Pros & Cons
-
- Potent AI performance
- Competitive, though not triumphant, gaming speeds at its price point
-
- Increased power consumption over closest Nvidia predecessor (RTX 4070 Super)
- Generally slower than like-priced AMD competition
- Negligible performance gains over RTX 4070 Super
Nvidia GeForce RTX 5070 Founders Edition Specs
| Board Power or TDP | 250 |
| Card Length | 9.5 |
| Card Width | double |
| DisplayPort Outputs | 3 |
| GPU Base Clock | 2330 |
| GPU Boost Clock | 2512 |
| Graphics Memory Amount | 12 |
| Graphics Memory Type | GDDR7 |
| Graphics Processor | Nvidia GB205 |
| HDMI Outputs | 1 |
| Number of Fans | 2 |
| Power Connector(s) | 12VHPWR |
Here in early 2025, the field of midrange graphics cards is getting a giant reset—all at once. Coming in at a suggested selling price of $549, the GeForce RTX 5070 is Nvidia’s middle-market competitor set to do battle with AMD's $549 Radeon RX 9070 and $599 Radeon RX 9070 XT. We've thoroughly tested a Founders Edition RTX 5070 card to see how it compares with those cards and its predecessors. The competition in this segment is fierce, with several last-generation graphics cards still in the mix. While you can undoubtedly draw an enjoyable gaming experience out of the RTX 5070, one option clearly edges out the others at this price for high-detail 1440p and light 4K gaming: the Editors'-Choice-award-winning Radeon RX 9070 XT. As for Nvidia's claims that the RTX 5070 can reach GeForce RTX 4090 performance levels, our testing here does not bode well for that comparison. (More about that at the conclusion of this article.)
A note before we begin: This is a wildly busy time in the consumer GPU world. Because the midrange Radeon RX 9070, Radeon RX 9070 XT, and GeForce RTX 5070 have launched within a two-day span this week, we're issuing our reviews of these three cards one by one as we complete them. But rest assured: All three GPUs have been thoroughly tested, and all have been factored into the charts and discussion here.
The Blackwell GB205 GPU: Nvidia's Smallest RTX 50-Series Chip Yet
The GeForce RTX 5070 is the first graphics card based on Nvidia’s GB205 graphics chip, which uses Nvidia’s Blackwell graphics architecture. Of all the RTX 50-series graphics processors, the GB205 is the smallest so far, with a die surface area of 263mm2 and a transistor count of 31.1 billion.
In terms of hardware resources, the Nvidia GeForce RTX 5070 has 6,144 CUDA cores, 192 texture mapping units (TMUs), 80 raster operation processors (ROPs), and 48 ray-tracing (RT) cores. These connect up to 12GB of GDDR7 video memory over a 192-bit bus. These figures represent only a modest increase in hardware resources compared with the Nvidia GeForce RTX 4070. Power consumption increased more than anything else, with the RTX 5070 rated to draw up to 250 watts, whereas the RTX 4070 topped out at 200W.
Design: An Nvidia Founders Edition, Through and Through
Nvidia's unmistakable Founders Edition brand identity makes this RTX 5070 look precisely like the Nvidia GeForce RTX 5080 Founders Edition and the GeForce RTX 5090 Founders Edition. This housing gives the card a premium look and feel, though as I will discuss later, it also leads to suboptimal thermal performance.
(Credit: Joseph Maldonado)To power this card, you will likely need to use a 12VHPWR to eight-pin PCIe power connector adapter, which Nvidia provides with the card. Officially, the 12VHPWR power connection is what the wider industry is supposed to be moving toward. However, it is still relatively new, and only Nvidia has seriously tried to adopt it, and that on its Founders Edition cards. Power supply manufacturers are also incorporating support for it, but it remains the exception rather than the rule.
(Credit: Joseph Maldonado)The RTX 5070 Founders Edition's rear I/O panel supports three DisplayPort ports and a single HDMI jack. This I/O arrangement differs from some partner versions of competing AMD cards, such as the Sapphire Pulse AMD Radeon RX 9070 and the Sapphire Pulse AMD Radeon RX 9070 XT, which both support two HDMI ports and two DisplayPort outputs each.
(Credit: Joseph Maldonado)Though DisplayPort is compatible with HDMI via a simple adapter, I prefer two of each. I use HDMI more often, and having a second HDMI port makes connecting a second display easier. Your preference may vary.
Note that although we tested a Founders Edition version of the RTX 5070 provided by Nvidia, this actual version of the card won't be available until late March. The only versions of the RTX 5070 on sale now are from Nvidia's board partners, which include Asus, Gigabyte, MSI, PNY, and Zotac. List prices may vary according to the thermal solutions those cards use, on a model-by-model basis.
