Nvidia thinks its frame generation technology could work on previously unsupported GPUs without needing optical flow on the hardware side. RTX 30 Series users may soon have a new option to improve the smoothness of their games.
During a Digital Foundry interview with Bryan Catanzaro, Nvidia’s VP of Applied Deep Learning Research, the brand’s representative indicated that the updated DLSS core could allow frame generation on older hardware. Fantastic news for RTX 30 Series owners who may have felt abandoned after frame generation’s launch.
“This is primarily a question of optimisation and also engineering and the ultimate user experience. So, you know we’re launching the best multi-frame generation technology with the 50 Series and we’ll see what we’re able to squeeze out of older hardware in the future,” says Bryan.
Nvidia has bundled a new and updated version of DLSS with each passing GPU generation, from DLSS 1 on RTX 20 Series and DLSS 2 on RTX 30 Series to DLSS 3/3.5 on RTX 40 Series. With the launch of RTX 50 Series chips, the brand is once more releasing a better version of its upscaler called DLSS 4.
Compared to its predecessors, DLSS 4 brings a new transformer model to further enhances the image quality. It also opens up new avenues to solve some of the remaining problems/limitations of upscaling. But most importantly, since it runs now on Tensor Cores, it makes it theoretically possible to run on RTX 30 Series, thus frame generation too.
“When we built DLSS 3 frame generation, we absolutely needed hardware acceleration to compute optical flow. We didn’t have enough Tensor Cores and we didn’t have an optical flow algorithm that was good enough. We hadn’t developed a real-time optical flow algorithm that ran on Tensor Cores that could fit our compute budget,” says Bryan.
Nvidia continues its journey to improve the trifecta of real-time graphics which include image quality, smoothness, and responsiveness. This comes in the form of DLSS 4 upscaling, delivering better than ever image quality and stability, multi-frame generation pushing the preserved smoothness by four folds, and Reflex 2 / flip metering for a reduced latency.
The updated flip metering system has up to x5 less frame time variability, which is important for multi-frame generation and frame pacing. It also removes the CPU from the equation of when to present each frame.
Lastly, the base frame rate for a good multi-frame generation experience remains the same, preferably above 60fps. The higher it is, the fewer artifacts you would notice as the AI model accesses more vector information.
If and when Nvidia brings frame generation to its older GPUs, AMD’s counterpart could become redundant on GeForce cards. A potentially great time for players who will have the choice to select the solution they prefer.
