We believed it. When Nvidia unveiled the RTX 5070 Ti, everything suggested that we finally had a card capable of reconciling high-end performance with a reasonable price. A sort of perfect balance between raw power, AI-driven innovations, and affordability. On paper, it was a masterstroke.
But spec sheets sometimes tend to embellish the truth. After several days of battering it in our benchmarks, pushing it to its limits on the latest AAA games, and then scrutinizing its thermal and energy performance, the verdict has been refined. Yes, this RTX 5070 Ti has what it takes. Yes, its DLSS 4 and AI-based upscaling capabilities are impressive. But not everything is as simple as hoped.
Behind the promises of record-breaking performance lies a card that's both fascinating... and frustrating. A concentration of technology that alternates between moments of grace and minor disappointments, often linked to its disastrous availability and a positioning that places it in a gray area of Nvidia's catalog. An emotional rollercoaster, as much for gamers as for testers.
So, is the RTX 5070 Ti graphics card really the card of the year or just a well-marketed mirage? We tell you everything, unfiltered, as if you were right there with us during these long hours of testing.
Table of Contents
A generation that changes the codes
As soon as you take the RTX 5070 Ti out of its box, the first surprise jumps out at you: there's no Founders Edition on the horizon. Nvidia has chosen to leave the field open to its partners, entrusting third-party manufacturers with the task of bringing this new generation to life. And on this point, the result is as varied as it is spectacular.
During our testing, several models were used on our test bench: the MSI Ventus, the Palit GameRock, and the MSI Vanguard. Three interpretations of the same GPU, three cooling philosophies. However, they all share one constant: an impressive dissipation system, almost oversized for the chip they house. It sometimes feels like you're handling a card designed for a higher-end GPU.
The Palit GameRock, in particular, left us speechless. Its shroud is identical to that of the RTX 5080 and 5090: a massive radiator, three oversized fans, a meticulously finished design, and RGB that's both discreet and elegant. The whole thing exudes quality. The downside is its considerable weight, which requires using the bracket included in the box to avoid twisting the PCIe port. Palit even goes so far as to provide an aRGB cable to synchronize the lighting with the rest of the setup—a small detail, certainly, but one that demonstrates a real concern for harmony for high-end configurations.
One question that stayed with us throughout the test remains: do you really need such a cooling arsenal for a 300-watt card? At this point, it's difficult to decide, but we feel that Nvidia wanted to impose a new design standard, even if it meant going a little overboard.
Another curiosity, more technical this time: the power connector. The RTX 5070 Ti adopts the same 12VHPWR with three times eight pins as the RTX 5080, while its TDP is limited to 300 watts, barely 15 watts more than the RTX 4070 Ti Super. From a purely electrical point of view, two connectors would have been more than enough. This choice is therefore not dictated by necessity, but rather by a logic of standardization, or perhaps a safety margin designed for overclocked models. A small detail, but revealing of the direction this generation is taking.
Features that leave you perplexed
Delving into the RTX 5070 Ti's spec sheet, a slight unease sets in. The numbers don't lie, but they're not exactly eye-catching either. Nvidia calls it a new product, but upon closer inspection, it feels more like a tentative, almost cautious, evolution.
The card boasts 8,960 CUDA cores, compared to 8,448 on the RTX 4070 Ti Super. That's barely 5% better, which is to say an insignificant margin at this level of the range. The memory does indeed upgrade to GDDR7, which is faster on paper, but the quantity remains fixed at 16 GB. The memory bus, however, hasn't changed one iota: it's still 256 bits. Nothing revolutionary here, just minor adjustments.
And that's the paradox of this generation. The RTX 5070 Ti looks like a new model, but its DNA betrays a clear kinship with its predecessor. In fact, it looks more like a slightly boosted 4070 Ti Super than a true architectural overhaul. It's almost as if Nvidia had simply turned a few sliders upwards to convey the message of a "new" card.
This feeling is reinforced when you remember that the 4070 Ti Super itself was already just a refresh of the original 4070 Ti, which capped at 12GB of VRAM. Three versions of the same technological base in less than two years: the brand now seems to prefer refining the existing rather than upending the market hierarchy.
