Now that Nvidia’s top-end GeForce RTX 50 series graphics cards have landed, ready to take on serious 4K gaming, it's high time for a deeper-value option for the 1440p crowd. Nvidia built its GeForce RTX 5070 Ti from the same 4nm GB203 GPU as the Nvidia GeForce RTX 5080 and pairs it with similar features, like 16GB of GDDR7 memory, but fewer silicon resources. At $749 for base-model cards, the GeForce RTX 5070 Ti looks to be a better value than the $999 RTX 5080: It’s roughly 11% slower at best and 21% slower at worst, but for 25% less cash at MSRP. The actual street prices of these RTX 50-series cards versus their list prices are the big question mark, though; whether you'll be able to get one easily at $749 remains to be seen. (Cards go on sale Feb. 20.)

We tested the RTX 5070 Ti in a three-fan Asus Prime GeForce RTX 5070 Ti, a $749-MSRP card that represents the GPU at a decent baseline with room for overclocking. In our tests, it's behind the RTX 5080, but overclocking, if you're intrepid, could make up some of the difference in performance. For that, we give the Nvidia GeForce RTX 5070 Ti our Editors' Choice award for high-end graphics cards; the RTX 5080 retains the award for "elite level" cards.

The Nvidia Blackwell GB203 GPU: Strategically Reduced

To create the Nvidia GeForce RTX 5070 Ti, Nvidia uses an old yet common strategy known as "binning," which sees chipmakers repurposing slightly under-spec silicon in lower-grade, lower-priced products. Nvidia didn’t do this much initially with the Nvidia GeForce RTX 40-series "Ada Lovelace" GPUs, which was surprising for various reasons. That changed with the launch of the RTX 40-series Super graphics cards, and with the RTX 50-series Blackwell parts, you can now see this strategy deployed from the beginning.

(Credit: Joseph Maldonado)

Essentially, this works as follows. Nvidia starts by creating large graphics chips, like the GB203, which powers the Nvidia GeForce RTX 5080, in that case with 10,752 CUDA cores, 336 texture mapping units (TMUs), 112 raster operation processors (ROPs), and 84 ray-tracing (RT) cores. TSMC provides the graphics die on its 4N 4nm fabrication process, and it is relatively large, covering an area of 378mm2 and containing 45.6 billion transistors.

(Credit: Joseph Maldonado)

Creating computer chips is not a perfect process, though. Inevitably, minor defects creep in, rendering some of the hardware resources on a fabricated chip unusable. Simply tossing these chips out would be wasteful, but stockpiling them is also costly, making it surprising that Nvidia didn’t bin the RTX 40 series from the start.

(Credit: Joseph Maldonado)

Binning deactivates the circuitry that isn’t functioning correctly, creating a lesser product with fewer total resources available that the chipmaker can charge less for at retail. For instance, the Nvidia GeForce RTX 5070 Ti is also based on the GB203 GPU die, but with just 8,960 CUDA cores, 280 TMUs, 96 ROPs, and 70 RT cores active.

The GeForce RTX 5070 Ti is also clocked a fair bit slower than the RTX 5080 in terms of its turbo clock, reducing performance. The memory side of the chip remains mostly unchanged, though, which is promising; that frequently gets reduced when creating cut-down cards like this. The RTX 5070 Ti and the RTX 5080 both have 16GB of GDDR7 memory connecting to the GPU over a 256-bit wide bus. The RTX 5070 Ti has slower 28Gbps GDDR7, versus the RTX 5080's 30Gbps GDDR7, reducing bandwidth.

(Credit: Joseph Maldonado)

The 20% reduction in core count separates the RTX 5070 Ti from the RTX 5080. However, the roughly 6% reduction in turbo clocks and the 7% reduction in memory bandwidth also play a key role in separating the two. These disparities in hardware match up well with the $250 price difference, but real-world performance could set the RTX 5070 Ti closer or further from the RTX 5080. We'll see below.

This design also allows Nvidia's GeForce RTX 5070 Ti to have considerable overclocking potential. The GB203 works at higher clock speeds in the RTX 5080, suggesting that the RTX 5070 Ti could also run faster when pushed. The RTX 5070 Ti's reduced clocks are likely in place to distance it from the RTX 5080, rather than due to any limitation of the chip itself.

