Steam Deck may well be the most mainstream example of the handheld gaming PC, but Valve is certainly not the first company to give the concept a shot. Other manufacturers got there first and continue to innovate with intriguing products. The AyaNeo Air is the most exciting of those handhelds yet, in that it’s something of an anti-Deck. The Air is small and easy to handle in the way that Steam Deck is not. It has the OLED panel with the inky blacks and incredible contrast that the Deck’s mediocre IPS LCD does not. Also – for good or bad – it runs on Windows 11, meaning that it plays everything, something that remains beyond the Deck when running the default Steam OS.
First impressions are compelling, exactly because of its miniature form factor, like a kind of hybrid between Switch, Switch Lite and PlayStation Vita. But it’s at this point we should point out that there are actually two versions of the AyaNeo air – a standard model with a 28Wh battery and a thicker, heavier Pro model with a 38Wh variant. Battery life, as we shall discover, is a real problem on the standard machine.
What both have in common are the same basic aesthetics, the same excellent analogue sticks and face buttons, the same lovely display and the same SD card reader. I found the shoulder triggers a little ‘spongy’ for my tastes while the shoulder buttons feel a little insubstantial but there are some genuine nice ideas here, such as USB-Cs top and bottom, both capable of charging.
In terms of specs, the AyaNeo Air has a bewildering array of options, so let’s boil it down to the bare minimum. There’s an ‘entry level’ version with 8GB of DDR4X-3200 memory and a 128GB SSD, which I think you should discount straight away. AMD APUs, like the Ryzen 5 5560U common to all variants of the standard Air thrive on memory bandwidth, so I’d recommend the more expensive 16GB DDR4X-4200. The processor itself is effectively a cut-down version of the Ryzen 7 5825U, available at the upper echelons of the AyaNeo Air Pro spec list. The full chip has eight CPU cores and 16 threads, reduced to six cores and 12 threads on the Ryzen 5. Eight Vega GPU compute units on the top-end Ryzen 7 drop down to six on the Ryzen 5. The only other spec variant to be aware of between Air and Air Pro is the battery – get the Pro with the larger battery. It’s bulkier, but not to any degree that actively interferes with gaming.
Setting up the unit with my test games, I’m reminded of the issues I encountered running Windows on Steam Deck. The OS isn’t really set-up for handheld gaming in the way that SteamOS is, meaning there’s far less hand-holding in the general UI. A USB-C hub is essentially essential in getting quickly set-up, using a keyboard and mouse. However, there is a bespoke AyaSpace app which acts as an uber-launcher of sorts – collating all of your installed games, while also allowing users to tweak controller settings and functionality… and to adjust the amount of power consumed by the Ryzen processor.
That latter point is crucial in managing battery life. Out of the box, three power levels are supported: 5W for battery conservation, an 8W ‘balanced’ mode and a 12W gaming mode, meant for modern titles. There’s also a user-configurable Pro mode, where you choose the power limit up to 15W on the standard Air and 18W on the Air Pro. The higher the power limit, the less battery life you’ll have, but also the harder the cooling assembly needs to work. And the machine can get rather warm, if not hot to the touch at higher wattages. In my tests, 5W is a write-off for gaming, 8W is fine for very low-end titles or 2D games (Cuphead works beautifully, as an example) while 12W is the place to be for the best ratio between power and performance. That said, 15W often provides the extra juice required to turn sub-30fps gameplay into a more palatable experience. This, again, points out the utility of the Pro model with the larger battery, even if you opt for one without the enhanced Ryzen 7 processor – the fact is that the bigger battery makes accessing those higher power limits more comfortable in terms of battery life.
|AyaNeo Air||AyaNeo Air Pro||Steam Deck|
|Processor||Ryzen 5 5560U||Ryzen 5 5560U/Ryzen 7 5825U||Custom AMD ‘Van Gogh’|
|CPU||Zen3+: Six Cores/12 Threads||Zen3+: Six Cores/12 Threads or Eight Cores/16 Threads||Zen 2: Eight Cores/16 Threads|
|GPU||Vega 6: Six Compute Units at up to 1600MHz||Vega 6: Six Compute Units at up to 1600MHz/ Vega 8: Eight CUs at up to 1600MHz||RDNA 2: Eight Compute Units at up to 1600MHz|
|Memory||8GB LPDDR4X-3200 or 16GB LPDDR4X-4200||8GB or 16GB LPDDR4X-4200||16GB LPDDR5-5500|
|Screen||5.5-inch 1080p OLED||5.5-inch 1080p OLED||7-Inch IPS LCD|
|Storage||128GB/512GB SSD||512GB/1TB/2TB SSD||64GB Flash/256GB/512GB SSD|
So let’s dig deeper into performance. We shouldn’t expect Steam Deck-like performance from a much older GPU architecture, but the existence of the Deck can’t be factored out of any performance comparisons: I’ve included benchmarks here of the same titles running under Windows, albeit Win10 – Steam Deck still lacks the BIOS support required for the latest rendition of the OS. Using both the standard and Pro models (many thanks to our friends at Time Extension for lending us their Pro!) we have measurements at 8W, 12W and 15W to get an idea of the scaling on both available processors, while the Deck metrics at 15W reveal how much of a game-changer RDNA 2 is in the mobile graphics space.