Testing the Nvidia GeForce RTX 5070: Setup and Competition
We tested the Nvidia GeForce RTX 5070 using our 2025 graphics card testbed PC that features a Gigabyte X870E Aorus Master motherboard and an AMD Ryzen 9 9950X processor with a large 360mm water cooler. Two 16GB sticks of DDR5 are installed on the motherboard and configured to operate under a 6,000MHz AMD EXPO memory profile. Two Crucial 2TB PCIe 4.0 SSDs are also inside the system, with one dedicated to games and the other holding Windows 11 and all other software. To ensure plenty of headroom, we used a 1,500W Corsair power supply.
As mentioned earlier, the RTX 5070's main competitors are AMD’s two new Radeon RX 9000-series GPUs. The Radeon RX 9070 is priced at $549, just like the RTX 5070, while the RX 9070 XT costs $50 more at $599. The closest recent parallels to the RTX 5070 from Nvidia’s camp are the Nvidia GeForce RTX 4070 Ti Super and the Nvidia GeForce RTX 4070 Super. The RTX 4070 Super and the older, vanilla RTX 4070 are both close to the RTX 5070 in terms of pricing (at launch $599 and $549, respectively), but the standard RTX 4070 isn’t quite as fast in terms of performance.
Synthetic Benchmarks
Our suite of synthetic graphics tests via UL's 3DMark and Unigine's Superposition suggested a less-than-promising outcome from the GeForce RTX 5070. In 3DMark Port Royal, the RTX 5070 was slower than both of its AMD competitors and was about halfway between the Nvidia GeForce RTX 4070 Ti Super and the GeForce RTX 4070 Super. It performed about the same in 3DMark Solar Bay, though the RX 9070 here fell into a tie with the RTX 5070.
The rest of the 3DMark suite tested here, and Unigine’s Superposition tests, showed much the same relative results across the board. The RTX 5070 was either behind the RX 9070 or tied with it. The RTX 4070 Super generally lagged behind the RTX 5070, but the RTX 4070 Ti Super tended to be faster.
AI Text Generation Benchmarks
Measuring AI performance is challenging, as AI work is implemented and carried out in myriad different ways, and different frameworks can have a big effect on performance. We’ve been using UL’s Procyon AI Text Generation Test to gather information on AI performance; the test measures token throughput and first-token generation times with four common large language models (LLMs).
We have come to expect significant advancements in AI performance with each new generation of graphics cards. Sure enough, we see this from the GeForce RTX 5070, but less than you might have expected. The RTX 5070’s AI performance is roughly comparable with the RTX 4070 Ti Super's and just marginally ahead of the RTX 4070 Super in some of the LLM tests here.
Nvidia’s AI speeds across the LLMs were much faster than AMD’s. If you value AI performance as a consumer (for whatever reason), this might be one key reason to buy an RTX 5070 over an RX 9070. We wouldn't take these results as a blanket judgment on the card's AI performance, however; this is just one set of AI scenarios in a vast sea. Simply put, with these particular LLMs and the test settings, Nvidia comes out on top.
Content Creation Benchmarks
Content creation performance is another area where we didn't see much generational improvement from the RTX 5070. It performed nearly identical to the RTX 4070 Super in the Adobe Premiere Pro and Blender tests.
Performance here was again ahead of the AMD Radeon RX 9070, but not by a drastic amount in Adobe—not enough to be a compelling reason to recommend the RTX 5070 over the RX 9070. This overall impression could change when we test the RX 9070 in the Blender Benchmark's GPU-centric trial, but that test doesn’t work with that card at this time.
Similar Products
Our Current Picks for The Best Graphics Cards for 4K Gaming in 2026
Screen Optimization Benchmarks
Black Myth Wukong supports DLSS 3 and FSR 3, which we test on Nvidia and AMD cards, respectively, with the super resolution sampling set to 100%. We then rerun these tests with frame generation on for all cards to gauge how this alters the performance. Note: This test does not show performance for DLSS 4, FSR 4, or Mutli-Frame-Generation (MFG) technology. With DLSS 4, Nvidia adopted a new AI model for DLSS work, and AMD made several changes to create its FSR 4 technology, too. Crucially, these technologies can create more than one intermediary frame between each conventionally generated frame, though this has some trade-offs. If you're interested in learning more about DLSS 4, check this article, in which I examined DLSS 4 performance on the Nvidia GeForce RTX 5090 more closely.
When using Nvidia's DLSS and AMD's FSR techniques for resolution super scaling and frame generation, it’s best to compare cards from the same manufacturer against each other, as image quality and workload can differ between using DLSS or FSR. The game Black Myth Wukong supports DLSS 3 and FSR 3, which we test on Nvidia and AMD cards, respectively, with the super resolution sampling set to 100%. We then rerun this test with frame generation on for all cards to gauge how this alters the performance.
The TLDR: The Nvidia GeForce RTX 5070 did not impress here. It was slightly faster than the RTX 4070 Super, but it was still clearly behind the RTX 4070 Ti Super.
With the RTX 5070 priced at $549 compared with the RTX 4070 Super for $599, this at least suggests some minor generational improvements in performance per dollar. Regardless, these results so far are lackluster.