For those who lived through the heyday when each Nvidia generation brought a 20-30% jump in performance, the contrast is striking. It's no longer a race for raw power, but rather a meticulous management of the ecosystem, almost a long-term optimization strategy. And for gamers, the question arises: should we still talk about a "next generation" when the gaps are becoming so narrow?
ChatGPT said:
Performance at stake: between promises kept and false hopes
This is where it all comes down to. A graphics card can shine with its silence and controlled temperature, but it's on the field, controller or mouse in hand, that it reveals its true nature. So we subjected the RTX 5070 Ti to a series of demanding tests, covering everything from recent blockbusters to the most capricious engines, in 1440p and 4K. And the result, while still solid, doesn't trigger the enthusiasm we'd hoped for.
On Cyberpunk 2077 , with high settings, ray tracing enabled, but without DLSS, the RTX 5070 Ti averages 96 FPS , while the RTX 4070 Ti Super tops out at 81 FPS . A jump of 18% , very decent. But it must be said: Cyberpunk is one of the best-optimized games for the Blackwell architecture. On other titles, the magic works much less.
Alan Wake 2 , known for bending many GPUs, shows 24% gain in 1440p and 27% in 4K , which is still impressive. On Black Myth: Wukong , the gap narrows, 17% in 1440p , 27% in 4K . Jedi Survivor remains in the average, 20% and 25% respectively. Indiana Jones and the Great Circle , more recent but less optimized, does not take off beyond 16% in 1440p and 20% in 4K . Conversely, The Witcher 3 takes full advantage of the improvements, with 30% better in 1440p and 35% in 4K, proof that older engines can still benefit from raw power.
On average, across all our tests, the RTX 5070 Ti offers a 22% gain in 1440p and 27% in 4K compared to the 4070 Ti Super. Enough to stay at the top of the range, but far from the 30% promised by Nvidia . And above all, below what we expected from a generational leap. It's not a revolution, but a controlled evolution.
Synthetic benchmarks confirm this observation. 3DMark Time Spy displays 27,862 points , compared to 23,943 for the 4070 Ti Super, a difference of 16% Time Spy Extreme , we climb to 25.5% , while Port Royal , oriented towards ray tracing, shows a more spectacular progression of 39% , a direct reflection of the new Blackwell RT cores.
The RTX 5070 Ti graphics card does the job, and sometimes even better than expected. But it struggles to create the “wow” effect we usually associate with a new generation of Nvidia. It's not a revolutionary monster, it's an efficient, disciplined card, almost too tame for its own good.
The real surprise: a disguised RTX 4080
It's when we dig a little deeper into our tests that the RTX 5070 Ti reveals its true nature. And there, surprise: the numbers don't lie, but they are unsettling. In 4K, the performance is almost identical to that of an RTX 4080. The difference? Less than 1% . In 1440p, it barely climbs to 3% . In other words, in the majority of cases, we literally play on the same level of fluidity. On Hogwarts Legacy or Indiana Jones and the Great Circle , the results are sometimes within a tenth of an FPS .
This observation changes everything. We thought we were testing a card that was supposed to be placed just below the RTX 5080, but we end up with a rebadged RTX 4080 , more compact, more sober, and above all much less expensive. It is no longer a competitor of the high-end, it is almost a disguised substitution. And this is the whole paradox of Nvidia: offering an incredibly powerful product for its price, while completely blurring the usual benchmarks of the range.
Because while the RTX 5070 Ti is on par with the 4080, it is only behind the RTX 5080 on average . A paltry difference when you consider that 300 euros . The performance/price ratio therefore clearly shifts in favor of the little sister, which de facto becomes the best compromise in the Blackwell catalog .
But this convergence between generations is not without consequences. When a so-called "70 Ti" card equals an "80" card from a few months ago, the entire balance of the range is disrupted. Nvidia plays a subtle, perhaps overly subtle, role: attracting gamers with tangible gains while maintaining the commercial hierarchy. And in the end, it's the informed consumer who comes out on top, at least as long as prices remain under control.
The confrontation with AMD: a duel at the top
The comparison was bound to happen. Up against the Radeon RX 7900 XTX , AMD's card that's supposed to represent the best power-to-price ratio at the moment, the RTX 5070 Ti finds itself in a battle of equals. And for once, the match is of rare intensity.
