I’ll cut right to the chase: If you’re an RTX 4070 owner looking to upgrade to NVIDIA’s latest video cards, the $549 RTX 5070 won’t be much different than what you’re used to. Aside from DLSS 4’s ability to generate multiple extra frames, the 5070 is practically the same as the 4070 Ti in our testing (and in some cases it’s worse). So if you’re already committed to spending more than $500 for a new GPU, you’re likely better off saving up a bit more for the $749 5070 Ti, which has more VRAM (16GB instead of the 5070’s 12GB) and more capable hardware for gaming in 4K.
Given the current state of the video card market, though, those prices are basically just theoretical. NVIDIA’s previous RTX 50-series cards are already selling well above their retail prices at most stores, if you can find them in stock at all. So I wouldn’t be surprised to see the RTX 5070 suffering the same fate. For now, though, Best Buy has NVIDIA’s Founder’s Edition card and an ASUS model listed for $549.
The RTX 5070 comes in $50 cheaper than the RTX 4070’s original launch price, so at least we’re edging closer to the days of the $500 RTX 3070 (again, if you were lucky enough to snag it for that MSRP). The 5070’s price alone made it the new NVIDIA GPU I was most looking forward to test. But after spending some time with it, I can see why NVIDIA was able to keep it relatively affordable: It’s simply not a huge jump from the 4070.
Hardware
As the runt of the litter (so far), the RTX 5070 is clearly behind its siblings on paper. It sports 6,144 CUDA cores, compared to the 5070 Ti’s 8,960, and the aforementioned 12GB of GDDR7 memory. Notably, the RTX 4070 also shipped with 12GB of memory, though that was slightly slower GDDR6X RAM. The new card has a 33 percent memory bandwidth advantage — 672 GB/s vs 504 GB/s — but it would have been nice to see more RAM, especially since AMD managed to deliver 16GB in the upcoming $549 Radeon 9070.
The NVIDIA Founder’s Edition RTX 5070 we reviewed looks similar to the 5090, with a gorgeous metal case and dual fans to cool down the PCB sandwiched in the middle. A dongle with two 8-pin PSU connections is included in the box, and the 5070 requires a 650W power supply at a minimum. (Though I’d recommend going for at least an 800W PSU if you want to future-proof a bit.) Around the back, the 5070 features three DisplayPort 2.1 connections and an HDMI 2.1 port.
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 |
In use: Solid 1440p and 4K, especially with multi-frame generation
I was a bit worried that testing the RTX 5070 would meaning scaling down my Avowed settings from maxed out 4K to 1440p. First-world problems, I know. But the 5070 managed to stay above 60 fps in 4K with ray tracing and graphics settings cranked up — so long as I was using DLSS upscaling. I typically saw between 65 fps and 75 fps in a variety of areas, which played smoothly on my 32-inch Alienware QD-OLED monitor.
I saw a smoother 120 fps while playing Avowed in 1440p, but I preferred sticking with 4K since it’s not a very fast-paced game. Instead, it’s a title where you’re often moving slowly through the world and absorbing breathtaking vistas and gorgeously constructed environments. Without DLSS, the 5070 crawled along at just 15 fps in 4K. In comparison, the 5070 Ti reached 90 fps in 4K with DLSS, and 32 fps in 4K without any upscaling.

