Nvidia and DLSS5: Or, We Used to Make Things

A week ago at time of writing, NVIDIA released the following demo. If you haven’t seen it, it’s like a minute long and it broke the internet for a few days. Watch it.

For the uninitiated, I will wildly oversimplify where we’ve come from and what we’re seeing. DLSS, or Deep Learning Super Sampling, is a feature of some of NVIDIA’s gaming graphics hardware. It uses machine learning to improve a game’s resolution, frame rate, and latency, per NVIDIA. Other vendors offer similar technologies, but with NVIDIA recently hitting 95% market share for gaming GPUs, it’s barely worth talking about anyone else. And nobody else is embracing generative AI in their tools like NVIDIA is here, putting what I’ve seen called a “slop filter” over the rendered image and “yassifying characters.”

I feel like I should disclose a bias here: NVIDIA has been in the doghouse with me for a long time. They were once a consumer-focused hardware company and then they weren’t. Now, they seem to be content to be market leaders in indulgence, creating world-class picks and shovels for the most cynical, terminally-online techno-fetishists. First, it was being content for consumer inventory to be snapped up by capital-rich speculators to build cryptocurrency mining rigs, only to begin to pivot back after inventory was almost unobtainable as the bloom came off that rose – and lying to their investors about where their GPU sales were going, to the tune of a $1B class-action suit. Now, they’re at the centre of a bizarre ourobouros of circular deals that NVIDIA, OpenAI, Microsoft, Oracle, and others have created, inflating the AI bubble and hoping that antitrust or the grim reality of the meagre productivity gains and cognitive issues related to intensive AI usage don’t pop the whole damn thing.

Also, DLSS5 looks bad. This is my opinion. I do not like how it looks. I think it looks like trash.

As I was loading the dishwasher this evening, I began listening to a long segment from a recent Linus Tech Tips podcast where Linus, Luke, and Riley talk about DLSS5, the backlash from gamers, and different perspectives on the tech and its use case. You can watch or listen to the whole thing here:

This conversation addressed a lot of the usual topics that you get in any generative AI conversation.

  • It looks bad.
  • It is very technically impressive.
  • It removes artists’ original intent.
  • It is more uncanny than the original art.
  • It ignores critical things about context and over-polishes everything.
  • It appears to be wildly resource-intensive (using two RTX 5090s currently).
  • It could be a tool to remaster old games.
  • People may want to make their games look and feel the way they want to (e.g. like how we use mods today).
  • “Good” graphics don’t make for good graphics – there’s artistry, taste, and style that produces results people like.
  • It’s extremely weird how it makes the female character “hot” as its first example.
  • Companies relying on this technology will make the gaming industry harder to enter, harder to work in, and even more competitive.
  • The games in the examples don’t even look like the same game after applying DLSS5.

I admit, I got a bit frustrated. I wanted to reach through the screen and grab my fellow Canadians by their cute little beards and whisper, “this isn’t what it’s about. This isn’t what NVIDIA is about any more. It’s not what they’ve been about for a very, very long time.”

And then finally, at around the 21 minute mark, Linus says this:

Let’s take this for what it is. It is NVIDIA’s marketing for a technology that they wish the industry to adopt, because it will make gamers and the entire industry more dependent on their products they sell.

Linus, The DLSS 5 Backlash is INSANE – LMG Clips. Retrieved March 23, 2026

And this is my problem with this DLSS5 thing. I think it’s marketing.

I think it’s fake.

I don’t think it’s fake in that it doesn’t work. I think it works as advertised. But I don’t think NVIDIA much cares if it becomes the next big things in computer graphics, and I definitely don’t think they care if it helps gamers or developers or artists. I think it’s for investors. I think it’s for NVIDIA to sell new compute to publishers who will pay NVIDIA money for training and processing and push it onto their studios to use. If trends continue, like most AI tools, uptake will be minimal, and people will be resistant to use it outside of edge cases. I think it’s entirely to keep this bubble inflating a little longer. And nobody liked it. Backlash forced NVIDIA’s stock down for a minute, but at time of writing it’s rebounding and trending upwards again. The bubble reinflates.

And this brings me to the core question of this piece. NVIDIA made an ostensibly fake thing, which is undoubtedly very expensive and very powerful, for almost nobody to do almost nothing with. Who is making actual things for real people to do real work with anymore?

According to Amazon’s earnings report, AWS is more profitable than Amazon’s retail business and both AWS and Amazon’s ad business are growing faster than retail.

Microsoft used to make devices and software that felt designed for me, specifically, in the Surface and Windows Phone. And now almost the whole Surface line has been retired, the remaining hardware is all B2B-focused, they’re having to grovel and promise to stop pushing Copilot into every inch of Windows, and take it out of Notepad. Notepad.

To not let Microsoft off the hook for one second, their most-successful consumer brand, Xbox, might be circling the drain as well following Phil Spencer’s departure. One of their AI execs is now in charge. It’s a good thing that they just recently bought Activision Blizzard, Obsidian, Bethesda, and so many other big, beloved gaming brands, if Xbox is getting sent off quietly into the night.

Even my user-serviceable, user-upgradeable Framework laptop has been impacted by this trend – not that it’s Framework’s fault. My laptop is currently humming along on 32GB of RAM from Crucial. And Crucial’s out of the consumer business because it’s so much more profitable to sell to AI data centres, sending RAM prices skyrocketing in what Tom’s Guide calls an “AI-driven pricing crisis.” I will not be repairing, replacing, or upgrading anything as long as prices stay this high, and if I’m having that thought, I have to imagine Nirav and the other Framework execs are shaking at the prospect of their whole install base feeling similarly.

Meta (née Facebook) recently killed the studio behind my favourite VR app, Supernatural, in a panic about its over-spending on the Metaverse as it pivots to AI. This was a murder so foul it brought former FTC chair Lina Khan back into the public eye to call out “a dominant platform eliminating choice by buying up the competition and then abandoning the market when corporate priorities shift.” Heck yeah, Lina.

Raspberry Pi. I feel like I don’t need to say more for most makers, but here’s the bottom line: availability has been fraught for a long time and now they’re subject to the same RAM shortage as everyone else. Those boards are now way too expensive.

One by one, big players pivot away from consumers to focus on B2B. Or smaller players get squeezed. When these players are so dominant, when they have purchased so many other companies, when there’s so much consolidation, when so much wealth rides on them continuing to grow on valuations that already make no damn sense, when they have sucked so much energy and so many resources out of the consumer marketplace to do so – I worry that there simply isn’t room for new consumer players to emerge. What marketplace can they emerge into?

Who is left making things for people?

Who is still making hardware at scale? Who can continue to democratize hardware? To empower people to make art, to build businesses?

I hope that we’ll find that nature abhors a vacuum and other players step in, but it’s tough to be a people-oriented, values-driven player in a system that not only rewards but forces growth and scale.

It didn’t used to be like this. Remember when Nvidia tech demos delighted us? Remember the GeForce 3 chameleon? Twenty-five years ago?

The chameleon was made to showcase tools that were made for artists. It was made to make consumers dream about what could be. It made us excited about the future.

DLSS5 is for nobody and makes us dread what’s next.

It feels like as a society, we used to make things for people. Now, increasingly, we don’t. We make them for markets, for investors, and for the narrative that will keep the bubble inflated just a little longer. We make things that are fake, for nobody to do nothing with.

Also, those things? They look bad.

As always, folks, paddle your own canoe.

– Trevor

Related posts

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.