The GeForce RTX 4070 officially turned three years old in April, and yet it seems like just yesterday that the Ada Lovelace GPU was announced. This is not nostalgia, far from it. Rather, it seems that there has been no significant change in the experience of owning a GPU, there is no lack of features, and furthermore, there is no reason to consider a possible upgrade in the last three years.
DLS 4.5 brought the second generation Transformer model which fixed most of the shortcomings of super resolution, while Multi Frame Generation took care of the frame output ceiling. That all of this is available on a card released three years ago could mean that Nvidia has created a truly future-proof line with RTX 40 seriesbut it could also mean that consumer hardware is no longer seeing innovation at the speed of light. Either way, an update is not in the cards, no matter how you approach the argument. However, that’s not great news. Here’s why.
Ada Lovelace could be the Pascal series of the 2020s
Turing and Ampere did not age in the same way
Looking back, it’s almost funny to see how many of the features that are integral to our gaming experience came unrefined. When real-time ray tracing debuted with the Turing range in 2018, Cyberpunk 2077 It brought the GPUs that launched with it to their knees, to the point that most people without an RTX 2080 Ti wouldn’t even enable it during gameplay. Ampere GPUs improved the experience slightly, but it wasn’t until the trifecta of Ada Lovelace, DLSS 3, and third-generation ray tracing cores finally made it usable.
On the topic of super resolution, DLSS followed a similar path. Subsequently, each generation perfected the trajectory until it became It’s hard to imagine modern AAA games. without him. Every feature that Turing and Ampere introduced in a rough proof-of-concept form was solidified with Ada, and it’s hard to argue otherwise. The 40 series It represented the fusion of the flagship features of GPUs released five years earlier, with an architecture advanced enough to deliver those features without compromising key aspects of the user experience.
DLSS 4.5 was the final piece and the 40 series was ready for it
The most complete GPU software stack, on purpose-built hardware
When DLSS 4.5 came out this year, it initially seemed like a win for all RTX series GPU owners. As Nvidia’s keynote at CES 2026, who could be excited about the second-generation Transformer? The promises were pretty big and worth a second look. Improved temporal stability and expanded access to Multi Frame Generation sounded exactly like what the successor to DLSS 4.0 needed. It felt like Nvidia opened up and went through the fifty thousand Reddit threads complaining about DLSS, ghosting, smearing, foliage glare, artifacts, and input lag, and addressed every single complaint enthusiasts had that made them avoid DLSS like the plague.
Then, the promises half met expectations. While Older RTX cards could “technically” support In the new feature set, only two GPU families possessed the native FP8 capabilities that allowed Transformer models to run without massive overhead that nearly destroyed gaming immersion. The first, of course, was the Blackwell series. The second was the Ada Lovelace lineup. Despite being released 3 years before the second generation Transformer, the 40 series did not need a complete architectural renovation to deliver the temporal stability and image quality improvements that made DLSS 4.5 worthwhile. It now appears that, from the day it was built, it was built to last.
For the first time in the history of consumer charts…
The future seems less exciting than the present
The RTX 40 series sits at a unique intersection that is difficult to manufacture by design. By this I mean that in this generation, raw raster performance remains on par with current titles, a full DLSS stack runs natively on the hardware it was built for, and MFG access exists without the ecosystem dependency that makes the Blackwell series seem premature. It is the last generation of GPUs in which the hardware itself took center stage, rather than software features built around its limitations.
For all those reasons, it seems like Ada Lovelace is the sweet spot between raw rasterization capabilities and AI-driven advancements in rendering. The same cannot be said for Blackwell and the coming era of neural rendering.
In this paradigm, Blackwell seems to arrive at the beginning of a structural change in rendering processes that software itself has not achieved. Perhaps the RTX 60 series will be designed to take full advantage of those advancements, but for now, the GPU family that offers the most value per dollar and the perfect combination of usable features and rasterization capabilities is, without a doubt, the Ada Lovelace.
The 50 series has been a tough sell and the 40 series is the reason why
For as long as I’ve been building PCs, new generations of GPUs have always seemed exciting to me because they promised something that my current hardware simply couldn’t do. With Blackwell, the feeling seems to have disappeared. The RTX 40 series is in a place where the hardware is still powerful, the software stack has continued to improve three years later, and the compromises are few enough that upgrading feels like a practice for chasing diminishing returns. As an enthusiast, that’s reassuring. As an analyst, it is unfortunate to witness that.
- Shading units
-
6144
- Accelerators/ray cores
-
48
