with 4K and RT Ultra, the 5070 ran at 17 FPS AVG. This is without upscaling and is intentionally the heaviest workload we run.
That positions it right between the 7900 XT and 7900 XTX. That’d be unfortunate positioning if AMD weren’t launching something with better RT in a day, but we’ll have to check in tomorrow for the rank.
The 5070 Ti continues its trend of being NVIDIA’s third or fourth iteration of a 4080 card and leads the 5070 by a huge 56% here. The workload is just too heavy for the 5070 to handle and it has neither the bandwidth or compute capability.
Ray Tracing - Cyberpunk 1080p RT Ultra
Here’s RT Ultra at 1080p. The 5070 held 64 FPS AVG with these settings, putting the 5070 Ti about 32% ahead. The 7900 XTX trails the 5070 here, unfortunately for AMD’s former flagship.
Ray Tracing - Cyberpunk 4K RT Medium
We also run RT Medium for Cyberpunk. We’ve found that RT Ultra and RT Medium can significantly affect the hierarchical ranking of NVIDIA and AMD, so we run both. NVIDIA runs away at Ultra.
With these settings and at 4K still, the 5070 now runs at 22.5 FPS AVG. This reduces the 5070 Ti’s lead to a more normal 35%. The 5070 just didn’t have the ability to keep up at 4K/RT Ultra.
RTX 5070 Thermals
We’ll keep thermals short and simple. Using our usual benchmark of Port Royal at 4K and looping, we measured the 5070 FE at steady state at around 74 degrees Celsius for GPU core temperature. The memory temperature was about 76 degrees. Both of these numbers are acceptable; memory is well within spec and is completely fine. Core has some room for a hotter computer case. The core temperature isn’t impressive, but is acceptable, and there’s a little bit of buffer there.
The 5070 FE’s fans ran at about 2500 RPM to maintain this temperature. We skipped acoustic testing this time since we have the two 9070 reviews we have to get through.
RTX 5070 Power Efficiency
For efficiency as usual, we use a power interposer in between the GPU and the power supply. That means we’re intercepting slot power and the PCIe power through the cables and we do that so we can isolate the GPU entirely, measure its power consumption during a workload. Then we take the frame rate numbers to do some simple math and produce an efficiency number.
In these charts, you’ll get a few things. You’ll get the total power consumption for that workload in watts and you also get its efficiency in FPS per watt not workload-normalized so we allow it to run at the frame rate that it can naturally run at.
Efficiency: FFXIV 4K
For efficiency with Final Fantasy 14 at 4K, we ended up with this data. This chart isn’t as dense as our 1440p chart that’s up next.
The 5070 FE ran at 0.33 FPS/W here, pulling 233W during the workload. The 5070 Ti pulled 264W, allowing it an improvement in efficiency to 0.37 FPS/W. AMD’s prior RX 7900 XT wasn’t particularly efficient, giving NVIDIA a large advantage in efficiency for the 5070. The FPS was close enough to be mostly observably equal to a player, but the 7900 XT ran at just 0.25 FPS/W from its 324W power draw during the test.
We’ll have to see what the 9070 series does to improve this, as this has been one of AMD’s GPU weaknesses over the years.
Efficiency: FFXIV 1440p
At 1440p, the 5070 ran at 0.65 FPS/W. That puts the 5070 Ti as about 12% more efficient than the 5070 when producing a variable workload. The AMD 7900 XT pulled 325W in this test, landing at 0.53 FPS/W. Its framerate is higher, but the power is also disproportionately higher, which hurts its efficiency despite a higher framerate.
Efficiency: FFXIV 1080p
At 1080p, the 5070 hits nearly 1 FPS/W. The 4060 Ti has passed it for efficiency, with the 5070 Ti improved by 0.11 FPS/W on top of the 5070’s 1.0 result. The 7900 XT is down at 0.75 FPS/W and is still pulling about the same power as previously.
Efficiency: F1 24 4K
F1 24 at 4K and with ray tracing is up next. In this one, the 5070 produced 0.17 FPS/W, pulling close to TDP at 246W in order to produce its hardly playable framerate. The 5070 Ti ran at 0.20 FPS/W, with the 4080 ranking at the top for its balance of power and framerate. Note that bar size changes from numbers that look the same are valid -- it’s just from the hidden decimal places.
The 7900 XT ran at 0.12 FPS/W, a significant fall from the 5070’s result. The 9070 series has a lot of work to do here.
Efficiency: F1 24 1080p
At 1080p and still with RT, the 5070 ran at 0.55 FPS/W, again giving the 5070 Ti a 13% efficiency advantage. We’re curious to see where the 9070 and 9070 XT land. The 7900 XT isn’t competitive in efficiency, with a lot of this particular result being because of its relatively low RT performance last generation. That’s what AMD is trying to tackle.
Efficiency: Black Myth 1080p
In Black Myth without RT and at 1080p, the 5070 Ti held a 0.53 FPS/W rank, putting it in the second slot. The 5070 is at 0.48, with the 4060 Ti still trading back-and-forth depending on the test. The 7800 XT was at 0.31 FPS/W, with the 7900 XT just below that. The 5070 is not NVIDIA’s most power efficient card due to the performance trade-offs, but is relatively efficient overall.
Efficiency: Dragon’s Dogma 2 RT 1440p
In Dragon’s Dogma 2 at 1440p and with RT, the RTX 5070 ranked at 0.36 FPS/W, sandwiching it between the 4090 that it’s not better than and the 5090. The 7900 XT is down at 0.28 FPS/W, yielding an efficiency benefit in a non-normalized framerate workload of 29%. That’ll be the mark for the 9070 series to hit.
Efficiency: Starfield 1440p
Finally, in Starfield rasterized at 1440p, the 5070 ran at 0.43 FPS/W and landed just below the 4090, which it remains not better than, again. The 7800 XT (read our review) from AMD ran at 0.32 FPS/W, so AMD has a lot of ground to gain here tomorrow. This will be an area we’ll focus to test for improvements, as theoretically, it should be better than it was.
RTX 5070 Conclusion
Visit our
Patreon page to contribute a few dollars toward this website's operation (or consider a
direct donation or buying something from our
GN Store!) Additionally, when you purchase through links to retailers on our site, we may earn a small affiliate commission.
One thing’s for certain: NVIDIA’s marketing about the 5070 being a 4090 was wrapped entirely in bull**** and built on a foundation of manure, which is only fitting for a company whose CEO is never found without a leather jacket.
The GPU market is insane right now. The fact that we sorted Newegg for $450 to $650 GPUs and got basically one is insane. The current buying experience matches that of the COVID-era boom around late 2020 and the 2017 crypto mining boom. That makes it hard to evaluate value, and broadly speaking, we’d say to just wait for things to cool off if your current machine can last you a bit longer. You’ll likely save money as pricing settles. If that doesn’t matter to you, maybe time will -- and unless you get lucky and snipe an early stock of the cards, you may at least save some time.
As for its value, we’re going to kick the can to tomorrow’s reviews of the 9070 XT and 9070. We don’t think you should buy this card until you learn about AMD’s competition, as the on-paper price is in similar territory. We’ll see how the in-reality price is for both of them.
Check back for the 9070 XT and 9070 reviews.
Connexion
