The NVIDIA RTX 5070 is launching this week, and I have already posted my first review of the GPU yesterday. I covered the ASUS TUF GeForce RTX 5070 model, and while the GPU itself delivers some great cooling and a small form factor, the RTX 5070 tech just isn’t there.
I say this because I had some pretty high expectations for the RTX 5070. This is the card that NVIDIA boasted had the same performance as the RTX 4090. Does that mean you should toss out your GPU and pick up this entry-level card instead? Hell no.
NVIDIA claimed that the RTX 5070 could theoretically match the RTX 4090 when pushing DLSS 4 AI upscaling and frame generation. This would lean on the new multi-frame generation feature that sees the card generate up to four times the frame for every single frame.
Now, on paper, this means the RTX 5070 could take a game that runs quite poorly without DLSS 4 and boost it up to an impressive frame rate. However, the card doesn’t use magical unicorn farts to do this. It relies on complex AI algorithms which have to be calculated at the tenth of a millisecond and sent to the ROPs for rendering.
These calculations take time to produce, especially if you’re multiplying them by four. So think of it this way. You move the camera in a video game, which triggers more new frames. These frames are now going through quite a substantial pipeline in order to get pumped out at four times the quantity. This whole process takes time, which means the result of your new frame is delayed compared to the input you have initiated by rotating your camera. Ladies and gentlemen, that is what we call latency.
The NVIDIA RTX 5070 has some of the most intense latency I have ever seen from a GPU. Testing out Alan Wake 2 with multi-frame gen X4 on the RTX 5070, I was able to get some impressive frames. We are talking about 60FPS at 4K with path tracing maxed out and every graphics setting on the max. However, this all came at the cost of latency.
During my tests, I measured latency as high as 351ms. This means every movement and action would only show up on the screen 0.3 seconds after I initiated it. The latency is so bad that the game is completely unplayable.
Granted, this was at 4K with everything maxed out. I then tested out the same scene but with lower settings. Even making use of Balance, Quality, and Performance DLSS 4 modes, the latency struggled to get down to 150ms. It makes the feature completely pointless and essentially, it means comparing the card’s performance to the RTX 4090 isn’t ethical. The RTX 4090 can deliver this same performance with a 20ms latency reading without even having to nudge into the Performance or Ultra Performance modes.
Then there’s the pathetic 12GB of VRAM that again makes the RTX 5070 feel like a card from a few years ago. I had instances where some games simply would not open, and some would crash when I enabled a basic ray tracing setting.
Sure, path tracing is a big ask from a card with these specs, but NVIDIA claimed it was just as powerful as the RTX 4090…. Cyberpunk 2077 nudged just into the 30FPS mark when using RT Overdrive, with MFG X4. A setting that the RTX 4090 breezes through.
So yeah, the RTX 5070 does have the same frames as the RTX 4090 in very certain situations, but those frames come at the cost of latency. It is impossible to generate these frames without the latency. Well, for now anyway. To stand in front of the world and tell them to buy an RTX 5070 and expect the same performance as the RTX 4090 is a blatant lie. I just thought you should know.
