Inspect your assumptions– Nvidia’s RTX 50-series standards do not inform the complete tale

Nvidia simply exposed its series of RTX 50-series GPUs at CES 2025, and made some huge efficiency cases. The firm states that the flagship RTX 5090 is two times as rapid as last gen’s RTX 4090, and it’s not the only one. Throughout every one of the standards Nvidia shared, from the RTX 5090 to the RTX 5070, Nvidia states the brand-new GPUs are two times as rapid as their last-gen equivalents.

We usually do not see that type of generational uplift. Did Nvidia placed some kind of wonderful GPU dirt right into its brand-new graphics cards? If you look very closely at the standards, it comes to be clear that, yes, Nvidia did precisely that. Which wonderful GPU dirt is called DLSS 4.

You can see what I suggest from the graph above. In the afterthoughts, Nvidia keeps in mind that it made use of DLSS 3 for the RTX 4090 and DLSS 4 for the RTX 5090. DLSS 4 sustains multi-frame generation– simply put, it can create several structures in between each “genuine” framework. Nvidia is supplying greater than two times the efficiency of the RTX 4090 with the brand-new RTX 5090, however that mostly begins the rear of multi-frame generation in DLSS 4.

That comes to be clear when you consider Much Cry 6 and A Plague Story: Requiem. The relative numbers in these video games are a lot less remarkable, with Nvidia revealing someplace around a 1.5 x renovation in efficiency. A Plague Story: Requiem just sustains DLSS 3, so the RTX 5090 does not have the multi-frame generation benefit. And Much Cry 6 does not sustain DLSS in any way, and it’s the weakest of the video games Nvidia flaunted.

This holds true throughout every one of the standards. You can see the outcomes for the RTX 5070 over, which is anticipated to get here in February. Throughout Nvidia’s discussion, chief executive officer Jensen Huang strongly stated that the $550 RTX 5070 provides efficiency on the same level with the RTX 4090, however once again, that insurance claim comes mostly on the back of multi-frame generation in DLSS 4.

There’s a trouble with multi-frame generation, which’s latency. As is currently the instance with DLSS 3, you’re basically using a hold-up with framework generation. It’s a mild hold-up, however the graphics card needs to provide 2 structures in order to execute the framework interpolation to “create” the structures in between. This isn’t a trouble at high framework prices, however DLSS 3 faces concerns if you’re going for a reduced base framework price.

That concern comes to be overstated when producing several structures, which is something I saw when checking out Lossless Scaling, which sustains 4x framework generation like DLSS 4. You might have the ability to go from 30 structures per 2nd (fps) to 120 fps with DLSS 4, however you’ll still obtain the responsiveness of 30 fps.

Currently, I have not made use of DLSS 4– nobody beyond Nvidia has. Based upon previous DLSS updates, I make certain it will certainly be better than what Lossless Scaling has the ability to provide, specifically with the brand-new Reflex 2 latency decrease attribute going for complete clip. And in justness to Nvidia, it’s still revealing a strong generational uplift in video games like Much Cry 6 and a Plague Story Requiem. It’s simply not two times the efficiency of the previous generation.

It should not be long till I have Nvidia’s brand-new GPUs in hand and I have the ability to evaluate them myself and see exactly how well DLSS 4 in fact stands up. No matter, it’s clear that the cases of two times the efficiency as the previous generation do not inform the complete tale regarding Nvidia’s brand-new graphics cards.