Nvidia GeForce RTX 5070 Review

Nvidia has been having a hard clip with its latest enactment of graphics cards. While the RTX 5070 and each of its much costly brethren are fine, they person failed to present the benignant of generational leaps you would expect aft 2 years. The RTX 5070 is the worst illustration of this yet. Because portion each RTX 50-series card, from the RTX 5070 Ti to the RTX 5080 has a tiny pb implicit their Super-variant counterparts from the past generation, the RTX 5070 is consecutive up tied with the 4070 Super – and is adjacent slower successful a batch of games.

So portion astatine $549 the Nvidia GeForce RTX 5070 is technically $50 cheaper than the RTX 4070 Super was astatine motorboat – if motorboat prices clasp up astatine slightest – you’re fundamentally conscionable getting a insignificant terms chopped connected an existing card. That’s not an breathtaking proposition, adjacent erstwhile Nvidia throws successful DLSS 4 multi-frame procreation to assistance niche users with high-refresh monitors.

Specs and Features

Just similar the RTX 5070 Ti earlier it, the RTX 5070 is built connected the Blackwell architecture. Before the RTX 5090 launched successful January, Blackwell was typically known arsenic the GPU tech down the supercomputers powering cloud-based AI applications. So, portion determination is simply a spot of betterment to the CUDA cores, Nvidia’s existent absorption was connected its caller Tensor cores, which powerfulness DLSS 4.

The RTX 5070 itself features 6,144 CUDA cores, dispersed crossed 48 Streaming Multiprocessors, oregon SM. Each of these SMs besides has an RT Core and 4 Tensor Cores, making for 48, and 192, respectively. Therein lies the occupation though. While this is fundamentally a akin configuration to the RTX 4070 – which had 46 SMs, the RTX 4070 Super, which replaced it successful 2024, has 56 SMs. So portion each SM successful a Blackwell GPU is simply a spot much powerful, the architecture doesn’t supply capable of a bump to marque up for the sheer quality successful silicon.

The RTX 5070 besides has the aforesaid magnitude of VRAM arsenic its predecessor, clocking successful astatine 12GB. However, dissimilar its predecessor, it's upgraded to GDDR7, which has higher bandwidth, adjacent though the RTX 5070 is connected the aforesaid 192-bit representation autobus arsenic the RTX 4070 and the 4070 Super. The faster representation does assistance successful games that are demanding connected memory, arsenic it’s capable to proviso information to the GPU faster. However, with the AMD Radeon RX 9070 launching soon for the aforesaid terms with 16GB of VRAM, it would person been bully to spot Nvidia supply a spot much framework buffer.

Continuing successful the contented of Blackwell graphics cards needing much powerfulness than their predecessors, the Nvidia GeForce RTX 5070 requires 250W of power, with Nvidia recommending a 650W powerfulness proviso to tally the thing. That’s a flimsy summation implicit the 220W of the RTX 4070 Super oregon the 200W of the RTX 4070, but its inactive debased capable to lone request 2 8-pin PCIe cables to powerfulness it, adjacent if you’ll person to plug them into the included 12V-6x6 adapter.

There is simply a caller spot of silicon successful the RTX 5070, though: an AI Management Processor, oregon AMP. This core, which sits connected its ain extracurricular of the SMs, intelligently assigns enactment to antithetic parts of the GPU, a task that has traditionally been handled by the CPU. This is the concealed condiment that allows Nvidia to propulsion its framework procreation tech to make aggregate frames disconnected of a azygous rendered frame, arsenic it present has a section halfway that paces retired the frames to debar artifacting oregon uneven framework times.

DLSS besides gets its biggest cardinal alteration since it was archetypal implemented successful the RTX 2080 successful 2018. Now, alternatively of the Convolutional Neural Network (CNN) that DLSS ran connected originally, DLSS tin present tally connected a Transformer Model. This doesn’t marque it immoderate faster per se, but it does greatly amended accuracy, making for a cleaner upscaled representation with little blur oregon ghosting artifacts. Nvidia was capable to get the Transformer Model moving connected older RTX graphics cards, too.

DLSS 4

What Nvidia is hoping volition truly merchantability the RTX 5070, on with the remainder of the RTX 5000 lineup, is DLSS 4. Now, DLSS has been astir for years and has grown to beryllium the modular upscaling solution. And portion DLSS 4 does amended the upscaling with a Transformer exemplary that’s overmuch much accurate, the existent propulsion this clip astir is Multi-Frame Generation, oregon MFG.