Design: A Look at the Asus Prime GeForce RTX 5070 Ti 16GB GDDR7

Asus' Prime GeForce RTX 5070 Ti, sent for review, features a large triple-fan thermal solution, predominantly black with some white stripes. The card has a metal backplate with a cutout showing the aluminum fins of the heatsink and several metal heat pipes that help with the cooling process. This cutout also allows air to pass through so that one of the three fans on the front of the card can blow air straight through the heatsink and the card.

(Credit: Joseph Maldonado)

The card extends out to fill 2.5 slots on the rear I/O panel, with three DisplayPort ports and a single HDMI port. It also requires a 12VHPWR power connection to operate, but Asus includes an adapter that converts three eight-pin PCIe power connectors to one 12VHPWR connection to work with conventional power supplies. (Nvidia suggests a minimum 750-watt power supply to run the RTX 5070 Ti.)

(Credit: Joseph Maldonado)

Testing the Nvidia GeForce RTX 5070 Ti: Setup and Competition

The graphics card testbed that we used to test the Asus Prime GeForce RTX 5070 Ti 16GB GDDR7 is built around a Gigabyte X870E Aorus Master motherboard with an AMD Ryzen 9 9950X processor paired with a Cooler Master 360mm Cooler Master water cooler. Two Crucial M.2 PCIe 2TB SSDs and two 16GB sticks of DDR5 RAM clocked at 6,000MHz and set in a dual-channel configuration sort out the storage and memory. Finally, a 1,500-watt Corsair power supply powers the entire system, which is housed in an Asus ProArt chassis.

The primary competition for the RTX 5070 Ti comes from the Nvidia GeForce RTX 4080 and the AMD Radeon RX 7900 XTX. Though both cards are pricier, they are also older cards that are now going to have a hard time justifying their $999 price tags. Both are close to the RTX 5070 Ti in terms of performance, too. (The Nvidia GeForce RTX 4070 Ti Super and the AMD Radeon RX 7900 XT, at their list prices, are more competitively priced than the RTX 5070 Ti at $749.)

Synthetic Benchmarks

The 3DMark results we gathered from the RTX 5070 Ti suggest it will slightly outperform or perform on par with the RTX 4080. Based on the 3DMark scores, it also has the potential to surpass all of the AMD graphics cards, though the Radeon RX 7900 XTX pulled ahead in a few of the 3DMark tests.

The RTX 5070 Ti and the RTX 4080 performed similarly in the Unigine Superposition test, which also suggests they might perform similarly in games. We'll see in a moment how that shakes out.

Content Creation Benchmarks

Nvidia's GeForce RTX 5070 Ti is quite capable of crunching on content creation tasks. It was essentially tied with the RTX 4070 Ti Super and the RTX 4080 in our Adobe Premiere Pro test. At the same time, it placed a notable step ahead of the AMD cards that we tested.

The RTX 4080 also outperformed the RTX 5070 Ti in the GPU-centric trial that we ran in Blender, and it did so by a more meaningful margin specifically in the Monster model (one of the three renders that the test performs). The two cards were more evenly matched in Junkshop and Classroom. Nevertheless, you might be better off sticking with an RTX 4080, if you have one, if you intend to use your graphics card more for content creation work than gaming.

Procyon AI Text Generation Benchmarks

Improving AI performance has been a core goal for Nvidia with each new generation of graphics cards, but the extent of the upgrade varies heavily depending on the task at hand. (Testing AI performance in any holistic sense is difficult, particularly as on-card AI muscle addresses a wide range of tasks--training versus inferencing, for one thing. It is impossible with current tools to test a very broad set of AI local-processing scenarios in an all-encompassing way.)

This test presses the GPU on a series of text-creation inferencing tasks via four popular AI large language models (LLMs) and produces an overall score in each case. (Here, Procyon uses versions of Mistral, Microsoft's PHI, and two flavors of Meta AI's Llama.) It also reports how many tokens (or discrete units of text, in the context of text generation) the hardware can produce per second with each model. It also measures the time it takes to produce the first token, as getting as much work done as fast as possible is basically AI’s mission statement.

Generally speaking, the RTX 5070 Ti topped the RTX 4070 Ti Super and the RTX 4080...

