Las Vegas, NV - The Consumer Electronics Show (CES) wrapped up recently, and one thing was noticeably absent from Nvidia's presentation: shiny new GeForce "Super" GPUs. Instead of unveiling the expected mid-cycle refresh, CEO Jensen Huang doubled down on software, highlighting the company's ongoing efforts to squeeze more performance out of existing hardware.
Nvidia Abandons GeForce Super?! What This Means Fo...
For those of us who've been covering the GPU beat for a while, this was a bit of a surprise. Nvidia typically uses CES to announce something exciting in the graphics card space. This year, however, the focus was squarely on software enhancements, primarily to their Deep Learning Super Sampling (DLSS) technology.
The headliner was undoubtedly DLSS 4.5, the latest iteration of Nvidia's upscaling and frame generation magic. The big improvement here is a new "second-generation transformer model." What does that actually mean? Essentially, it's a smarter algorithm trained on a massive dataset, allowing it to predict missing pixels more accurately. According to Nvidia's Bryan Catanzaro, this is especially crucial in "Performance" and "Ultra Performance" modes where the upscaler is doing a lot of heavy lifting because the source image is lower resolution.
DLSS Multi-Frame Generation (MFG) is also getting a boost with a new 6x mode. This, coupled with Dynamic Multi-Frame Generation which adapts the number of generated frames based on your system's performance, sounds promising. However, let's be clear: MFG still requires a relatively modern GPU (think RTX 50-series for the full experience, 40-series can only generate one extra frame). Older cards are left out in the cold. More importantly, it's really meant for games that are already running at decent frame rates; it's not a miracle cure for slideshow-like performance. I've seen it firsthand – it's great for adding a little extra smoothness, but it won't turn a stuttering mess into a silky smooth experience.
The good news is the new DLSS 4.5 transformer model is compatible with older GeForce 20- and 30-series cards, giving those users a taste of the improved upscaling. The catch? Nvidia warns of a potentially greater performance impact compared to the newer 40- and 50-series cards. And early benchmarks seem to confirm this. Some outlets have reported performance dips on an RTX 3080 Ti when using DLSS 4.5, ranging from 14% to 24% compared to the previous DLSS 4.0 model. Still, even with that hit, performance is generally better than running games at native resolution, so it's a worthwhile trade-off for many.
One thing that was conspicuously absent was any mention of Reflex 2, the updated input lag reduction tech unveiled at last year's CES. Considering that upscaling and frame generation inherently add lag, Reflex is a crucial part of the equation. Nvidia previously touted a potential 75% lag reduction with Reflex 2 on 50-series cards, but there were no updates on its release timeline this year. Perhaps we'll hear more about that later in the year. For now, it seems Nvidia is betting on software to carry the torch, at least until the next generation of GPUs is ready for prime time.
Comments
Please sign in with Google to post a comment
No comments yet. Be the first to comment!