In raw performance , it's hard to separate the two rivals. In 1440p as in 4K, the gap never exceeds 2% , a margin so thin that it is almost a statistical coincidence. On some titles, the balance tips towards Nvidia, on others, AMD takes back the lead. Rainbow Six Siege perfectly illustrates this alternation: 321 FPS for the RTX 5070 Ti , 268 FPS for the Radeon , a clear gap... but reversed on Forspoken or Avatar: Frontiers of Pandora , where the RX 7900 XTX takes back the upper hand. In short, it's a toss-up in each benchmark.
But the real difference isn't in raw power. It lies in two areas where Nvidia still holds the advantage. The first is ray tracing , where its RT Cores continue to widen the gap. Reflections, shadows, and lighting effects display superior precision and stability, especially at higher resolutions. While AMD may be making progress, Nvidia's lead in this area remains tangible.
The second point is that of proprietary technologies . On one hand, DLSS 4 and its multi-frame generation which literally enhances performance while maintaining a stunningly sharp image. On the other, FSR 3 , more open but also more irregular depending on the title. And it is precisely here that Nvidia plays its best card . Its AI ecosystem is more mature, better integrated, and above all more coherent. The player has nothing to configure, everything works, and it is this invisible simplicity which often makes the difference once in the game.
DLSS 4: the long-awaited revolution?
This is where Nvidia plays its biggest card. DLSS 4 and its multi-frame generation aren't just an evolution of super-sampling: they're a demonstration of pure AI power. Enabling it literally sends the counters skyrocketing.
On Cyberpunk 2077 , in 4K, ray tracing enabled and DLSS 4 in performance mode, the RTX 5070 Ti smashes the limits with 252 FPS on average. Yes, 252 FPS on one of the most demanding games on the market. We spend a few seconds in disbelief, checking the numbers three times because they seem absurd. On Counter-Strike 2 , the results are supernatural: 410 FPS in 1440p, compared to 183 FPS for an RTX 4070 Ti Super equipped with the old frame generation. The same is true on Jedi Survivor , where we pass the 355 FPS , and on Shadow of the Tomb Raider , which climbs to 215 FPS in 4K. Performances that, just two years ago, seemed simply impossible.
But the feat doesn't stop at the numbers. Image quality has also reached a new level. After spending several dozen hours on Cyberpunk 2077 with DLSS 4 enabled, it's difficult to distinguish the native rendering from the generated rendering. The artifacts, once common, have virtually disappeared. Only a few occasional micro-glitches remind us that the image isn't entirely "real." The secret? A Transformer model boosted by deep learning, capable of reconstructing each image with almost photographic precision.
Yet, behind this debauchery of power, hides a disturbing paradox. Because if DLSS 4 multiplies the images on the screen, it does not reduce the input latency . In short, even at 252 FPS displayed , the response time remains that of the original frame rate. If the game ran at 60 FPS natively , the responsiveness will remain that of 60 FPS, even if the screen displays four times more. The visual result is stunning, but the mouse inputs do not gain in liveliness. It is a strange, almost disconcerting experience: the game seems smoother, but does not “respond” any faster.
Nvidia is banking on Reflex to compensate for this latency, and to some extent, it works. But Reflex already existed, and its effectiveness doesn't depend on DLSS 4. In other words, the next-generation AI upscaling isn't a game-changer for competitive gamers. For them, DLSS 4 remains a double-edged sword : spectacular to the eye, but potentially misleading in terms of the sensation of responsiveness.
Overclocking finally rewards
Good news for tuning enthusiasts: overclocking is finally becoming interesting again. After several generations of RTX cards hampered by overly tight TDPs and insignificant gains (2 to 3% at best), the RTX 5070 Ti marks a true return to flexibility.
The GDDR7 memory proves surprisingly docile: we were able to push it to +1700 MHz without the slightest instability. The GPU, for its part, takes +250 MHz without flinching. Result: +6% performance on average , with peaks of +10% on certain demanding titles like Horizon Forbidden West .
All this, with temperatures contained around 67 °C , proof of a comfortable thermal margin.
