NVIDIA GeForce RTX 5070 FE review: Not the 'RTX 4090-level' card Jensen Huang teased, but solid for 1440p gaming
Meanwhile, AMD is lurking behind.
#nvidia #rtx5070 #pcgaming
By Aaron Yip -
The GeForce RTX 5070 is NVIDIA's mainstream GPU from the RTX 50 Series. Image: HWZ
The fourth GPU in the RTX 50 Series to launch after the GeForce RTX 5090, 5080 and 5070 Ti, NVIDIA's GeForce RTX 5070 is geared at the mainstream 1440p gaming market and is also the direct successor to the RTX 4070. At US$549, the RTX 5070 is very likely the graphics card that most of us might actually be able to afford to buy, given that it is also the lowest-price GPU in the RTX 50 Series until the 5060 Ti and 5060 come along (if, eventually).
Just like its faster siblings, the RTX 5070 is based on the Blackwell architecture and so offer the same technologies and features such as DLSS 4, Multi Frame Generation and neural rendering (read about them here). NVIDIA has also made a Founders Edition, after skipping one with the RTX 5070 Ti GPU, and there’s a lot going for this 1440p-focused GPU. In our time with it, it has consistently shown an average 20% gen-on-gen improvement over its RTX 4070 predecessor. It may not exhibit as significant a performance improvement gap as the RTX 5070 Ti, but consider this: NVIDIA is launching the RTX 5070 at a US$50 lower price point than the RTX 4070, but with more bells and whistles added inside.
The RTX 5090 FE has shrunk compared to the RTX 4090 FE but is still much larger than the RTX 5070 FE. Image: HWZ
Of course, the launches of the RTX 5090 and RTX 5080 have shown that SRPs are often meaningless, as retail prices for these graphics cards are marked up significantly due to GPU scarcity. This is one reason why NVIDIA is launching the RTX 5070 later, allowing time to stabilize its supply chain. This will be a story for another day, though.
GeForce
Graphics Card | RTX 5090 | RTX 5080 | RTX 5070 Ti | RTX 5070 | RTX 4070 Super | RTX4070 |
GPU | Blackwell (GB202) | Blackwell (GB203) | Blackwell (GB203) | Blackwell (GB205) | ADA Lovelace (AD104) | Ada Lovelace (AD104) |
Process | 4nm (TSMC) | 5nm (TSMC) | ||||
Transistors | 92 billion | 46billion | 46
billion | 31 billion | 36 billion | 36 billion |
Streaming Multi-processors (SM) | 170 | 84 | 70 | 48 | 56 | 46 |
CUDA cores | 21760 | 10752 | 8960 | 6144 | 7168 | 5888 |
Tensor Cores | 680 (Gen 5) | 336 (Gen 5) | 280
(Gen 5) | 192 (Gen 5) | 224 (Gen 4) | 184 (Gen 4) |
RT Cores | 170 (Gen 4) | 84 (Gen 4) | 70
(Gen 4) | 48 (Gen 4) | 56 (Gen 3) | 46 (Gen 3) |
GPU base / boost clocks (MHz) | 2017 / 2407 | 2295 / 2617 | 2300/ 2452 | 2325 / 2512 | 1980 / 2475 | 1920 / 2475 |
Memory | 32GB GDDR7 | 16GB GDDR7 | 16GB GDDR7 | 12GB GDDR7 | 12GB GDDR6X | 12GB GDDR6X |
Memory bus width | 512-bit | 256-bit | 256-bit | 192-bit | 192-bit | 192-bit |
Memory bandwidth | 1,790 GB/s | 960 GB/s | 896 GB/s | 672 GB/s | 504 GB/s | 504 GB/s |
Interface | PCIe 5.0 | PCIe 4.0 | ||||
TDP | 575W | 360W | 300W | 250W | 220W | 200W |
Price | US$1,999 | US$999 | US$749 | US$549 | US$599 | US$599 |
Unlike like the RTX 5070 Ti (which is closer to the 5080), the 5070 model is built on the Blackwell GB205 silicon – unlike the RTX 4070 Ti and RTX 4070 that shared the same ADA Lovelove silicon. Memory wise, NVIDIA has stuck with 12GB of VRAM and this could yet prove to be a mistake in 2025 as AMD’s soon-to-be-launched Radeon RX 9070 cards will have 16GB. However, NVIDIA is also now using the faster GDDR7 memory on a 192-bit bus that gives it 672GB/s of memory bandwidth – it’s a lot zippier than the RTX 4070 Super and non-Super cards. Power consumption has risen by 25% over the RTX 4070 too, with NVIDIA recommending a 650W PSU for the RTX 5070’s 250W TDP.
