China is getting first crack at a British technology that might change how mobile games are made and played. Today, the Xiaomi 18 Fold launches in mainland China with an Arm Mali G2-Ultra NX graphics processor inside its custom Xring O3 chip. What’s so special about that? After five years of development, Arm now has an AI gaming graphics accelerator inside its GPU.

By following Nvidia’s playbook, Arm-powered phones are about to have PC-like technologies to make games run faster, at higher resolution, with advanced lighting techniques, all without using more power than they do today. The jagged edges you see in the mobile versions of popular games might begin to go away.

And these advancements won’t just be locked to Chinese phones. The chips, devices, and games are already inbound.

“You will be able to buy it; you will be able to get it in phones next year, you will be able to get it in other types of larger screen devices, whether that be tablets or Chromebooks,” Arm’s Chris Bergey, EVP of Edge AI, tells The Verge in an interview.

While he can’t say which other devices will contain the Mali G2-Ultra NX, he hints that some of Google’s upcoming Googlebooks might use this graphics tech as well.

“We think it’s going to be huge,” Bergey tells me.

DLSS for Android

There are two ways to look at Arm’s new Arm Mali G2-Ultra NX. One is that it’s stronger than prior mobile GPUs, period, with an alleged 85-percent performance increase over the GPU that Xiaomi was using previously, and a 14 percent increase over Arm’s previous flagship in “non-AI gaming.”

But the other, more important way is that it enables a suite of hardware-accelerated graphical techniques that sound very similar to Nvidia’s DLSS, including AI upscaling (where it imagines a higher-resolution image), frame generation (where it generates new “fake” frames between existing ones), and even ray reconstruction like Nvidia’s DLSS 4.5.

Arm is calling all of these techniques “neural graphics,” but it doesn’t mean quite the same thing as when Nvidia uses that term — we’re not talking about enhancing faces by reimagining how light and shadow work, at least not just yet.

Instead, these improvements are about speed, offering what Arm claims is up to four times the frame rate when “Neural Super Sampling,” “Neural Frame Rate Upscaling,” and “Neural Denoising” are all applied simultaneously, because each technique lets the system do less work rendering actual pixels and frames, counting on AI to fill in the rest.

Note that games likely won’t feel four times faster, though, even if they look that way. Half of the boost is because your phone will imagine new frames between existing ones — and when Nvidia, AMD, and Intel do “frame generation,” it typically comes with the cost of additional latency.

Regardless, these techniques won’t all arrive immediately. “The first titles will largely be super sampling,” says Bergey, with frame gen coming in early 2027 and denoising “soon after that.”

But when you put them together, he says a phone can now genuinely play an experience like this Neural Dawn technology demo, below — and without draining your battery fast. “It might be like, hey, you can now play 1080p, 60 frames per second for three hours,” Bergey tells me. “In the past, it was 30 minutes.”

With the Xiaomi Pad 9 Pro Max tablet, which pairs Arm’s GPU with a larger battery, screen, and surface area for cooling, the chip may offer even more performance. The Chinese company is showing off 3.2K resolution at 120 frames per second.

Bergey says that an entire phone with these chips will consume no more than 2.5 watts, with only 1.5 watts allotted to the GPU. “That’s basically what you want for real sustained gaming where it’s not getting warm in your phone, you’re not like ‘Oh, where’s my charger,’ that kind of thing,” says Bergey.

Or, he says, device and chipset makers can run this GPU at lower wattage to last longer. It’s not just for gaming phones: “This is not overclocking, this is mainstream,” he says. “This is about AI replacing traditional rendering; anywhere you can use it, you will use it,” he says.

It’s not just for premium phones and tablets, either: “While Mali G2-Ultra NX targets flagship smartphones, neural graphics will also be available across the Premium and Pro configurations, bringing the technology to a broader range of mobile segments and tiers.”