Clearly, the RTX 5070 is best-suited to 1440p gaming, just like the 4070. But it’s more capable if you don’t mind the interpolated frames from DLSS 4’s multi-frame generation. It can generate up to three frames from every one that’s natively rendered. That led to me seeing 168 fps in Dragon Age: The Veilguard while playing in 4K with graphics settings cranked up. The 5070 Ti, meanwhile, hit over 200 fps while playing Veilguard in 4K, while the 5090 reached 250 fps.
Your experience with the RTX 5070 will mostly come down to the amount of games you play with DLSS 3 and 4 support. Halo Infinite, which doesn’t work with any upscaling technology yet, hit an acceptable 97 fps in 4K. Still, I think most players would prefer the smoother 138 fps performance in 1440p, or even the 178 fps I saw in 1080p. Cyberpunk 2077, meanwhile, flew in 4K with multi-frame gen in ray tracing overdrive mode, reaching 115 fps.
None |
3DMark TimeSpy Extreme |
Geekbench 6 GPU |
Cyberpunk (4K RT Overdrive DLSS) |
Blender |
NVIDIA RTX 5070 |
10,343 |
178,795 |
115 fps (4x frame gen) |
6,015 |
NVIDIA RTX 5070 Ti |
12,675 |
238,417 |
153 fps (4X frame gen) |
7,365 |
NVIDIA RTX 4070 Ti Super |
11,366 |
220,722 |
75 fps (1x frame gen) |
7,342 |
NVIDIA RTX 4070 |
8,610 |
N/A |
55 fps (1x frame gen) |
6,020 |
Benchmarks paint a far starker picture for the RTX 5070. It was practically identical to the RTX 4070 Ti in 3DMark’s TimeSpy Extreme and Speedway tests. And it lagged behind that card in the Port Royal and DirectX ray tracing tests, as well as in Blender’s benchmark, where it was a full 1,200 points behind the 4070 Ti! NVIDIA clearly leaned more on tensor core performance for DLSS for the 5070, rather than raw rasterized or ray tracing speeds.
Unsurprisingly, NVIDIA’s Founder’s Edition cooling setup shined on the RTX 5070, where it kept temperatures between 35C and 40C while idling, and under 75C while I was gaming and running benchmarks. The fans were audible but not annoying when I was really stressing the card, thoughI didn’t notice much since I also had audio blasting on my speakers.

Should you buy the RTX 5070?
As I mentioned up top, I don’t think the RTX 5070 is a compelling upgrade if you’re already satisfied with an RTX 4070. But if you’re coming from an older GPU, and you’re not ready to shell out $750 (or likely more) for the 4070 Ti, it’s currently the best value NVIDIA is offering. The 5070 will handle 1440p gaming well, and it’ll be acceptable for basic 4K gaming for non-DLSS titles. Throw multi-frame generation in the mix, and things look even better.
Still, for the price, I really wish the 5070 was more of a step up from the 4070. If the 5070 Ti ever comes back down to its $750 retail price, it’s likely a better longterm purchase since it has 16GB of RAM. That’ll make it better equipped for handling massive 4K textures, and it’s also more powerful in general to deal with more demanding gaming engines and ray tracing. The 5070 is a GPU that might satisfy you for a few years, whereas I think the 5070 Ti could handle most of what you throw at it for the next four years.
For now, though, my advice is the same as it was in my 5070 Ti review: Don’t rush out to buy a new video card right away. NVIDIA’s pricing and stock availability is incredibly volatile, and it’s not worth paying over $900 for something like the 5070 Ti. In a few months, prices may fall back down to Earth, especially with AMD’s more reasonable $549 and $599 Radeon RX 9070 and 9070 XT in the mix. In fact, it’s worth waiting a bit just to see how well those two cards compare to the 5070.

Wrap-up
While NVIDIA mostly succeeded with delivering a semi-affordable midrange GPU with the RTX 5070, it’s competing against itself when compared to the excellent RTX 4070. It’s limited by having only 12GB of VRAM, and without the magic of DLSS AI upscaling, its performance isn’t that much different than the 4070 Ti. In the end, the 5070 feels a bit like a missed opportunity to deliver a truly performant card for close to $500. Instead, it’s more of the same, with a healthy dose of AI.
This article originally appeared on Engadget at https://www.engadget.com/gaming/pc/nvidia-geforce-rtx-5070-review-basically-a-4070-ti-thats-better-at-4k-140016718.html?src=rss
Recent Comments