The RTX 2070 Super is uncomplicated. It's a product of a far simpler time. That's what the Turing architecture, raster performance, and CUDA core count suggest. Released to perform in an era where AI features didn't dominate every launch presentation and rasterization was still king, Nvidia's upper midrange card from 2019 makes every enthusiast on a 1080p platform ask the same question, even in 2026. Is an upgrade now worth it?

And sure enough, seven years later, it's still a difficult question to answer. In part, because DLSS 4.5 gave it a new lease on life, and in part, because it continues to handle modern games and most productivity tasks you throw at it with surprising confidence. Owning a PC with a 2070 Super makes one feel that there's much to gain with an upgrade, but not a lot to lose even if they decide not to. Here's every reason why the GPU makes upgrading a choice, and not a necessity.

It's still an excellent 1080p GPU today

60-90 FPS in AAA titles isn't a GPU that needs replacing

One of the strongest arguments in favor of keeping the RTX 2070 Super firmly planted in my second PC's PCIe slot also happens to be simple and uncomplicated, and it's the mere fact that it performs remarkably well at 1080p. Having tried multiple modern AAA releases this year and the year before on the card, I can confidently say that titles like Black Myth: Wukong, Pragmata, and Forza Horizon 6 remain playable and enjoyable without requiring major compromises to the experience, provided one does not care far too much about ray tracing; something that first-gen RT cores were never particularly good at anyway. Anyone who has ever owned an RTX 20-series card is probably privy to that fact.

For the overwhelming majority of games in my library, the GPU continues to do a phenomenal job, because Nvidia didn't cut any corners when it came to raw rasterization in this generation, and it shows till date. There are, of course, exceptions to this, particularly among poorly optimized PC ports, but that's where community-developed tools such as OptiScaler and Lossless Scaling earn their keep. For the vast majority of the titles that I've emulated on my two rigs, any port that required running Lossless Scaling on the one with the 2070 Super also required it on my workstation housing the 4070 Ti Super, and the latter is a card with 16GB of VRAM.

Upgrading to another card with 8GB of VRAM doesn't make much sense

Modern GPUs are still shipping with this configuration

A conversation surrounding a GPU can't be complete unless the VRAM is considered, and I'll admit, that's where RTX 2070's biggest weakness lies. The dependence of many newer AAA titles on AI-driven enhancements like DLSS and frame generation has exposed that limitation, and if you take into account ultra texture packs and any amount of ray tracing, you're looking at the very definition of a VRAM-induced bottleneck.

But to simply argue that the 2070 Super is a bad card due to the VRAM constraint alone would be misguided, especially when you consider that many other recently released GPUs haven't moved beyond that constraint either. Whether it's an RTX 4060, RTX 5060 or a Radeon RX 7600, buyers shopping in the entry-level and lower end of the midrange segments can often be seen making similar compromises as they would on a seven-year-old Turing card. If I'm already comfortable dialing back texture settings or reducing ray tracing, there isn't a great deal separating it from newer hardware at similar price points. That's a thought that makes the 2070 Super easy to live with.

DLSS 4.5 completely changed the value proposition

Not even Nvidia is counting the Turing card out just yet

Perhaps the biggest surprise of owning a 20-series GPU came recently, when Nvidia signaled their intent to invest in software improvements for Turing cards by releasing the second-gen transformer-based DLSS 4.5 for the lineup. If you've read my past coverage, you'll know that I ran the benchmarks on my RTX 2070 Super to see whether the new transformer model would overwhelm the aged Turing architecture, especially when there was every reason to believe so. Nvidia itself warned that older GPUs could experience a heavier performance penalty because they lacked native FP8 processing as opposed to the newer GPUs.

My testing, however, indicated quite the opposite. Despite the additional computational demands, the performance overhead remained manageable, while image quality saw a dramatic improvement. Games greatly benefited from the added temporal stability and appeared richer. As a matter of fact, I almost couldn't distinguish whether I'd been using my Turing card or the Ada Lovelace GPU in the testing at 1080p, and that's not really something you hear anyone say about a card that's as old as the 2070 Super.

A good GPU doesn't suddenly become a bad one

Seven years later, the RTX 2070 Super, and to an extent the 20-series family, still finds itself to be relevant. A well-engineered card is a well-engineered card, no matter how old it gets, but that's something the GTX 1080Ti taught the market anyway. Between respectable 1080p performance, continued software improvements and the card's ability to leverage them, Nvidia's old Turing card assumes the spot of the old and reliable GPU that's not afraid to take on the AAAs. At least, not today.