Arm isn’t the only one attempting to push these speed-boosting PC technologies into the mobile realm. The developers of the graphically intensive Stellar Blade, for example, recently revealed they’re making their game run at 60fps on the Nintendo Switch 2 by using both Nvidia’s DLSS and a version of AMD’s frame generation technique. The Switch 2 has an Arm-based mobile chip too, though it contains a custom Nvidia GPU.

Like with the most advanced versions of Nvidia, AMD, and Intel’s techniques, game developers will need to integrate Arm’s features into their games, though participating game engines like Unreal Engine and Unity can help. Bergey doesn’t think it’s a switch gamers will be able to flip for just any game they like.

Where Winds Meet, Infinity Nikki, and Arena Breakout: Infinite are the first three games confirmed to take advantage of Arm’s new graphics accelerator, and Bergey tells us more games are on the way. Some will be announced in the coming weeks, he says — he’s just not allowed to reveal them quite yet.

Officially, Nvidia’s controversial DLSS 5 AI rendering was supposed to launch this evening with only a single game, only on Nvidia’s latest RTX 50 GPUs, and with developers in full control of their artistic vision.

But unofficially, modders have already ported a leaked version of DLSS 5 to run on just about anything while cranking sliders to the max. Now, Nvidia is changing its plans. DLSS 5 is still launching first on RTX 50 GPUs, but it’s officially coming to older RTX 40-series GPUs as well, Nvidia spokesperson Rajal Maharaj confirms to The Verge.

“Our current focus is on optimizing performance for the GeForce RTX 50 Series with model updates expected later this fall. Once RTX 50 Series performance is more fully tuned, we plan to work on expanding official support to the GeForce RTX 40 Series,” reads part of Nvidia’s statement.

What about earlier RTX 30 and RTX 20-series GPUs, which modders successfully targeted as well, assuming you count “not exactly playable framerate and plenty of crashes” as successful? “Right now the focus is of course on 50-series,” Maharaj tells me. “Once we feel like the 50-series is in a good place, we’re going to look at the 40-series.”

Now that the cat’s out of the bag, will gamers get full control over DLSS 5’s sliders the way modders already do? Nvidia says no: “For gamers, DLSS 5 is a simple on-or-off toggle,” the company confirms. “The more granular artistic controls are for developers, including model selection, Structure Intensity, Tone Intensity, automatic masking and custom per-pixel masking.”

That said, the company tells me that developers could pass along certain controls in their own settings menus, like they’ve done with other graphics tech. (Early DLSS Super Resolution was also just an on/off toggle, but today’s games let you pick between varying levels of quality or even an exact percentage, depending on the game.)

Assuming you don’t mind DLSS 5, one of the benefits is that it’s easy to integrate. Developers don’t need to tell DLSS 5 which objects are “silverware” and which are “furniture” because the AI models can automatically recognize and isolate objects in a scene, distinguishing between faces, textiles, foliage, and more to apply different amounts of AI processing. Game engines that already know which video game objects are which can help DLSS 5, too:

For more targeted control, semantic AI masking can automatically recognize scene objects, while engine-level masking allows artists to isolate specific props or groups of assets, such as glassware, water droplets or foliage, and apply custom neural lighting adjustments to those elements.

Unreal Engine 5 has a native plug-in, and Nvidia also has SDKs.

Nvidia says developers can apply DLSS 5 to their entire game this way, or do it differently for different scenes, cutscenes, or portions of gameplay, and the company tells The Verge that its DLSS 5 models run on every single frame, so there’s no minimum frame rate for it to work.

We’re still waiting to find out if the early backlash against DLSS 5 scared away publishers and developers that had originally committed to the technology. Nvidia’s response to us today suggests that perhaps some companies have backed away.

When we asked if the company could reconfirm that it will come to Assassin’s Creed Shadows, Hogwarts Legacy, Starfield, Resident Evil Requiem, and the other 12 originally announced titles, this was the company’s only reply:

“More games are on the way. Many developers are actively integrating DLSS 5, and we’ll share updates on upcoming titles as partners finalize their integrations.”

