When the Steam Deck came out late last year, it was great in spite of its relatively older hardware, most of which was a generation or two out of date. On paper, you can improve the handheld gaming PC market by quite a bit just by updating the hardware, but that hasn't panned out. Many newer handhelds use the latest hardware, and we're still not making generational progress.

To get to the next generation of handheld PC gaming, companies need to make a lot of crucial improvements, whether we're talking about devices from Asus or Ayaneo or even Valve. It's not just about making everything bigger and better, but being very deliberate when it comes to upgrading today's handhelds in order to transform them into ones for the next generation.

Custom APUs are going to be necessary to get the right balance

One of the main problems with the ROG Ally and every other current generation handheld has to do with their processors. Virtually all handhelds are using AMD's Z1 Extreme, which, despite the name, is literally just a top-end Ryzen 7040 series APU with the full eight cores and 12 CUs (graphics cores). It's also clocked pretty highly (though it's missing Ryzen AI for whatever reason).

Now, you might think it's impossible to lose with this incredible APU, but I would argue this thing is part of the problem and not the solution. Firstly, the CPU is just way too much in every respect. It uses eight big Zen 4 cores clocked crazy high for maximum performance — and power draw. Yet the Deck, with its measly 4-core Zen 2 chip, is always within 80% to 90% of the Ally's framerate at the same power level, mostly because you don't actually need eight cores for gaming. You can hit about 50% more frames per second if you increase the power draw, but that means less battery life. Not exactly the next generation of handheld gaming, is it?

Even if the Z1 Extreme was significantly faster than the Deck's APU, it wouldn't matter since we're not gaming at 120+ FPS on handhelds, which is where a CPU advantage would actually matter. Besides, the GPUs inside these handhelds aren't nearly powerful enough to push super high framerates without reducing the graphics settings so hard that it looks like you're playing on a Wii.

The end result is that the Z1 Extreme is a little faster than the Deck APU but costs way more to manufacture and has features you won't even use. By contrast, the Deck APU was purpose-built and not simply sourced from a product line that's made for laptops and desktops. If companies want to develop a handheld gaming PC that can be as cheap as the Deck, run as efficiently as the Deck, and actually beat the Deck in raw performance, a custom design isn't optional. It's necessary.

What a great, next-generation APU ought to be

While I'm sure many people are happy to pay $700 or even upwards of $1,000 for a handheld that's only 20% faster than the Steam Deck or 50% faster with a much higher power draw, most people aren't going to be interested. If some company asked AMD to design a custom chip for a next-generation handheld gaming PC, I see a few things it would need.

A big question is whether an iteration on the Deck APU or the Z1 Extreme would be ideal. The Deck APU has the benefit of being on the cheaper 7nm/6nm node, and AMD could update it with up to Zen 3 and even RDNA 3, which made it to 7nm with the RX 7600. The Z1 Extreme, on the other hand, benefits from the more power-efficient and denser 5nm/4nm process and also has access to Zen 4, which isn't much better than Zen 3 or even Zen 2 when it comes to gaming. But Zen 4c is much smaller than Zen 4 while being identical except for a lower frequency. Starting from the Z1 Extreme makes the most sense, especially because 5nm will eventually get cheaper, while 7nm is already as cheap as it can be.

In order to mitigate the fact that 5nm is already expensive to make, I suggest two big changes. Firstly, let's cut off four of the CPU cores as well as half the L3 cache, leaving just four cores and 8MB of L3. Let's also swap out Zen 4 with Zen 4c. By just altering the CPU cores, the size of the Z1 Extreme would shrink down to about 85% to 90%. There's probably even more that can be removed without issue, like PCIe connectors for graphics cards.

A smaller APU is cheaper to manufacture, and cutting out CPU cores will help to reduce power consumption, all without impacting gaming performance. This APU would be more expensive than the Deck APU at first, but it would be far more affordable than the Z1 Extreme and just as fast. For devices that want more speed, why not just put more GPU cores where those four useless CPU cores used to be? A 16 CU APU would provide a decent boost in performance over the existing Z1 Extreme while being about the same size, and that's a huge increase in value.

Other hardware can also be better

You can probably tell I think the processor is the most important piece of the puzzle when it comes to handheld gaming PCs. Still, there are other important things Valve, Asus, and other companies should consider, like the display, the controls, and the software.

At the top, display technology could be improved across the board in the next generation. I'm not just talking about reducing bezels or getting OLED technology, but 900p screens would make a lot of sense. It's a great middle ground between 720p and 1080p and works for both lower-end APUs as well as higher-end ones. I think the Deck would look significantly better with a 900p display without reducing performance very much (for reference, it has a native 1280x800 display). Also, variable refresh rate (VRR) should be the standard and would be especially good for lower-end devices.

One thing that the Deck proves is that controls are super important, and it's one area where the ROG Ally and other Ryzen 7000-powered handhelds aren't up to scratch. Honestly, upcoming handhelds should just copy the Deck. While the haptic trackpads and gyros aren't super necessary, it would be best if they became standard. At minimum, I would expect higher-end handhelds to match the Deck when it comes to controls, especially for simple features like back buttons, of which the Deck has four, the Ally only has two, and some consoles have none at all.

Maybe everybody should switch to Linux?

The most important place for improvement, though, is software. While Linux isn't the best overall OS for gaming, it works well enough, so right now, Linux is just better for handhelds than Windows. I'm particularly disappointed in how poorly Windows works with controllers, which forced Asus to build its own SteamOS-like UI on top of Windows to make up for it. But if you ever need to access the underlying OS, you need a keyboard, whereas on the Deck, you can do a lot without even having to go into desktop mode.

Here's a free idea for Microsoft: Why not just port the Xbox UI to Windows for handhelds and console-like PCs? The Xbox runs on what is basically Windows, and the Xbox even uses AMD hardware just like the Deck and Ally. At least then, the ROG Ally wouldn't be hamstrung by an OS that has the best back-end for gaming but the worst front-end if you use a controller.

All of these uber-expensive handhelds aren't pushing the medium forward

Overall, my main problem with the way the handheld gaming PC space has evolved so far is that the Steam Deck is still the only "affordable" option. I think the ROG Ally is mostly fine since it does offer good hardware, at least, but when I look at the Ayaneo 2S and OneXFly, I'm just uninterested. I have a hard time taking these alternative handhelds seriously when they do stuff like offer 64GB of RAM. Who in the world needs that much RAM to run any game in 2023, let alone on a handheld?

The Steam Deck didn't invent handheld PC gaming, but it did revolutionize it by proving you can make all custom hardware and offer it for a low price. We've had expensive handhelds with ridiculous specs for a while now, but they're clearly not moving this nascent segment of PC gaming forward. For the next generation, we need other companies to think like Valve and actually cater to what gamers as a whole want, and not just gamers with deep pockets that solely care about performance.