Tellingly, we reached the limit of MSI Afterburner's slider , and had to switch to Asus GPU Tweak to continue climbing. The gains were still measurable, a sign of real hidden potential.
One small drawback: on the MSI Ventus we tested, the TDP limit remains stuck at 100% , with no possibility of raising it to 105% or 110% as with some competitors. A surprising choice, because with an already generous TDP of 300W , the card could have offered a little more freedom to those who like to flirt with the limits.
Production performance: where the RTX 5070 Ti really makes the difference
While the in-game gains remain modest, it's in the production and creative work Blackwell architecture finally reveals its full potential. We've put the card through its paces in several professional software programs, and this time, the numbers speak for themselves.
On DaVinci Resolve , rendering times drop significantly compared to the previous generation. Blender takes full advantage of the new CUDA cores and the RT Cores : rendering under Cycles accelerates dramatically. The same is true for D5 Render , used in the world of architecture, where the fluidity of real-time rendering is impressive.
Even more surprising, local language models (LLM) also benefit from this new architecture, even if, on Windows, some tools are not yet fully optimized for the RTX 5000.
However, it's V-Ray that everyone agrees on: the gains are simply massive. For those who alternate between gaming and content creation, video editing, 3D rendering, and audio mixing with AI plugins, the RTX 5070 Ti offers a real and tangible leap.
Adobe Premiere Pro and DaVinci Resolve Radeon cards continue to lag behind.
Ultimately, this is perhaps the best selling point for those who already own an RTX 4000: professional use . In pure gaming, the progression remains modest. But for production, the RTX 5070 Ti stands out as a true workflow accelerator .
The launch drama: when price and stock betray the promise
This is where the dream begins to crack. On paper, the RTX 5070 Ti had everything going for it: an MSRP of €884.95 , or €100 below the Radeon RX 7900 XTX and on par with the RTX 4070 Ti Super. With performance equivalent to an RX 7900 XTX, and Nvidia technology to boot, the choice seemed obvious.
But in reality, the launch was a nightmare. On D-day, at 3 p.m. sharp , we were behind our screens frantically refreshing Materiel.net, TopAchat, LDLC and Cybertek. And then, the cold shower: zero stock in a fraction of a second . Impossible to add anything to the basket. The few cards available displayed crazy prices: 1100, 1200, 1300 euros , and some premium models even climbed up 1400 euros . Calculated, the real average price is around 1130 euros , or 245 euros more than the announced MSRP .
This inflation completely changes the situation. At 885 euros , the card was the best possible recommendation, a clear alternative to AMD. At 1130 euros , it becomes even more expensive than the RX 7900 XTX , which is slightly less equipped. The performance-price equation is reversed, and it is the gamers who suffer the consequences.
But who's responsible? The answer isn't simple. Nvidia sets the MSRP, manufacturers add their cooling system and adjust their pricing, and resellers apply their margins. Somewhere along the way, MSRP discipline is being lost, and it's the consumer who suffers. This scenario isn't new: the RTX 5080 suffered the same fate, starting at €1,180 but ending up in practice at around €1,530 . Nvidia continues to tout attractive prices that, in reality, are almost impossible to find.
The launch thus shows the fragility of a card which, technically brilliant, comes up against the merciless reality of the market and prices.
Who should really consider this card?
Putting aside the actual inventory and pricing drama , and the RTX 5070 Ti sits at its MSRP of €885 , then the question becomes: who does this card make sense for?
The first profile is the one who builds a new PC after several generations of GPU: RTX 2000, 3000 or equivalent. The leap is immediate and palpable: high-performance ray tracing , DLSS 4 , and performance capable of handling 4K in good conditions , or 1440p in ultra with more than 144 FPS constant . For these users, the card represents a real revolution compared to their previous generation.
Second profile: those looking for a versatile gaming + production solution . If your days alternate between Adobe Premiere, DaVinci Resolve , VST plugins leveraging AI, or any software that takes advantage of CUDA cores , the RTX 5070 Ti becomes a logical investment. Against AMD, it has a clear advantage in this area.
Third situation: players focused on recent AAA games compatible with DLSS 4 and multi-frame generation . Your Steam library filled with titles like Cyberpunk 2077 , Alan Wake 2 , Indiana Jones and other recent games will take full advantage of the technology, transforming the visual fluidity and perceived responsiveness of the gameplay.
