Nvidia’s GeForce RTX 5090 and RTX 5080 are actual, and nicely, they’re not as spectacular as many players hoped at first look – a minimum of while you aren’t utilizing DLSS 4.
I’m frightened about what meaning for the $550 RTX 5070 PC that players appeared so jazzed about, and whether or not your entire RTX 50-series may wind up being a flop if DLSS 4’s implausible new Multi Frame Generation fails to choose up steam quick. (It ought to!) And I’ve the numbers to again up my fears.
The “Blackwell”-powered RTX 50-series doesn’t ship tangible enhancements in worth or uncooked rendering efficiency from technology to technology, full cease. The GeForce RTX 5090 brute forces its method atop the benchmark charts due to Nvidia cramming it with 33 % extra CUDA cores than the RTX 4090, which helps it obtain 27 % sooner speeds at 4K decision – for an 18 % improve in value and a lot larger energy consumption. It’s undoubtedly brutal and badass, however meh for a brand new technology.
The RTX 5080 is even much less spectacular, delivering solely 15 % extra efficiency than the RTX 4080 Super it replaces at 4K, and a paltry 11 % enchancment at 1440p. Plus, it nonetheless solely has 16GB of GDDR7 reminiscence regardless of costing $1,000. Meh.
Balancing this out, in fact, are the huge efficiency positive factors supplied by Nvidia’s new DLSS 4, and extra particularly, its Multi Frame Generation function. MFG inserts as much as three AI frames between two “traditionally” rendered frames to ship body charges and visible smoothness hovering. The tightly paced frames it spits out could make even janky video games like Star Wars Outlaws really feel as snappy because the legendary Doom 2016. As I mentioned in our RTX 5090 review, it’s a real, literal game-changer, and it’ll work in 75 video games and apps from day one.
But the RTX 5080 exhibits that Nvidia is keen to bet an entire graphics card’s existence on DLSS 4’s promise. And the RTX 5080 isn’t the one one.
Worrying in regards to the RTX 5070 collection
Shortly after Nvidia revealed the Blackwell technology, we analyzed the RTX 50-series specs, making an attempt to estimate the place “traditional” gaming enhancements will fall. We got here rattling near nailing it, guessing that the RTX 5090 would wind up roughly 33 % sooner (it wound up at 27 %) and the 5080 would wind up 10 to 15 % sooner (actuality: 11.5 to 15 %).
How’d we try this? Simple: We merely in contrast their CUDA core counts to their predecessors. Blackwell GPUs are constructed utilizing the identical 4N course of at TSMC, so there was no simple uplift coming from a transistor node bounce, and as a substitute of prioritizing conventional efficiency positive factors, Nvidia centered a lot of its consideration on beefing up the RTX 50-series’ AI tensor cores to accommodate DLSS 4’s pleasant DLSS 4 Multi Frame Generation function, in addition to extra future-looking applied sciences like Neural Rendering and Mega Geometry.
Again: DLSS 4 is a lot extra than simply “fake frames.” It drastically improved gaming fluidity and smoothness in our playtesting in a handful of video games made accessible to press earlier than launch. AI upscaling is killing native rendering now, and it is sensible for Nvidia to prioritize it in a technology the place AMD’s Radeon crew already admitted that it isn’t targeting highest-end performance.
If the DLSS 4 dream fails to materialize, nonetheless – if the remainder of the 75 launch video games exhibit visible artifacts or glitches extra pronounced than we’ve seen, if players reject the additional latency AI frames require, 1,000,000 different ifs – the RTX 50-series may wind up being probably the most disappointing GPU generations in latest reminiscence.
Adam Patrick Murray / Foundry
We’ve already seen CUDA core counts correlate to the RTX 5090 and 5080’s efficiency. That’s a nasty omen for the RTX 5070 collection, regardless of beings priced $50 decrease than earlier than.