Ray-Traced Gaming Benchmarks
The GeForce RTX 5070’s performance doesn’t improve in the ray-traced games we tested, performing marginally better than the RTX 4070 Super in most games. The only place the RTX 5070 clearly beat out the RTX 4070 Super was in Far Cry 6 at 1440p. This was a rather hollow victory, though, as the RTX 4070 Super was then much faster in the game at 4K. You'll find some other minor wins for the RTX 5070 over the RTX 4070 Super, like in Returnal, but the gains here were tiny, with the RTX 5070’s main advantage against that card being its $50 lower price. And given today's price volatility for the RTX 50 series, that may not even be a real-world advantage.
Nvidia's RTX 5070 also struggled against AMD competition in many of these games. The Radeon RX 9070 XT was unquestionably faster in all titles and at all resolutions, while the RX 9070 was a closer competitor for the RTX 5070. The RX 9070 stood out as a bit faster in F1 2024 and had a more pronounced lead in Far Cry 6 and Call of Duty: Modern Warfare III.
For its part, the RTX 5070 was faster than the RX 9070 in Cyberpunk 2077, Avatar: Frontiers of Pandora, and Returnal, but only by a slim margin in each case.
Raster Gaming Benchmarks
With Nvidia's GeForce RTX 40 series and AMD's Radeon RX 7000 series, ray-tracing performance was a strength for Nvidia, while more traditional rasterized games were a strength for AMD. Oddly enough, this scenario nearly reversed with the RTX 5070 and its competitor, the RX 9070.
The RTX 5070 and the RX 9070 essentially tied in Shadow of the Tomb Raider, but the RTX 5070 was faster in Total War: Three Kingdoms. The AMD Radeon showed more strength overall in ray-tracing games, which is arguably more critical as ray tracing has become widely used in the gaming industry.
Power and Thermals
We use a Kill-A-Watt wall meter to collect power usage information from our graphics card testbed. Our testing showed that the Nvidia GeForce RTX 5070 consuming roughly the same amount of power as the RTX 4070 Super in content creation tests, but the RTX 5070 drew markedly more power in the gaming tests.
The RTX 5070 also pulled in more power than the AMD Radeon RX 9070 in the gaming tests, which suggests the RX 9070 to be the more efficient of the two. The RX 9070 also stayed cooler during the testing process than the RTX 5070, but this could be due to the Nvidia Founders Edition thermal solution used on this card being less effectual. Sapphire also equipped its RX 9070 with a dual-fan thermal solution, but it didn't suffer the same problem.
Verdict: An Nvidia Letdown, an AMD Beatdown
So far, the new generation of graphics cards has eight new core-GPU combatants, with two each from AMD and Intel, and now four from Nvidia. The GeForce RTX 5070 is the most disappointing of the lot, and Nvidia’s marketing around it has contributed to that disappointment.
(Credit: Joseph Maldonado)When Nvidia first announced the RTX 50 series at CES 2025, CEO Jensen Huang made an audacious statement about how the RTX 5070 would be able to match the RTX 4090 in gaming performance. Typically, with new graphics cards, we expect them to perform at least one tier better than their predecessors. For example, when the RTX 4060 was released, we expected it to perform on par with the RTX 3060 Ti, as that was the next tier up from the RTX 4060’s predecessor, the RTX 3060. It did this without issue and, in some places, exceeded that mark.
We sometimes see more significant generational gains than this, too, but the RTX 5070 competing on the RTX 4090's level would have been one of the largest generation gains in graphics-card performance ever. I was skeptical about this (and I was far from alone), expecting it to be thanks to the RTX 5070 using DLSS 4 and Multi-Frame Generation via upgraded AI hardware, while the RTX 4090 used DLSS 3. We haven’t done that specific testing yet, but based on the above results, I’m dubious that the RTX 5070’s DLSS 4 performance will even match the RTX 4090 with its DLSS off.
(Credit: Joseph Maldonado)In truth, we saw mostly marginal performance improvements for the RTX 5070 over its direct predecessor, the RTX 4070 Super, and these at the cost of higher power consumption. The RTX 4070 Ti Super remains clearly ahead of the RTX 5070, too, which means the RTX 5070 is disappointing in terms of the normally expected performance gains, not just the inflated RTX 5070-vs.-RTX 4090 performance.
Nvidia's GeForce RTX 5070 doesn’t impress versus the new-gen AMD competition, either. For $50 more at MSRP, the AMD Radeon RX 9070 XT is far faster in several tests and is an unquestionably better option for gamers. The RX 9070 often ties with the RTX 5070, and it takes the lead by a sizable margin in a few cases. The RTX 5070 only tallied up some minor leads, and only in a few instances.
So, unless you need the RTX 5070's better AI performance as a consumer (again, why, in a midrange card?), the slightly pricier Radeon RX 9070 XT is the better option, which is no small reason why it earned our Editors' Choice award. Unless you catch the RTX 5070 on a steep sale (unheard of for GPUs, these days), or the AMD options are all sold out, you can find a better graphics card than an RTX 5070 for 1440p play.