Last week, I powered on a 26-year-old copy of Pokémon Pinball, one of the best Game Boy games ever made. My save file was no more. The Charmander I’d caught for my daughter, the high scores we’d set together… gone. Before flash memory, Nintendo cartridges used tiny watch batteries to save their games, and those batteries were never meant to last this long.

But it’s not a huge deal, because I saved my family’s save file months ago. I backed it up with an Epilogue Operator, a device that connects your real Nintendo cartridges to Windows PCs, MacBooks, even a Steam Machine. 

The $50 GB Operator is a transparent trapezoidal cartridge reader that reads and writes to real Game Boy, Game Boy Color and Game Boy Advance cartridges over a USB-C cable; the company’s new $60 SN Operator does the same for Super Nintendo and Japanese Super Famicom titles. Epilogue’s also got a N64 version coming.

They’re not just for backups! As you can see in my embedded video, these gadgets can play your carts, too. My Xbox Ally now takes Super Nintendo cartridges, my Steam Deck plays Game Boy carts, and my phone is a Game Boy Camera, because I attached an Epilogue Operator to each one. (Double-sided magnet rings and Mechanism mounts helped.) 

Technically, the SN Operator and GB Operator are just cartridge readers, there are no original Nintendo chips or FPGAs inside. When you play a game, your computer saves the ROM to its drive and plays it in an emulator.

But Epilogue’s Playback software makes that easy —  just plug in the reader, plug in the cart, hit play — and it casts a convincing illusion that you’re playing your games on actual carts. Plug one in, and your game pops up on screen. Whenever you save in the game, Epilogue should automatically save back to your cartridge as well. Whenever you yank a cart, the game instantly winks out. 

That last one’s a bit of a double-edged sword, because Epilogue’s Operators can be touchy about cartridge detection. They aren’t as forgiving as an actual Super Nintendo or Game Boy if your carts are slightly dirty or damaged. I’ve got Japanese copies of Bahamut Lagoon and Sailor Moon R that’ll quit mid-session if I nudge them wrong. 

But I love that I can write savegames to my cartridges: I can play a game on my PC, move the cartridge over to my Steam Deck, then pop it into an actual original Game Boy or Super Nintendo and pick up where I left off. If you don’t need to automatically write saves, you can even switch to an emulator core that offers RetroAchievements, plus optional creature comforts like saving anywhere in a game with save states, or the ability to rewind time and try that tricky jump once more. And, you can backup your ROMs for use in a traditional emulator if you prefer.

Want to hear something especially nerdy? My family recently played Super Metroid on the Super Famicom Classic Edition that Nintendo released in 2017, and there’s actually a way to pull your savegames off of those emulator consoles, too. I converted and wrote that savegame to my actual Japanese Super Metroid cartridge, and now I’m playing that save on 30-year-old original Nintendo hardware. (Speaking of: did you know a Japanese copy of Super Metroid is downright affordable and entirely in English?)

I never had a Super Nintendo growing up, and I don’t have a working Super Scope light gun or SNES Mouse now. But the SN Operator is also helping me play amazing games like Metal Combat: Falcon’s Revenge and Mario Paint and share them with my kids, no crusty old peripherals necessary. There’s an option to let your PC mouse cursor emulate a light gun or mouse, and it was simple to map those to my Steam Controller’s gyroscope aiming. 

While you can do most if not all of these things by pairing emulators with other hardware, including the Open Source Cartridge Reader and a variety of Game Boy readers, Epilogue’s hardware is surprisingly high-quality considering its low cost, with attractive minimalist circuit boards clearly visible through classy polycarbonate. You know how we love our transparent gadgets around here. 