Frame procreation debuted with the RTX 4090 a mates years ago. The thought primitively was that your GPU would render 2 frames, and past make an AI-interpolated framework successful betwixt the frames, successful bid to boost your framework rate. This attack drastically improves framework rates, but due to the fact that the GPU is fundamentally holding a rendered framework hostage until the generated framework is produced, it could summation latency somewhat – usually by a fewer milliseconds.

To compensate for the accrued latency, Nvidia Reflex is required successful bid to alteration immoderate benignant of framework generation. Essentially, Reflex zeroes retired your render queue, and synchronizes your CPU and GPU truthful that frames are rendered conscionable arsenic they’re needed, alternatively than having a queue of frames waiting for your GPU to render them. This tin dramatically trim latency, making the added latency successful Frame Generation little noticeable.

With the RTX 5070, however, Nvidia present supports Multi-Frame Generation, truthful alternatively of having 1 AI-generated framework based connected each rendered frame, the GPU tin make up to 3 frames astatine a time. Obviously, this makes your framework complaint spell up significantly, but it does adhd a spot much latency to the equation.

In Cyberpunk 2077, astatine 1440p and with the Ray Tracing Overdrive preset, the Nvidia RTX 5070 gets astir 70 fps with an mean PC latency of 31ms. That’s not bad, and is wholly playable, but what if you privation adjacent much frames? Well, with 2x framework gen – what’s supported connected the last-generation hardware – the framerate astir doubles, up to 116 fps, and with 36ms of latency. Frame gen straight adds 5ms of latency, but that’s not overmuch astatine all. With Frame Gen acceptable to 4x, however, that framerate goes up to 200 fps, astir tripling the archetypal framework rate. Though, the latency gets a spot worse, going up to 38ms of mean latency.

Keep successful mind, though, that MFG isn’t determination to crook an unplayable framework complaint into a creaseless experience. If you’re playing a crippled that’s lone getting 20 frames per second, framework procreation volition marque the framework complaint spell up, but you’ll inactive beryllium subjected to the precocious effect times that travel with debased framework rates. MFG besides needs much frames successful bid to accurately foretell however objects successful the country are moving from framework to frame, and a little framework complaint means determination are bigger differences betwixt each rendered frame. This makes it harder for the algorithm to accurately make images, and volition present errors and artifacts into the scene. So, don’t crook connected framework gen unless you’re already getting thing similar 60 fps already.

Multi-Frame Generation, then, makes for a somewhat little responsive gaming experience, but with much frames being sent to your display. If you person a show with a precocious refresh rate, this results successful amended ocular smoothness, adjacent if you’re not getting the boosts to responsiveness you’d get from faster autochthonal rendering. This makes it hard to urge for anyone playing fast-paced shooters similar Black Ops 6 oregon Marvel Rivals, but is perfectly worthy it if you play single-player games similar Cyberpunk 2077 oregon Indiana Jones and the Great Circle.

Performance

When a caller procreation of graphics cards travel out, it’s tenable to expect them to person a large betterment implicit its predecessor. However, portion the Nvidia GeForce RTX 5070 has a decent uplift implicit the RTX 4070, it falls connected its look arsenic soon arsenic the RTX 4070 Super enters the conversation. The RTX 4070 Super deed the marketplace astir a twelvemonth aft the archetypal RTX 4070, replacing it astatine its $599 price, and delivering immense improvements successful performance, particularly for a mid-generation refresh. That makes it a hard enactment to follow, and unfortunately, the RTX 5070 does not propulsion it off.

Across each tests astatine each resolutions, the RTX 5070 was connected mean conscionable 4% faster than the RTX 4070 Super, adjacent portion it was 20% faster than the archetypal RTX 4070. When I dial successful the results to lone see 1440p, that fig does amended a spot to a 5% pb implicit the 4070 Super. However, instrumentality the synthetic 3DMark tests retired of the equation, and the 1440p betterment is conscionable 3%. What makes it adjacent worse is that there’s 1 crippled wherever the RTX 5070 soars implicit the 4070 Super: Total War: Warhammer 3. Take that one game retired of the equation, and the mean betterment implicit the RTX 4070 Super is 0%. That’s embarrassing, but it really feels worse erstwhile I look astatine it game-by-game.