The RTX 5070 Ti outpaced the 4080 by varying amounts depending on the test, with the most considerable performance advantage when using the Mistral 7B model. It is faster when using other models, but to a lesser degree. With that, Nvidia has achieved measurable generational gains in AI performance, which will become more relevant as the technology becomes more applicable to everyday users.

Screen Optimization Benchmarks

Black Myth Wukong's benchmark utility supports Nvidia DLSS 3 and AMD FSR 3, with frame generation supported for both. It also supports Intel XeSS, making it ideal for testing these screen-optimization technologies against each other. (As a refresher: frame generation uses AI to insert predictive frames between traditionally rendered ones.)

This is a highly demanding title that pushes even the fastest graphics cards to their limit, making frame generation and resolution scaling all but necessary for smooth frame rates. The RTX 5070 Ti couldn't quite keep up with the RTX 5080 in this benchmark test, but it landed in a dead heat with the RTX 4080.

Activating DLSS frame generation enabled the GeForce RTX 5070 Ti to slightly outperform the RTX 4080, but enabling FSR's own frame generation also helped the Radeon RX 7900 XTX outperform the RTX 5070 Ti at 1080p and 1440p.

Ray-Traced Game Benchmarks

The next set of in-game benchmarks more accurately reflects the performance of the graphics cards we test without AI-driven assistance. The results of the synthetic tests suggested that the Nvidia GeForce RTX 5070 Ti should perform about on par with the RTX 4080. In most cases, this is also what the in-game benchmarks showed, but at times, the RTX 5070 Ti lagged behind the older RTX 4080. Still, 1080p and 1440p frame rates were very respectable across the board, even with the killer Cyberpunk bench.

The tests in which the RTX 5070 Ti and the RTX 4080 performed pretty close to each other were Cyberpunk 2077, F1 2024, Far Cry 6, Call of Duty Modern Warfare III, and Returnal. We saw several near-ties between those two graphics cards in these games, but whenever we didn't see a tie in one of these games, the RTX 4080 tended to be on top.

In Cyberpunk 2077, for example, the RTX 4080 was 5% faster than the RTX 5070 Ti at 1080p, but the two cards tied at 1440p and 4K. The test scores from F1 2024 showed the same relative results, with the RTX 4080 again slightly ahead at 1080p but tied at other resolutions. The only game that showed the RTX 5070 Ti clearly behind the RTX 4080 without ties was Avatar: Frontiers of Pandora; the RTX 4080 was about 5% faster on average across all three resolutions.

Taking a brief look at the RTX 5070 Ti’s performance relative to the RTX 4070 Ti Super or the Radeon RX 7900 XTX, we see both cards are clearly behind the RTX 5070 Ti. At least in AMD's case, this is because Radeon RX 7000-series cards simply have less capable ray-tracing hardware than the Nvidia competition, which holds them back in these tests that support ray tracing. (The RTX 4070 Ti Super also has fewer RT cores.)

Nvidia’s RTX 5080 is no doubt faster than the RTX 5070 Ti, but the degree to which the RTX 5080 beats the RTX 5070 Ti is interesting. We’ll spare you a blow-by-blow of percentage differences between these two cards in each game and at each resolution, but overall, the test results showed the RTX 5080 with an 11%-to-21% advantage over the RTX 5070 Ti. That excludes Far Cry 6, which showed signs of a bottleneck by either the processor or the game engine, putting the two cards even closer together. The RTX 5070 Ti costs 25% less, so it's a compelling value.

Raster-Only Game Benchmarks

Turning to games that don’t support ray tracing, AMD's Radeon RX 7900 XTX could more competently compete with Nvidia's GeForce RTX 5070 Ti. The RX 7900 XTX didn't pull ahead here except in the Total War: Three Kingdoms at 1440p, but it could at least tie with the RTX 5070 Ti in most tests.

One interesting thing these games that lack ray-tracing support show us is actually related to ray-tracing performance. In ray-traced games, the RTX 5070 Ti tended to tie with or lag behind the RTX 4080. This is due to the former's smaller count of RT cores, comparatively speaking, despite architectural improvements with Blackwell.