And Epilogue isn’t just riding on the coattails of emulator developers, either. While the hardware’s proprietary, the Romanian company contributes some money and some code to open source emulators. Co-founder Petru Sebastian Toader also tells me he built his software’s excellent new CRT filter himself, the company’s working on online multiplayer, and every single SN Operator has its clockspeed factory-calibrated using satellites so they can dump certain SNES games without errors.  

Epilogue’s also teamed up with the developers of Delta Emulator so you can plug real carts into your phone, and even built its own Flashback app if you want to turn your phone into a Game Boy Camera.  

With all these amazing features, I just think Epilogue should be a little bit more upfront on its product pages about the compromises you need to make.

You should know the limitations: the Operators don’t work with multi-game flashcarts, they don’t work with the Super Game Boy, and their heavily marketed “Counterfeit cartridge detection” has been very hit or miss in my tests, to the point I wouldn’t really trust it. The company advertises that it can turn your Game Boy Camera into a webcam, but it’s more of an OBS screen recording hack than I would have guessed.

And if you, like me, have original Game Boy cartridges with original batteries, it may be too late to save some of your saves by now. It’s far too late to pull Pokémon off second-gen games like Pokémon Gold, Silver and even Crystal, whose real-time clocks drained their backup batteries and might not have lasted more than a decade. 

As for first-gen Game Boy games, I pulled a Charmander off an original Pokémon Red cart last year, but even those carts are reaching their battery’s end of life. I have a 30-year-old copy of The Legend of Zelda: Link’s Awakening whose battery was working last month. Last week, I found its save, too, was gone. Thankfully, the GB Operator backed up my progress a few months ago.

Now I just need to install a new battery, flash my save, and I’ll be back in business.

Nvidia knows that DLSS 5 left a bad first impression. In March, many gamers took one look at Resident Evil Requiem protagonist Grace Ashcroft’s DLSS 5-ified face and declared it AI slop. So this week, the company’s re-introducing its controversial “neural rendering” with firm messaging about how it “preserves artistic intent,” “respects the rendered frame,” and gives game developers full control.

Now, ahead of its September 3rd launch, with an official Nvidia briefing under our belt and some hands-on time with the version that leaked onto the web, we think we finally get the picture. Here’s what’s going on.

Is DLSS 5 AI slop?

The critics weren’t wrong: DLSS 5 largely is a generative AI filter for video games. Kind of like a camera filter on your phone, DLSS 5 analyzes and modifies a real-time stream of 2D images after your 3D game engine is done spitting them out. It can indeed make game characters look terrible if applied poorly, like they’ve aged 20 years in a split-second. It uses a lot of GPU power, and can easily cut framerate in half — even applied well, you could argue that existing games look better without it.

But DLSS 5 isn’t what we typically think of when we think of generative AI. It doesn’t change shapes, or hallucinate words on a sign, or place brand-new fake objects in view. Instead, it imagines light, shadow, color, and detail that make the same shapes appear differently.

DLSS 5 can make dull, pasty characters look far more photorealistic. It can fill their hair and clothing and nearby foliage with signs of realistic light. It can ground characters and objects and clothing so they don’t look like they’re floating above video game surfaces.

Here’s an example I captured myself in Yakuza Kiwami: with DLSS on, see how contact shadows underneath each character make them seem like they’re actually standing on the ground?

(Later in this story, you should really check out how Cade Cunningham’s jersey casts shadows on his video game body, so it looks more like he’s wearing it.)

And unlike AI-generated photos and videos, which usually produce different results with the same prompt, Nvidia claims DLSS 5 is deterministic so games should look the same every time.

DLSS 5 is also not one-size-fits-all. Despite what it sounds like, “DLSS 5 On” and “DLSS 5 Off” isn’t just an on-off switch. Developers (and apparently modders) can drag various sliders to change how DLSS 5 plays with color and light, and they have three different AI models to work with at launch, giving your graphics “default,” “natural” or “cinematic” looks.