The RTX 5070 FE retains the same 2-slot design as its bigger FE siblings. Image: HWZ
Looking at the backplate alone is impossible to tell which RTX 50 Series card it is. Image: HWZ
Looks wise, the RTX 5070 FE retains the same design language as the 5090 and 5080 Founders Edition cards, including the 2-slot design and 12V-2x6 power connector. The card is much shorter though and resembles more like the RTX 4070 FE, which is great for a small-factor PC build.
Performance wise, NVIDIA talked a big game for the RTX 5070 when it was first announced at CES 2025 – CEO Jensen Huang had boldly claimed “it can deliver RTX 4090 performance for just S$549”. Let’s find out how much of that is true, if at all.
Our test rig
Hello, RTX 4070 FE old friend. Image: HWZ
We are using the updated same test rig that was used to benchmark all three previous GeForce RTX 50 Series cards and comes with the following specifications:
- AMD Ryzen 9 9950X
- ROG Maximus Z890 Hero
- Samsung 990 Pro 1TB SSD
- Kingston Fury 32GB DDR5
- Windows 11 Home 64-bit
- ASUS ROG Swift OLED PG32UCDM 4K Gaming Monitor
For this review, I’ll be comparing the RTX 5070 not only with the RTX 4070 Super and vanilla 4070 but also the card from the RTX 40 Series. Naturally, this includes the RTX 4090 to test NVIDIA's CEO's bold claims. Spoiler: It doesn't.
Gaming performance (rasterization)
My game list includes a mixture of old and recent games to test the RTX 5070's performance, and while the list isn’t exhaustive by any measure there are enough different game engines and APIs variety to give us an idea of broader performance trends.
The higher the fps, the better.
The higher the fps, the better.
The higher the fps, the better.
The higher the fps, the better.
The higher the fps, the better.
The higher the fps, the better.
In rasterization performance, the RTX 5070 is a solid card for 1440p gaming but it doesn't offer massive leap in performance over the last generation at resolutions lower than 4K. Though if you're coming from a 30 Series card or older, then the RTX 5070 would be a massive improvement. If you're gunning for 4K gaming, then you should think about the RTX 5070 Ti instead but that comes with a big jump in price too. 12GB won't be enough for the next couple of years on too, in my opinion, as modern games becomes hungrier for VRAM for 4K gaming. This is worth keeping in mind, as AMD will soon ships its RX 9070 and RX 9070 XT cards with 16GB of memory for the same price.
DLSS 4 and ray tracing
Like the RTX 5090, and by extension all GPUs in the RTX 50 Series, the true power of the RTX 5070 lies in its Multi-Frame Generation (MFG) feature.
The higher the fps, the better.
The higher the fps, the better.
The higher the fps, the better.
Cyberpunk 2077 is one of the earliest game to fully support DLSS 4 and Multi Frame Generation (MFG) and is a great test title, as it demonstrates the motion clarity improvements with higher frame rates that NVIDIA's new technique brings to the RTX 50 Series.
As we can see in the charts above, at higher resolutions like 4K with full settings and ray tracing switched on, the 5070 becomes slightly more playable at 53fps once Frame Generation is activated. But with 3x Multi Frame Generation, the game's framerate jumped to a smoother 76fps. Of course, the caveat here is that the 76fps may not feel "real" for those with more discerning hand-eye coordinations, because the game still maintains the same input latency it has after DLSS is applied – so it's less reactive than a real 76fps when you hit the mouse button to shoot.
But guess what? With MFG turned up to 4x, the RTX 5070 did match up to the RTX 4090. Jensen's not wrong about his claim, but I can almost see the rolled eyes and slow clapping from gamers with this.
Power and Temperature
The lower the power and watts, the better.
The RTX 5070 sports a 250W TDP, which is 25% more than the RTX 4070. But when running on the same application, the RTX 5070 showed better power and thermal efficiency, drawing a mere 10% more wattage than the older card but with 20% higher performance. It's a pretty good trade-off, in my opinion.
Conclusion
Hold on to your RTX 5070 purchase first. Image: HWZ
Some gamers will muse that the RTX 5070 only showed modest generational improvements in rasterization, but I think a 20% average improvement across most games tested for a card that is launched at a lower price point than its predecessor is more than decent. The RTX 5070 is a very competent card for 1440p gaming, and might even offer some 4K gaming performance with DLSS enabled (and with settings dropped down). But alas, it's difficult to recommend the RTX 5070 right now because of one simple reason: AMD's Radeon RX 9070 and RX 9070 XT cards.
It's not a stretch to say that AMD literally rained on NVIDIA's RTX 5070 parade when it surprised everyone by announcing a US$549 price tag for the RX 9070 and US$599 for the RX 9070 XT. Both cards are direct rivals to the RTX 5070 and RTX 5070 Ti cards, and AMD's renaming of its GPUs' model numbers and purposeful use of the "070" here is not lost on us.
Until we know how the Radeon RX 9070 and RX 9070 XT perform against the NVIDIA equivalents, I will recommend holding off on buying a GeForce RTX 5070 card then. But we won't have to wait long.
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.