I tested each graphics paper connected their existent nationalist drivers. That means each Nvidia cards were tested connected Game Ready Driver 572.60, and each AMD cards were tested connected Adrenalin 24.12.1. The Nvidia GeForce RTX 5070 itself was tested connected a reappraisal operator provided by Nvidia.

3DMark doesn’t needfully bespeak real-world gaming performance, but it does springiness a bully glimpse astatine what the imaginable of a graphics paper could be. In Speed Way, the RTX 5070 gets 5,845 points, compared to 5,190 from the RTX 4070 Super and 4,498 from the RTX 4070. That’s a tiny 12% pb implicit the RTX 4070 Super, but it’s a wider pb against the archetypal 4070. However, the leads constrictive successful the adjacent mates tests. In Steel Nomad, the RTX 5070 lone leads the 4070 Super by 10%, with lone an 8% pb successful Port Royal.

The problems commencement to amusement up successful Call of Duty: Black Ops 6. Rather than getting a show pb implicit its predecessor, the RTX 5070 is 5% slower than the RTX 4070 Super, portion maintaining conscionable a 7% pb implicit the RTX 4070 astatine 1440p. At 131 fps, it’s acold from slow, but it’s inactive slower than what came before.

In Cyberpunk 2077 the losses proceed astatine 1440p. The RTX 5070 gets a respectable 90 fps connected the Ray Tracing Ultra preset with DLSS acceptable to balanced, but the RTX 4070 Super gets 96 fps, making for a 6% show loss gen-on-gen. What’s more, the RTX 5070 is really slower than the AMD Radeon RX 7900 GRE, which gets 143 fps successful the aforesaid test, with FSR acceptable to balanced. That doesn’t bode well, with the Radeon RX 9070 astir the corner.

I tested Metro Exodus: Enhanced Edition with nary upscaling, due to the fact that it only supports DLSS, which means I can’t get comparable results from AMD graphics cards. But adjacent here, the RTX 5070 loses 4% show implicit the RTX 4070 Super, adjacent if it’s 19% faster than the RTX 4070. Luckily, the RTX 5070 stays implicit 60 fps, which is saying something.

The RTX 5070 gets 115 fps successful Red Dead Redemption 2 astatine 1440p, which is wrong the borderline of mistake with the 4070 Super’s 116 fps. But that’s inactive a dead-even effect with a one-year spread and a caller architecture.

Total War: Warhammer 3 is simply a spot of an outlier, though. The RTX 5070 gets 134 fps successful this game, which is simply a 22% pb implicit the RTX 4070 Super’s 110 fps. That’s conscionable a azygous game, but it does possibly suggest the RTX 5070 has a spot of country to turn successful the future. This is apt owed to the improved representation bandwidth.

Assassins Creed Mirage is different wash, with the RTX 5070 getting 163 fps to the RTX 4070 Super’s 165 fps. That’s wrong borderline of error, but it’s inactive a 1% nonaccomplishment for the caller graphics card. This crippled besides shows the lowest betterment implicit the RTX 4070, which gets 151 fps: that means the RTX 5070 lone has an 8% pb here.

In Black Myth: Wukong, the RTX 5070 pulls up of the RTX 4070 Super, but conscionable barely. It gets 66 fps astatine 1440p, compared to 64 fps from the RTX 4070 Super and 55 from the archetypal 4070. But that’s inactive conscionable a 3% pb implicit the RTX 4070 Super.

Luckily, Forza Horizon 5 looks overmuch amended for the RTX 5070. In this game, the RTX 5070 gets 168 fps astatine 1440p with the Extreme preset and DLSS acceptable to Balanced. With the aforesaid settings, the RTX 4070 Super gets 149 fps, which makes for a 13% pb successful favour of the 5070. That’s much successful enactment with the imaginable shown by 3DMark, adjacent if it’s not the astir breathtaking generational improvement.

It would person been atrocious capable if the RTX 5070 had single-digit improvements implicit the RTX 4070 Super, but with truthful galore games showing the caller GPU being slower than its predecessor, it becomes truly hard to spot wherefore anyone should bargain it. The lone crushed you should get this is if you person an older graphics paper and you don’t person the currency to walk connected an RTX 5070 Ti. And adjacent then, you’re amended disconnected getting an RTX 4070 Super – if you tin find one.

Jackie Thomas is the Hardware and Buying Guides Editor astatine IGN and the PC components queen. You tin travel her @Jackiecobra