In addition to announcing the GeForce RTX 40-series (Ada Lovelace) graphics cards, Nvidia also revealed the company's latest DLSS 3 technology. With the chipmaker claiming up to a 4X performance boost, DLSS 3 obviously received a lot of attention.
The million-dollar question is whether DLSS 3 will come to older GeForce graphics cards — and Bryan Catanzaro, Vice President of Applied Deep Learning Research at Nvidia, has answered the question on Twitter (opens in new tab).
Catanzaro confirmed that DLSS 3, in its current state, only works on the GeForce RTX 40-series graphics cards. The Nvidia engineer did leave the door open for backwards compatibility, stating that it's theoretically possible to get DLSS 3 working on previous generation GeForce graphics cards, such as the RTX 30-series (Ampere) or RTX 20-series (Turing) offerings. However, Catanzaro believes that DLSS 3 won't offer much benefit for owners of legacy GeForce graphics cards.
Nvidia's DLSS 3 technology relies on the Ada's fourth generation Tensor cores and the new Optical Flow Accelerator (OFA) to work. OFA isn't some new invention that Nvidia recently cooked up with Ada. OFA has been present since the Turing days. The difference is that Nvidia has tremendously improved it in Ada versus Ampere, and it now delivers higher performance and better quality. The OFA in Ada is reportedly 2 to 2.5 times faster than on Ampere. Nvidia has seemingly made some algorithmic changes as well.
Ampere and Turing can theoretically leverage DLSS 3, but they won't yield the same benefits. Catanzaro said that DLSS 3 likely won't boost frame rates on Ampere or Turing — on the contrary, owners would probably experience laggy gameplay and bad image fidelity.
DLSS is known to add some degree of latency. This is the main reason why DLSS 3 integrates Nvidia Reflex with the aim to deliver higher frame rates at the same system latency. Nvidia claims that DLSS 3 is able to reduce latency by as much as 2X when compared to the native latency. In a demo (opens in new tab) of Ray tracing: overdrive mode in Cyberpunk 2077, DLSS 3 in conjunction with DLSS Super Resolution, DLSS Frame Generation, and NVIDIA Reflex showed 4X higher performance and 2X improved responsiveness.
DLSS 3 has proven to be beneficial in CPU-intensive games such as Microsoft Flight Simulator. Nvidia's demo (opens in new tab) showed DLSS 3 doubling the frame rates in Microsoft's title, which heavily taxes the processor.
DLSS 3 will debut on October 12 alongside the flagship GeForce RTX 4090. Nvidia expects over 35 games and applications to support DLSS 3 at launch. More importantly, DLSS 3 builds upon the foundations of DLSS 2 so game developers can easily enable DLSS 3 on current games that already embrace DLSS 2 or Nvidia Streamline.
Neat. Good use of resources, Nvidia, thanks for passing that cost onto your customers.
I look forward to when it gets added to Control, so I can enjoy shill youtubers acting like we will literally die without it.
I can already see the clickbait thumbnails in my head: Poorly groomed, poorly photoshopped millennial millionaires making a goofy face in front of an ultra-saturated primary color background. An image layout that is algorithmically engineered to sell product to children.
Title: "We weren't ready for this ..."
Delightful. Not tired, predictable, or irritating at all.
Nvidia knows they are price gouging on the next gen and needs more bait to get fishies to bite :D
Of course, finally now that a lot more have bought 3000 series after 2 years of unavailability, the best move is to lock newer features for newer cards, since there are no technical details on the surface it definitely sounds like Nvidia being a scumbag, like always.
Also, YouTube does not allow people to find content made by small channels. In the rare cases people do find your small channel, YT will immediately autoplay into a "related" video from a big channel over the other videos from the series. Most non-YouTube sites get pretty mad when you try to give people an external way to find your content. So I can't even directly link anybody to what I'm talking about.
It wouldn't matter if I had a complete exclusive to beat people to the trend. The first "preferred partner" to repost, react, or comment about that exclusive would likely outperform my original video by at least 10,000 to 1. People may even credit that celebrity influencer for having the exclusive.
I actually did the math recently. If my time investment had been paid like my engineering salary, Then would be close to paying YT around $200 per each of my ~750 subscribers. Close to $150k in labor over 18 months (mostly editing) with $5-$10k invested in equipment, software, building out my studio space, etc. That investment does not show on screen, and it has not brought in even a single cent in revenue.
I would be closer to breaking even if I had bought into a literal Ponzi scheme.
I can afford an expensive life-consuming hobby like that, but I feel really bad for the kids who think hat they can just grab a camera and turn this into, like, an actual job. It doesn't work that way, especially post-covid. YouTube requires you to be an incredibly well-connected insider with a team of people to support you, just like Hollywood or anywhere else in the big money entertainment/celebrity/arts industry. One of the most common way to get that support, is by being directly and indirectly paid by the companies selling the tech products you review.
Bringing me back to my original point: any YouTube reviewer that anybody has ever heard of is just shilling for money and access. They all do slightly different versions of the same mental gymnastics to justify it to themselves, but it still is what it is.
Nvidia buys a heckton of ad reads on many channels, so you'd better believe these people are going to treat their new expensive tech like the second coming of GabeN, even though it won't even be available to 99.9% of gamers for the forseeable future. Games occasionally support current DLSS 2, but DLSS 3, I'm not so sure.
I see and understand you are passionate about content creation, however, not everyone that is a reviewer on youtube is shilling for companies. For instance Gamers Nexus is reasonably independent. There are not many who are as good but Project Farm is also a relatively unbiased reviewer of tools and the such. Many things you have said are in general true, but also not necessarily fair to what people have made their channels into. People only do things in a certain way because it works and they are incentivized to do so monetarily. The idea that you need to have contacts and luck to succeed on Youtube or even Twitch is not true as a whole. Certainly it helps to get lucky with a collaboration or you were in the right place at the right time with a video, but that is not the whole picture. I believe Ludwig made a separate YouTube channel anonymously to prove as much and it did well.