The Nvidia GeForce RTX 5090 and RTX 5080 are real, but they're not as spectacular as many gamers initially expected – at least not if you don't use DLSS 4. 😕

I'm worried about what that means for the $550 RTX 5070 PC that so many gamers have been waiting for, and whether the entire RTX 50 series could end up being a failure if the incredible new Multi Frame generation of DLSS 4 fails to take off fast. (I should!) And I have the numbers to back up my fears. 📈

The RTX 50 series, powered by “Blackwell,” doesn’t offer much of an improvement in value or rendering performance from generation to generation, plain and simple. The GeForce RTX 5090 dominates the benchmark charts thanks to Nvidia equipping it with 33 percent more CUDA cores than the RTX 4090, allowing it to achieve 27 percent faster speeds at 4K resolution – in exchange for an 18 percent increase in price and a a lot elderly energy consumption. It's definitely brutal and powerful, but meh for a new generation. 🔋

The RTX 5080 is even less impressive, offering just 15 percent more performance than the RTX 4080 Super it replaces at 4K, and a dismal 11 percent improvement at 1440p. Plus, it still only has 16GB of GDDR7 memory despite costing $1,000. Meh. 😒

Of course, this has to be balanced against the huge performance gains offered by the new DLSS 4. Nvidia, and more specifically, its Multi Frame Generation feature. MFG inserts up to three AI frames between two “traditionally” rendered frames to make frame rates and visual fluidity skyrocket. 🚀 The frames it generates can cause games to struggle, such as Star Wars Outlaws, feel as agile as the legendary Doom 2016As I mentioned in our RTX 5090 review, it's a true, literal game-changer, and will run 75 games and apps right out of the box. 🎮

But the RTX 5080 proves that Nvidia is willing to stake the existence of an entire graphics card on the promise of DLSS 4. And the RTX 5080 isn't alone. 🤑

Concerns about the RTX 5070 series

A little after Nvidia revealed the Blackwell generation, we take a look at the RTX series specifications 50 to try to estimate where the “traditional” gaming improvements will fall. We almost nailed it, guessing that the RTX 5090 would end up being about 33 percent faster (it ended up being 27 percent) and that the 5080 would be somewhere between 10 and 15 percent faster (really: 11.5 to 15 percent). 🤓

How do we achieve that? Simple: we compare their counts cores CUDA processors with their predecessors, leaving room for more power and faster clock speeds. Blackwell GPUs are manufactured using the same TSMC 4N process, so there's no obvious increase due to a jump in transistor node. And rather than prioritizing performance improvements traditional GPUs, Nvidia focused much of its attention on powering the RTX 50-series Tensor AI cores to accommodate DLSS 4's delightful Multi Frame Generation feature, as well as future technologies like Neural Rendering and Mega Geometry. 🔧

Again: DLSS 4 is much more than just “fake frames”. It has dramatically improved the fluidity of the gaming and smoothness in our gaming tests in a handful of titles that were made available to press ahead of launch. AI upscaling is winning the battle over native rendering, and it makes sense for Nvidia to prioritize it in a generation where AMD's Radeon team has already admitted it's not chasing high-end performance. 💪

However, if the dream of DLSS 4 doesn't materialize – if the other 75 launch games show more pronounced visual artifacts or glitches than we've seen, if gamers balk at the extra latency that AI frames require, a million other "ifs" – the RTX 50 series could turn out to be one of the most disappointing GPU generations in recent memory. 😱

Adam Patrick Murray / Foundry

We have already seen how core counts CUDA correlates with RTX performance 5090 and 5080. That bodes poorly for the RTX 5070 series, despite being pegged at $50 less than before. ❌

The same CUDA comparisons project the $750 RTX 5070 Ti as the “star” of this generation, with an increase of approximately 17 percent over its predecessor, in addition to an improvement in memory capacity, bandwidth and speed that could help further increase performance. A ~20 percent increase from gen to gen is still pretty meh in terms of progress, but it's a more tangible improvement. 🔝

The reception for the $550 RTX 5070 could be even cooler if our estimates hold up; that GPU only has 4 percent more cores than the existing 4070. It also keeps the same 12GB of memory as its predecessor, albeit of the faster GDDR7 variety. We’ll have to wait to see how it performs in real life when it launches later this spring, but the 5070 has the potential to make the RTX 5080’s minimal generational gains seem comparatively massive. 😮

Nvidia hasn't announced any low-end GPUs yet. What we've seen of the RTX 5070 laptop GPU makes us nervous about the RTX 5060, as a graphics card with just 8GB of VRAM in 2025 just isn't enough. We can't say too much about that until the card is announced and tested. ⏳

What's clear is that Blackwell's traditional rendering performance isn't moving the needle forward much. The RTX 50 series was designed for an AI-assisted future and it will depend on how DLSS 4 is received. 🔮

Betting everything on DLSS 4

Great news! I absolutely love the multi-frame rendering of DLSS 4 in practice. My fellow PCWorld video team members Adam Patrick Murray and Will Smith are also excited about it. If you want to hear our more in-depth thoughts, check out our lengthy discussion of DLSS 4 in the time-stamped video below. 🕒

But make no mistake: Nvidia's entire RTX 50-series lineup is betting that DLSS 4's massive AI-powered gains in smoothness and frame rates will be welcomed by PC gamers. And that's certainly a bet. 🎲

DLSS 4 impressed me during testing, but many PC gamers disparagingly call AI-generated frames “fake frames.” Generating a single frame RTX Series DLSS 4 Frame 40 faced devastating reviews at launch due to visual issues and latency concerns. I've seen far fewer of those issues in my time testing with Multi Frame Generation – or perhaps they just happened so quickly that my eyes couldn't notice them, rendering at incredibly high frame rates. ⚡

We've only had access to a handful of MFG-equipped games, and the previous generation, DLSS 3, could be very hit and miss depending on what game you were playing. (Microsoft Flight Sim, for example, notoriously had issues with glitchy text). Making DLSS 4 available in 75 games and apps at launch will offer a lot of software in the hands of gamers to test, which is excellent and a big step forward compared to previous standards, but it remains to be seen whether they all offer the same impressive performance than Cyberpunk 2077, Star Wars Outlaws, and Alan Wake 2. 🕹️

Adam Patrick Murray / Foundry

Enabling DLSS 4 in all of those games won't necessarily be a no-brainer. While some games are being updated for DLSS 4 by developers, many of the 75 games and apps will require you to enable “DLSS Override” options within the Nvidia App, which tells the software to replace older DLSS 3 files with the new DLSS 4 capabilities. 🔄

Will gamers be able to do it? Will games with DLSS Override render as smoothly as native DLSS 4 titles? Will lower-end cards like the RTX 5070 and 5060 offer enough performance to hit 60 frames per second or more, the bare minimum needed to mitigate the lag that MFG causes, which feels devastating and dizzying? Will developers rush to add DLSS 4 to even more titles? 🤔

Nvidia is clearly betting on a “yes” to all of the above. Based on what I’ve played, DLSS 4 can deliver heavenly, super-snappy gaming experiences that are simply impossible with traditional rendering. But if Nvidia can’t convince gamers that AI-generated frames are more than just “fake frames,” the entire series is in trouble. GeForce RTX 50 could end up being seen as a failure – because as we've already seen with the RTX 5080 and 5090, Nvidia bet the future of this generation on AI and DLSS, No in traditional rendering. 📉