On the other hand, for those who already own an RTX 4070 Ti Super, RTX 4080 or RTX 4070 Super , the upgrade is completely irrelevant. The gains are too modest to justify the additional investment. The same goes for an RX 7900 XT or XTX : these cards will continue to perform well for years to come.
Finally, for competitive gamers, those who prioritize pure framerate without ray tracing or DLSS, or for those playing older titles, the Radeon RX 7900 XTX retains an advantage. It remains solid, and above all, its current price of 979 euros reflects reality , unlike the speculation surrounding the RTX 5070 Ti.
✔ Performance equivalent to an RTX 4080 (theoretically cheaper).
✔ DLSS 4 with Multi-Frame Generation: Multiplies FPS by 3 or 4 in 4K with stunning image quality (99% fidelity to native).
✔ GDDR7 and Overclocking: Gains of 6 to 10% possible with an additional 1700 to 2000 MHz on GDDR7.
✔ Excellent cooling: Very low temperatures (49-63°C) and quiet operation, even under load.
✔ Production performance: 15-25% gains on professional software (better than in-game).
✔ VRAM: 16GB of GDDR7 providing significant headroom for the future.
✔ Ray Tracing Gains: Significant 39% improvement over the previous generation thanks to RT cores.
✔ Reasonable consumption: 300W TDP, 850W power supply sufficient.
✔ Outperforms the RX 7900 XTX in raw performance.
✖ Outrageous real price: €1130 to €1400 against €885 advertised (ruins the interest of the card).
✖ Non-existent stocks at MSRP price (purely marketing theoretical price).
✖ Modest gains vs RTX 4070 Ti Super: Only 12 to 17% (far from the 30% announced).
✖ DLSS 4 input lag: Remains that of the base framerate (limit for competitive gaming).
✖ No Founders Edition model (no price regulation by Nvidia).
✖ Minimal architectural evolution (+5% CUDA cores): More of a refresh than a new feature.
✖ 3-pin 12VHPWR connector for 300W.
✖ Overclocking limited by some manufacturers (e.g. MSI blocks the TDP).
✖ **Does not justify an upgrade from an RTX 4000**.
✖ At actual price, the RX 7900 XTX is a better buy for €150 less.
Verdict: An excellent card sabotaged by its launch
The Nvidia RTX 5070 Ti is, on paper, a remarkable card. It manages to offer the performance of an RTX 4080 in a smaller and theoretically more affordable form factor. Overclocking becomes interesting again after years of stagnation, DLSS 4 visually impresses despite input latency remaining problematic, and performance in professional production is simply excellent .
But market reality dampens the enthusiasm. At an average price of €1,130 , well above the advertised MSRP of €885, the card loses all appeal compared to the competition . Some custom models even reach €1,400, a price that's hard to justify. For us, the reasonable ceiling remains €1,000. Beyond that, alternatives like the RX 7900 XTX or even an RTX 4080 Super on sale become more relevant and rational.
This is extremely frustrating, because Nvidia has delivered a mature product this time around: ray tracing is finally becoming usable in practice, DLSS 4 opens up new possibilities for gaming and content creators, and the Blackwell is truly effective. Yet, between the marketing and the reality of inventory and pricing, there is a gap that Nvidia seems unable, or unwilling, to bridge.
So our recommendation remains simple and unchanged: wait . Wait until stocks stabilize, real prices align, and perhaps even until the arrival of the Radeon RX 9000 forces the market to reprice.
What if you really need a card today? Ironically, the Radeon RX 7900 XTX , released in late 2022, remains the best option in 2025. It's sad to see, but it's the reality of today's market.
Our best articles on graphics cards
NVIDIA's new graphics card, the GeForce RTX 5080, is finally here. If you have
PNY, a key player in the graphics card market, returns to the forefront
MSI is hitting hard in the very high-end with the MSI RTX 5090 Suprim Liquid,
The GeForce RTX 4090 made a spectacular entrance onto the market, driven by
After weeks of rumors, leaks, and feverish anticipation, it's finally here.
With its new Blackwell architecture, NVIDIA drives the point home and confirms its dominance in the