In games that don’t support ray-tracing, the RTX 5070 Ti held a clear advantage at all resolutions. This reflects Nvidia's claimed architectural improvements to its Blackwell CUDA cores, TMUs, and ROPs over the RTX 40-series Lovelace parts. This explains why the RTX 5070 Ti pulled ahead in these pure rasterization-based game tests. At the same time, however, it also shows the ray-tracing hardware of the RTX 5070 Ti isn’t quite as capable as that found on the RTX 4080, which is why the RTX 4080 was able to pull ahead in those games that do support ray-tracing.

Though the Nvidia GeForce RTX 5070 Ti stole a win here from the RTX 4080, consider it a hollow victory. Total War: Three Kingdoms and Shadow of the Tomb Raider are both 2018 games, which we continue to use because they are some of the last games to feature in-game benchmarks that don’t support ray tracing. Ray tracing is now nearly universal for major AAA titles, so you likely will only enjoy this advantage on the RTX 5070 Ti in older games.

Power Consumption Benchmarks

We measure the power consumption of our graphics card testbed as a whole using a Kill-A-Watt wall meter. This gives us insight into the power draw of each card tested, as the card is the only part that changes between tests. All of the Nvidia GeForce RTX 50-series cards to date have been more power-hungry than their direct predecessors, and the same is true for the RTX 5070 Ti. It doesn’t use that much more than the RTX 4070 Ti Super, only about 5% more at most, according to the test results we gathered.

The Asus Prime GeForce RTX 5070 Ti 16GB GDDR7 model we tested also seemed to stay reasonably cool during our tests. It did run a bit warmer than the RTX 4070 Ti Super we tested and a bit warmer than the RTX 4080, but running a bit hotter is also a trend with the RTX 50 series. However, the Prime GeForce RTX 5070 Ti 16GB GDDR7 was cooler than the RTX 5080 Founders Edition during gaming, and its temperatures were well within the typical safe range for graphics cards.

Verdict: A Compelling Card...If the Pricing Holds

The Nvidia GeForce RTX 5070 Ti is a competent high-end graphics card dunked into a soon-to-be highly competitive market segment. From a value perspective, this might be Nvidia's most attractive offering in its GeForce RTX 50-series family of graphics cards so far, thus earning it our Editors' Choice award. It doesn’t cost $2,000 like the Nvidia GeForce RTX 5090, and it costs 25% less at MSRP than the RTX 5080 with an 11%-to-21% drop in speeds, adding up to a better overall value. Better yet, if you're willing to push the card, overclocking could compensate for some of what you lose in performance relative to the RTX 5080.

What might be more challenging to overcome for the RTX 5070 Ti is the older Nvidia GeForce RTX 4080 Super. We didn’t retest the RTX 4080 Super for this review, but look at the RTX 4080 as a proxy: The RTX 4080 Super is generally about 5% faster than the RTX 4080 and priced at $999 MSRP. That price point for that card isn’t tenable anymore, with the RTX 5080 also at $999, so price drops could set the RTX 4080 Super price closer to that of the RTX 5070 Ti, if you can find one. That would make the RTX 4080 Super the better option, as it also has slightly better performance, but that all depends on what the price will fall to--if, indeed, it even does. That’s difficult to tell for now, as the RTX 4080 Super is sold out or priced above MSRP everywhere we can find it online.

(Credit: Joseph Maldonado)

The RTX 5070 Ti clearly has the upper hand against AMD’s competing RX 7000-series graphics cards. The Radeon RX 7900 XTX initially launched at $999 and has since been reduced to $899, making it more expensive than the RTX 5070 Ti, which outperforms it.

As the market stands, the Nvidia GeForce RTX 5070 Ti is an excellent option for anyone looking to avoid paying out for an RTX 5080 but still get high-end 1400p and decent 4K gaming performance. The lack of in-stock competition also makes it a dominant choice in the market at publish time, but just be sure to evaluate against what is in stock when shopping. (The RTX 5070 Ti may well see its own in-stock challenges, like the first two RTX 50 GPUs have.) For the right price, an RTX 4080 Super could be more appealing. Also know that AMD has new cards coming soon, on an announced Q1 2025 timeline, and they could well be right in the RTX 5070 Ti's wheelhouse. Stay tuned, and if you can't find the card readily on sale in the weeks after launch, be patient; more options from AMD and Nvidia alike are coming soon.