It’s funny how things in reality can be so far removed from what we imagined.
A classic example of this is how I imagined there to be a horde of scientists at Nvidia HQ hunched over their PCs and all working to train the next generation of Nvidia DLSS algorithms — between enjoying bouts of Call of Duty with colleagues, of course. But as it turns out that’s only part of the story…
Yes, there are scientists at Nvidia working on these projects, but doing a large portion of the work in training and developing new DLSS technology for us PC gamers to enjoy is also an AI supercomputer, and it’s been doing that non-stop 24/7 for going on six years now.
That nugget of information was delivered by Brian Catanzaro, Nvidia’s VP of applied deep learning research at CES 2025 in Las Vegas.
Catanzaro dropped that gem on stage casually as a throwaway comment while discussing details about DLSS 4. But as it turns out, that reference has been the catalyst for a ton of talk about the topic.
Further reading: Nvidia’s DLSS 4 is so much more than just ‘fake frames’
Just about every journalist in the room was left pondering it, judging by the number who’ve written about it since then (me now included). So why is it such a fascinating point?
Personally, I think it’s because we don’t know much about how supercomputers are being utilized in PC gaming. This information is understandably obscured by the games, hardware, and technologies that succeed their input — not by design, but simply because these are the things we directly interface with in our gaming experience.
Despite that fact, supercomputers are propping up the PC gaming ecosystem; they’re helping produce technologies like DLSS — heck, even providing code for the games themselves, but we know little or nothing about the specifics of where or how they’re being used.
The obscurity of that information could be because gaming devs don’t want to give all the credit away to the mere tools they’re using. After all, supercomputers are essentially just the reference textbooks they need to consult for physics simulations and coding. But that just means their impact to the industry is being diminished.
Nvidia DLSS 3 technology with RTX 40-series GPUs improves rendering by up to 4x over traditional rendering.
Pexels: Alena Darmel
That Nvidia should use a supercomputer to create technology as technical as DLSS 3 or DLSS 4 really shouldn’t shock anyone, though — it’s a fine piece of technology only the likes of Nvidia could create.
Most gamers know that DLSS harnesses AI super resolution and tensor cores in their GPU, but digging down to the nitty-gritty of what makes it all tick and I can’t help but feel awe.
Underlying the technology that multiplies rendering performance by up to 4x are a whole lot of complex mathematical computations. The DLSS 3 frame generation convolutional autoencoder is taking input from four inputs at once, for example, so it’s a masterful piece of software.
Further reading: Nvidia GeForce RTX 5070 Ti review: DLSS turns a beastly GPU into a gaming monster
With his comment about the supercomputer at Nvidia, Catanzaro brings all that home. He’s inadvertently given us a little glimpse of how important these devices are to PC gaming in 2025, and in the process, whet my interest and a whole lot of other people’s interests to know more.
So, as a callout to any devs reading this, I want to know more about how supercomputers are being used to simulate complex physics, weather patterns, and AI behavior to create realistic and immersive game worlds and the characters that are more believable.
Adam Patrick Murray / Foundry
Additionally, I want to know more about how supercomputers power cloud gaming services and the data centers that gamers connect to remotely — services like GeForce Now and Xbox Cloud Gaming.
I’d also be very keen to hear how supercomputers are involved in research and development of new gaming technologies, being used to train AI models and algorithms, and to do things like run more realistic simulations of VR and AR — making modern games a far cry from the indie games built by MS-DOS programmers in their basements in the 1990s.
Nvidia’s supercomputer fits somewhere into that last category of training, research, and development, according to Catanzaro. It’s a critical part of the teaching process for new generations of Nvidia DLSS, harnessing the power of thousands of graphics cards to forward the development, but it’s also constantly looking for failures and analyzing images for things like ghosting, flickering, and blurring too, Catanzaro said.
These failures are then fed back into the computer to develop better, more efficient models. You could say it’s on a mission to constantly improve image rendering, the kind that’s only going to reap big benefits for gamers in years to come.
Games physics is a category we can expect to see the biggest future gains from supercomputers, in my view.
Thanks to the efforts of supercomputers, we can likely look forward to things like: The creation of smarter physics algorithms that will speed up game development; more efficient offloading of complex physics computations to the cloud; and the development of next-gen particle simulations that will give us more realistic fire, air, and water in game environments.
It’s for those reasons — and many others I could mention — that we need more vigorous discussion about the role of supercomputers in PC gaming. The more talk and collaboration we have about these devices, the closer we’ll get to the zenith of PC gaming experience.