The identical CUDA comparisons undertaking the $750 RTX 5070 Ti as being the “star” of this technology, with a roughly 17 % improve over its predecessor, plus an improve to reminiscence capability, bandwidth, and pace that would assist enhance efficiency much more. A ~20 % efficiency improve generation-to-generation remains to be fairly mediocre so far as a progress goes, nevertheless it’s a extra tangible enchancment.
The reception for the $550 RTX 5070 could possibly be even frostier if our estimates proceed to carry; that GPU solely has 4 % extra cores than the existing 4070. It sticks to the identical 12GB of reminiscence as its predecessor, too, albeit of the sooner GDDR7 selection. We’ll have to attend to see the way it handles in actual life when it launches later this spring, however the 5070 has the potential to make the RTX 5080’s minuscule generational positive factors look comparatively large.
Nvidia hasn’t introduced lower-end GPUs but. What we’ve seen from the RTX 5070 laptop GPU makes us nervous about the RTX 5060, as a graphics card with simply 8GB of VRAM in 2025 doesn’t minimize it. We can’t opine an excessive amount of about that till the cardboard is introduced and examined although.
What’s clear is that Blackwell’s conventional rendering efficiency doesn’t transfer the needle ahead by a lot. The RTX 50-series was constructed for an AI-augmented future and it’ll dwell or die based mostly on DLSS 4’s reception.
Betting all of it on DLSS 4
Good information! I completely adore DLSS 4’s Multi Frame Gen in apply. So do Adam Patrick Murray and Will Smith, my cohorts on PCWorld’s video staff who’ve been knee-deep testing RTX 50-series GPUs for weeks now. Check out our lengthy dialogue about DLSS 4 within the time-stamped video beneath if you wish to hear our deeper ideas.
But make no mistake: Nvidia’s whole RTX 50-series lineup bets that DLSS 4’s humongous, AI-powered positive factors in fluidness and body charges shall be embraced by PC players. And that’s certainly a chance.
DLSS 4 blew my hair again throughout testing, however many PC players derisively name AI-generated frames “fake frames.” The RTX 40-series’ Single Frame Generation confronted withering criticism at launch as a consequence of visible glitches and latency issues. I’ve seen many, many fewer of these with my hands-on time with Multi Frame Generation – or maybe they’ve merely blown by too quick for my eyes to note, churning out at insanely excessive body charges.
We’ve solely had entry to a handful of MFG-equipped video games, and final technology, DLSS 3’s Single Frame Gen could possibly be very hit and miss relying on the sport you’re taking part in. (Microsoft Flight Sim notoriously had points with glitchy textual content, for instance). Making DLSS 4 accessible in 75 games and apps at launch will put a lot of software program in players’ fingers for testing, which is glorious and an enormous enchancment from previous norms, nevertheless it stays to be seen if all of them ship the identical finger-licking efficiency as Cyberpunk 2077, Star Wars Outlaws, and Alan Wake 2.
Adam Patrick Murray / Foundry
How to allow DLSS 4 in all these video games gained’t essentially be apparent, both. While some video games are being up to date for DLSS 4 by builders, most of the 75 video games and apps would require you to allow “DLSS Override” choices deep within the Nvidia App, which tells the software program to exchange older DLSS 3 information with the brand new DLSS 4 capabilities.
Will players know to try this? Will these DLSS Override video games carry out as easily as native titles? Will lower-tier playing cards just like the RTX 5070 and 5060 ship sufficient uncooked efficiency to hit 60 or extra frames per second, the baseline wanted to mitigate MFG inflicting devastating, swimmy-feeling lag? Will builders rush so as to add DLSS 4 to much more titles?
Nvidia clearly is playing on “yes” to the entire above. From what I’ve performed, DLSS 4 can ship heavenly, snappy gaming experiences which might be merely inconceivable with conventional rendering. But if Nvidia can’t persuade players that AI-generated frames are greater than merely “fake frames,” your entire GeForce RTX 50-series may wind up being considered as a flop – as a result of as we’ve already seen from the RTX 5080 and 5090, Nvidia guess this technology’s future on AI and DLSS, not conventional rendering.
Further studying: AI upscaling killed native graphics gaming. We’re better off for it