Plus, developers can decide whether to apply DLSS 5 to everything in a scene, just a character’s face and skin, just foliage, just silverware or furniture or even just food.

You’re saying game studios could leave character faces alone?

Yep. The toolset can automatically cut out a mask, like in Photoshop, to isolate a character’s face and skin from everything else:

If they don’t like the automated mask, devs can cut one out themselves. That’s the little “Developer Masking” checkbox in the image above. “There’s no limit to how many masks, how many groups, they could implement,” says Nvidia’s director of real-time content technology Gabriele Leone.

Would game studios actually leave the faces alone?

Probably not! The first game brave enough to ship with the tech is NBA 2K27, and I suspect that’s because Cade Cunningham actually looks a bit better with detailed skin, a uniform that doesn’t seem glued to his body, and a bit more glint in his eye. Take a look at Nvidia’s images for yourself:

That basketball, too — doesn’t it feel a bit more lifelike now most of its detail is in shadow?

But what about DLSStelle and all the other horrific AI faces I’ve seen since modders got hold of the tech?

Truly deranged. But these games are supposed to look like animation, and surely their creators would never do anything like this to them, right? Right??

“The soul and goal of GeForce is to achieve photorealism in real-time rendering,” Nvidia’s director of applied deep learning Edward Liu told journalists. And games with more realistic art styles don’t look anywhere near this bad. But even realistic games can look wrong if the sliders are set too aggressively.

What do you mean by “too aggressively?”

I’m willing to bet that if you’ve seen the leaked DLSS 5 modifying an existing game, the unofficial DLSS 5 mod was set to its default level of around 80 percent intensity or even higher. Some creators will even pump it up to 200 percent to make the difference between “DLSS on” and “DLSS off” especially clear.

But I believe even 80 percent is far too aggressive for many existing games. It can feel like you’re artificially aging a character by weathering their skin and deepening the wrinkles and creases in their face.

To show you, I fired up Shadow of the Tomb Raider. Here’s what Lara Croft looks like with DLSS 5 at 0 percent, 80 percent, and 200 percent intensity:

She seems… weathered, no? But at 40 percent intensity, she still looks like the same video game character to me, only with a bit more life in her eyes:

Similarly, here’s Kazuma Kiryu from Yakuza Kiwami once more. In this part of the game, he’s supposed to be 27 years old, and nowhere near the grizzled veteran he is by Yakuza 6. But as I ramp up the intensity from 0 to 40 to 85 percent, you can see his furrowed brow gets more furrowed, his hair lightens, and the increased shadow across his cheeks makes his face appear slightly gaunt.

He looks older — even if the underlying geometry of his face hasn’t actually changed. In a cutscene at 85 percent, it even looks like his hair might be tinged with grey:

But again, it’s something developers and modders can control with some granularity. I’d probably enjoy a version of DLSS 5 for older games that ignores the faces but enhances a character’s gear. Kiryu’s antagonist Majima, below, doesn’t quite look right to me — but his eyepatch and necklace look excellent.

By the way, these are just me playing with the overarching DLSS 5 slider when set to its “default” model. There are actually three different AI models, each with a slightly different look, and two different sliders, one for “tone” and one for “structure.”

What’s the difference?

I’m telling myself that “tone” is about light and color, while “structure” is about shadows and definition. But it’s probably easier to show you by isolating each:

Officially, Nvidia says that “structure intensity” controls a mix of ambient occlusion, contact shadows, reflections and subsurface scattering (read: how light scatters through materials like hair, skin, and leaves), while “tone intensity” controls color and “broader lighting” details.

What about the three different AI models?

Nvidia only called them Model A, Model B and Model C in my briefing, but the DLSS 5 mod I used describes them as “default,” “cinematic” and “natural.” Here’s the image Nvidia provided to show the difference:

And here’s my own comparison in Shadow of the Tomb Raider — you may need to blow these up and cycle through to spot the differences.

Is this really an AI filter? It doesn’t use the game engine at all?

