SerpApi.Org
On the surface, Nvidia’s newest lineup of budget-centric GPUs appears to offer a massive performance bump over the past two generations, at least on paper. Take a look at the fine print, and almost all of these benchmarks account for some rather heavy upscaling, often with frame generation enabled.
It’s exactly this very convenient illusion that irks me to no end. It’s technically a sizable boost, but the implementation has so many shoddy drawbacks that seem to be glossed over in an attempt to prove the “bigger number better” trend.
A look at Nvidia’s current budget lineup of the RTX xx50 series
Long gone are the days of proper budget cards
Nvidia xx50 cards used to be the king of value, with the likes of the GTX 1650 Ti and 750 Ti coming to mind. It was not unreasonable to expect an xx50-series Nvidia GPU to easily run a few of the more demanding games at the time, with reasonably playable frame rates.
Take into account that these cards were used to natively render games with no DLSS or frame-generation trickery, and it becomes a lot more impressive. Upscaling just wasn’t as egregious back then, often looking absolutely horrid with muddy textures and poor antialiasing. Rendering at native resolutions was the expected approach.
In comparison, modern RTX GPUs rely on the whole smoke-and-mirrors approach. Yes, the latest and greatest RTX 5000 series cards support a more modern implementation of DLSS Dynamic Multi-Frame Generation, but at the end of the day, you're still using tricks to make up for marginal improvements in raw rasterization performance.
Generational improvements between the RTX 3xxx and 5xxx series have been painfully mediocre at best. The only major upsell of the RTX 5xxx series is in the fact that they run the latest and greatest DLSS tech from Nvidia, but in terms of raw performance, they are quite close, making it have some shockingly poor value.
DLSS and, more importantly, frame generation technologies might have caused more harm than good
When tools become chains
Let’s get one thing straight: DLSS is some pretty impressive technology, and there’s no denying that. It has many legitimate use cases, and perhaps the most important is its intended purpose.
This is to help run games just a little bit smoother on weaker hardware. It’s not exactly an apples-to-oranges comparison, but the custom DLSS being used within the Nintendo Switch 2 is largely responsible for making games playable on it in the first place.
DLSS was just a way to get some free performance out of an aging card, and not a crutch.
The industry seems to be heading in a very weird direction, choosing to use upscaling to offset poor performance and optimization issues.
It doesn’t help that Nvidia itself seems to be pushing both DLSS and DLSS frame generation quite hard, with a special emphasis on the latter. Nvidia did quite famously compare the 5000 series with the 4000, but these benchmarks never quite told the whole truth.
With multi-frame generation, a 5070 can outperform even a last-gen RTX 4090, but there is a huge catch. Frame generation inherently introduces a degree of latency, distortion, and ghosting, which is very noticeable.
Even more so when you enable 4x or higher frame gen, where everything turns into a hallucinated, janky mess.
It just comes off as disingenuous, and frame generation (or even upscaling, for that matter) introduces far too many variables to be considered as a straight-up FPS boost. Fake frame counts are not indicative of real-world performance, and you will most definitely notice the input latency, which, depending on the game, can make it entirely unplayable.
No amount of upscaling and frame generation can offset weak hardware
Higher raw raster performance always helps
DLSS is good. Very good, in fact, but not black magic. No amount of upscaling can make up for lower-end hardware, and you are bound to run into problems sooner or later. This is especially apparent with the more budget-focused cards, which have 8GB (or in some cases, 6GB) of VRAM.
8GB of VRAM is unfortunately not enough to run modern titles anymore, and you’re bound to run into a bottleneck, followed by some rather nasty stutter.
Sure, the alternatives are a bit pricier (especially given modern hardware pricing), but you do get a sizable performance boost without having to rely on DLSS or frame generation to achieve the same.
It’s also a far superior gaming experience — especially in something that’s a lot more fast-paced like a first-person shooter. The used market has decent enough deals, and I can’t personally recommend gunning for the latest RTX 5xxx series at all. The last-gen RTX 4xxx cards hold a lot more value in my eyes. Or you could consider switching to team red.
I'm not very hopeful for the future
I wish I could end things on a more positive note, but given the way things are going in game development, I’m not sure I feel very hopeful. We have conformed to the assumption that games need tons of VRAM to run properly, despite the rather mild and disappointing improvements to in-game visuals.
It doesn’t help that most budget cards still seem stuck in that weird 8GB middle ground, where they have barely enough VRAM to run modern games.
