Breaking the Price Barrier: NVIDIA GeForce 5070 Ti Emerges as a Capable 4K Gaming Solution for $749

It’s hard to recall the days when buying video cards was straightforward, prior to the impact of cryptocurrency enthusiasts, supply chain challenges, and increased demand during the pandemic, which drove GPU prices well above their manufacturers’ suggested retail prices. Ideally, I want to convey that NVIDIA’s $549 RTX 5070 and $749 RTX 5070 Ti represent more balanced options compared to the $2,000 RTX 5090 and $1,000 5080. Unfortunately, card manufacturers and retailers have already inflated the prices of the RTX 5070 beyond those suggested retail values. Currently, our review model, the ASUS 5070 Ti Prime, is priced at $900 at Best Buy and $750 at Newegg (we’ll see how long these prices stick). Naturally, both retailers are currently out of stock.

While it’s impossible to predict the exact price of any RTX 5070 Ti, one thing is clear: these cards are poised to deliver excellent 4K performance at a fraction of the cost of the RTX 5080 and 5090. However, if you’re not in urgent need of an upgrade, it may be wise to wait a few months for prices and availability to stabilize.

ASUS

85100Expert Score

NVIDIA’s RTX 5070 Ti emerges as a formidable GPU for 4K gaming, largely due to DLSS 4 enhancements. It’s a substantial step up for players seeking a bit extra power and memory compared to the 5070.

Pros

  • Excellent 4K performance
  • DLSS 4’s multi-frame generation yields significant benefits
  • Remains cool during heavy use
Cons

  • Prices are highly variable
  • Availability is a significant issue
  • $750 remains steep for a mid-range GPU

$900 at Best Buy

Explore More Buying Options

$750 at Newegg

Evaluating its features and (anticipated) price point, the RTX 5070 Ti positions itself as the most well-rounded choice for performance and value within NVIDIA’s offerings. It boasts 8,960 CUDA cores and 16GB of GDDR7 VRAM—significantly fewer than the 10,752 CUDA cores found in the 5080—but both models share the same memory capacity. The more economical 5070 model has merely 12GB of VRAM, which might pose issues when gaming at 4K resolutions.

ADVERTISEMENT
Advertisement

Our ASUS 5070 Ti card doesn’t stand out much visually, featuring three fans, a plastic chassis, and a conventional heatsink architecture. Users have the option to select between performance and silent BIOS configurations, which simply adjusts the fan speed. Its design takes up 2.5 slots, making it compact enough for smaller cases, although I found it to be slightly larger than the RTX 5090 Founders Edition.

RTX 5090

RTX 5080

RTX 5070 Ti

RTX 5070

RTX 4090

Architecture

Blackwell

Blackwell

Blackwell

Blackwell

Lovelace

CUDA cores

21,760

10,752

8,960

6,144

16,384

AI TOPS

3,352

1,801

1,406

988

1,321

Tensor cores

5th Gen

5th Gen

5th Gen

5th Gen

4th Gen

RT cores

4th Gen

4th Gen

4th Gen

4th Gen

3rd Gen

VRAM

32 GB GDDR7

16 GB GDDR7

16 GB GDDR7

12 GB GDDR7

24 GB GDDR6X

Memory bandwidth

1,792 GB/sec

960 GB/sec

896 GB/sec

672 GB/sec

1,008 GB/sec

TGP

575W

360W

300W

250W

450W

The 5070 Ti fits seamlessly into a variety of gaming configurations without necessitating a power supply upgrade. Its peak consumption reaches 300 watts, in contrast to the 5080’s 360W and the 5090’s dizzying 575W. This means the 5070 Ti can comfortably operate with an 850W power supply, without needing to upgrade to a substantial 1,000W unit.

Devindra Hardawar for Engadget

However, what truly distinguishes this GPU is its complete support for multi-frame generation through DLSS 4, NVIDIA’s AI-powered upscaling solution. This feature enables the card to generate up to 3 AI-driven frames for every real-time frame rendered, allowing NVIDIA to assert that the 5070 can rival the performance of the $1,599 RTX 4090. While some may argue these additional frames are merely «fake» and serve to enhance benchmark results, my experiences with both the RTX 5070 Ti and 5090 have demonstrated that they contribute to a noticeably smoother gaming experience.

ADVERTISEMENT
Advertisement

Besides multi-frame generation, various DLSS 4 enhancements are also becoming available for older NVIDIA models. As I pointed out in my review of the 5090, «RTX 40 cards will benefit more from single-frame generation efficiency, while RTX 30 and 20 series will receive upgrades from the AI transformer models that enhance ray reconstruction (resulting in steadier ray tracing), Super Resolution (improved texture quality) and Deep Learning Anti-Aliasing (DLAA).»

Devindra Hardawar for Engadget

To start with, the RTX 5070 Ti outperforms the 4070 and 4070 Ti in most benchmark tests only slightly. Specifically, it is 17% faster than the 4070 Ti in the 3DMark Timespy Extreme benchmark and 21% quicker than the 4070 Ti Super in the Speedway benchmark. The advantage is even more negligible in raw computing and rendering tasks, where the 5070 Ti posted only an 8% increase over the 4070 Ti in the Geekbench 6 GPU test.