Nvidia says that DLSS 5 can take some cues from game engines. But it’s based on the colors and motion vectors that it sees in the flat 2D frames, after they’re rendered and after other techniques like DLSS Super Resolution are applied.

When we asked how DLSS 5 knows the difference between a character model, an object, and a piece of foliage, Liu replied that the model was trained to recognize what it sees. “Just like a person, if you look at anything, you know where things are.”

You can even technically run DLSS 5 on video capture from a Nintendo Switch 2, and modders have already found ways to inject it into earlier games that used OpenGL and DirectX 9.

Will game developers actually buy into DLSS 5? Will they use it well?

That’s one of the big remaining questions — and one that Nvidia left unanswered during and after our briefing.

While Nvidia said in March that DLSS 5 would come to at least 16 different games including Assassin’s Creed Shadows, Hogwarts Legacy, Starfield, and Resident Evil Requiem, the company did not answer questions about when DLSS 5 might come to any game beyond NBA 2K27, nor did it confirm that other titles are still coming following the original backlash. Perhaps Nvidia and its partners will wait and see how reviewers and gamers react first.

It’s worth noting that this may be a niche technology to start: while it no longer requires two RTX 5090 GPUs to function, it still uses a lot of horsepower. Originally, Nvidia announced it would only come to the latest RTX 50 series GPUs, which have become quite expensive: even a lower-mid-range RTX 5060 retailers for around $470 these days.

But after modders brought DLSS 5 to earlier GPUs, Nvidia has decided to bring it to the RTX 40 series too. “Our current focus is on optimizing performance for the GeForce RTX 50 Series with model updates expected later this fall. Once RTX 50 Series performance is more fully tuned, we plan to work on expanding official support to the GeForce RTX 40 Series,” Nvidia tells The Verge.

Just how much of a performance hit are we talking?

Nvidia claims that an RTX 5060 is enough to play NBA 2K27 at 1080p with ultra settings, ray tracing, and DLSS 5 enabled — assuming you’re pairing it with a fast gaming CPU and a whole heap of RAM.

With NBA 2K27, Nvidia claims the performance hit is 50 to 60 percent on average across all Nvidia RTX 50-series GPUs. When I ran the Shadow of the Tomb Raider benchmark at 1440p with on an RTX 5090 at Highest spec, DLSS Quality and an unofficial DLSS 5 mod, I saw framerate exactly cut in half — 312fps with DLSS 5 off, 156fps with it on.

But Nvidia says it’s not done optimizing performance; the cost should go down.

Will gamers be able to tweak DLSS 5 the way they like?

The DLSS 5 mods are so easy to apply, it seems like the cat is out of the bag. But Nvidia’s message is that the developer has full control, not the gamer.

“For gamers, DLSS 5 is a simple on-or-off toggle,” Nvidia tells The Verge. But developers could pass along certain controls to gamers, the company says.

Which mod should I install?

I don’t yet feel comfortable recommending one — there’s been a feeding frenzy for DLSS 5 mods ever since the leak and we haven’t been able to properly vet. I installed one on a completely fresh system, just in case something went wrong.

Where can I learn more?

Nvidia now has its video explainer and blog post up on the web:

How have gamers reacted to Nvidia’s re-introduction of DLSS 5 so far?

The reactions have been all over the place since DLSS 5 leaked and modders began posting their work. I’ve seen people say that DLSS 5 looks like a free next-gen remaster of any game you like, and I’ve seen people say it looks as terrible as it did before. I’ve seen people suggest that even if it looks good, it’ll make every game look too similar once developers start using it as a crutch.

But many of those impressions were based on images and videos of the tech at its extremes. Once people realize there’s granular control, perhaps reactions will become more measured.

Update, September 3rd: Added more answers from Nvidia.

Correction: An earlier version of this article misidentified characters in screenshots; they are NBA player Cade Cunningham and Majima from Yakuza.