None

3DMark TimeSpy Extreme

Geekbench 6 GPU

Cyberpunk (4K RT Overdrive DLSS)

Blender

NVIDIA RTX 5070 Ti

12,675

238,417

153fps (4X frame gen)

7,365

NVIDIA RTX 5090

19,525

358,253

246fps (4X frame gen)

14,903

NVIDIA RTX 4070 Ti Super

11,366

220,722

75fps (1x frame gen)

7,342

NVIDIA RTX 4070

8,610

N/A

45fps (1x frame gen)

6,020

Nevertheless, actual gaming performance is more relevant than benchmark numbers. When playing titles that support DLSS 4, you’ll definitely observe enhanced performance. For example, Dragon Age: The Veilguard achieved a consistent 200fps in 4K with 4X multi-frame generation, ray tracing, and graphics settings maxed out. Meanwhile, on the 4070 Ti, I usually experienced between 90fps and 100fps under those same settings and DLSS 3.5’s single frame generation.

ADVERTISEMENT
Advertisement

I cannot claim that the game appeared to run at double the smoothness on my Alienware 32-inch QD-OLED display, but it certainly felt incredibly smooth throughout my testing period. There were no noticeable upscaling artifacts; everything felt authentic. Additionally, the RTX 5090 consistently reached 240fps in Dragon Age with the same graphics settings. My CPU, a Ryzen 9 7900X, may have held back performance slightly, but the 5070 Ti still delivered impressively close results while being a significantly more affordable option.

Cyberpunk 2077 performed exceptionally well in 4K ray tracing overdrive mode with multi-frame generation, averaging 150fps. Though it falls short of the RTX 5090’s astonishing 250fps performance, it’s still remarkable for a game known to challenge even the most robust systems. At 1440p, Cyberpunk achieved 230fps with those same settings, which nicely upscale to 4K displays.

In titles lacking DLSS 4, such as Halo Infinite, the 5070 Ti still delivered strong performance, averaging 140fps with maxed-out graphics and ray tracing. To compare, the 5090 averaged 180fps. For those equipped with a 240Hz 4K monitor, I’m confident that even the most demanding gamers would be satisfied with the 5070 Ti’s performance. Nevertheless, for those prioritizing framerate over resolution, you’ll also be well taken care of; I recorded 220fps in Halo Infinite at 1440p, and 320fps at 1080p.

The ASUS 5070 Ti generally idled between 30C and 35C and quickly reached up to 65C when under load. While its cooling system isn’t as advanced as that of the 5090 Founder’s card, it still managed to cool the card down to below 40C in roughly 15 seconds.

Devindra Hardawar for Engadget

In summary, the RTX 5070 Ti performed admirably across the board, and I didn’t miss the 5090 as much as I anticipated (excluding the bragging rights). I haven’t had the opportunity to try the RTX 5080, but its high price tag makes it hard to advocate for.

ADVERTISEMENT
Advertisement

The most critical consideration for gamers currently is: Do you truly require the RTX 5070 Ti’s 16GB of VRAM and enhanced CUDA core count? If you primarily play games at 4K, having more than the 5070’s 12GB of memory is advantageous. With games evolving in complexity yearly, it’s probably not long until 16GB of VRAM will be necessary for comfortable 4K gaming. Conversely, for those who game at 1440p, 12GB should suffice for years ahead.

The multi-frame generation offered by DLSS 4 is the standout feature of NVIDIA’s 50-series cards, particularly for 4K gaming. So, if you’re satisfied with your current 40-series GPU and don’t aim to push a 4K 240Hz monitor to its limits, there’s limited motivation to upgrade. However, for owners of 30- and 20-series GPUs, your patience will pay off.

As previously mentioned, it’s advisable to wait a few months to see how prices adjust. If luck is on your side and you can snag the RTX 5070 Ti for $750, seize the opportunity. However, at $900 or above, it loses its appeal. At that point, you are approaching the $1,000 MSRP for the RTX 5090.

We are still in anticipation of AMD’s upcoming RDNA 4 Radeon 9070 and 9070 XT GPUs, which are intended to compete directly with the 5070 and 5070 Ti. Additionally, AMD is set to introduce DLSS-style AI-driven upscaling in the near future, making the differences between their offerings and NVIDIA’s a bit less pronounced. Nonetheless, NVIDIA enjoys a significant head start, and it could take some time before AMD’s Fluid Motion Frames technology catches up in multi-frame generation.

Devindra Hardawar for Engadget

The RTX 5070 Ti has exceeded my expectations in ways I didn’t foresee. While I anticipated it would be slightly quicker than the 4070 Ti Super, its ability to handle multi-frame generation has made it a far superior option for 4K gaming. Plus, it’s definitely more future-proofed than the 5070, as it shares the same 16GB of VRAM as the 5080.

Although I believe that the $549 5070 is the most captivating offering in NVIDIA’s latest series, it’s reassuring to see a viable option for discerning gamers nestled between that and the $1,000 5090. And indeed, it’s somewhat odd to designate a $750 graphics card as «sensible.»