Kicking off with Forza Horizon 5, this was actually the triple-A game that gave me the best experience on either of the AyaNeo Air units. Using the internal 30fps cap, I could run at 12W for a virtually locked performance level across most of the game content, barring heavy storm weather. This was using 720p resolution, medium settings and TAA rather than MSAA, which caused speckling artefacts. The benchmark sequence is taken from a taxing area of gameplay and is representative of the actual game experience, unlike many other benchmarks.
There are a couple of takeaways here. Firstly, scaling from Ryzen 5 5560U to Ryzen 7 5825U is not exactly compelling – seven percent at 12W and just three percent at 15W. Power is clearly more important than silicon, until you factor in the relatively stellar performance of RDNA 2. Also curious is that at 8W, the Ryzen 7 seems to be power-starved and performs worse than the Ryzen 5 – which I double-checked and found to be the case across all tested games.
Forza Horizon 5: 720p, DX12, Medium, TAA
The Marvel’s Spider-Man cutscene-into-gameplay sequence taken from the beginning of the game isn’t representative of the actual way the game plays but rather puts graphics hardware to the test in an entirely like-for-like scenario played out across all test units. Here, we find the same kind of scaling: Ryzen 5 outperforms Ryzen 7 at 8W (which, to be fair, is hardly useable for triple-A titles anyway), while once again, the Ryzen 7 offers greater performance boosts vs its cut-down equivalent at 12W rather than 15W.
Steam Deck’s boost over the AyaNeo Air models is at its lowest here, but we’re still looking at a big improvement overall. In terms of settings, I choose to play Marvel’s Spider-Man at medium settings, with traffic, crowd density and depth of field set to low – so that’s how I benched it. Just like Forza Horizon 5, the game comes with a 30fps cap option that delivers properly consistent frame-pacing but the difference is that dynamic resolution scaling is essential in delivering anything close to a locked experience (I’d recommend avoiding IGTI and FSR 2.0 options as they both add a lot of blur). Even then, there are still problems with stuttering in fast web-swinging across the city.
You can, of course, drop settings even lower but even with the beautiful OLED screen, I do think there’s a point you reach where you strip away too much. This title also highlights the hard performance limits of the Vega architecture: it’s a struggle to play this locked to 30fps, while Steam Deck can do it all day, with only minimal stutter in the web-swinging sections where AyaNeo Air obviously has issues.
Marvel’s Spider-Man Remastered: 720p, DX12, Medium/Low, TAA
Death Stranding continues the trend of the Ryzen 5 outperforming the Ryzen 7 at 8W, but the point is moot really because the game runs like a slideshow. What is intriguing here is that there is fairly decent scaling, not just between wattages but also when comparing Ryzen 7 to Ryzen 5. Notionally there is an extra 33 percent of GPU compute available with the extra compute units, but due to the brutal power limit and architectural reasons, you never see the full potential performance. Still, an eight to 10 percent performance uplift here depending on power puts Death Stranding in front.
However, what I discovered in gameplay is that there’s no real point reducing settings below the default and it was tough for any AyaNeo Air to maintain 30fps locked, something that Steam Deck has no real issue with. In this scenario, effectively the only way forward is to use upscaling or lower internal resolution.
There is another issue only hinted at in the bar charts here: yes, the Steam Deck is a lot, lot faster but the lowest one percent scores hint at the real issue facing the Vega-powered AyaNeo Air – stuttering. Mid-way through the bench, Vega graphics faces off-putting stuttering issues that the Steam Deck does not. I’m wondering if this is down to how graphics memory is portioned off by Windows but it is what it is – and was repeatable across multiple runs, so it’s not an issue caused by shader compilation (which Death Stranding does while loading, by the way).
Death Stranding: 720p, DX12, Default Settings, TAA
We’ll close out with metrics on Remedy’s Control, which does not have an internal benchmark sequence. This causes some problems for us in that the only like-for-like content that’s easily accessible is the beginning of the game, which is considerably lighter on the graphics hardware than the most hardcore of action within the experience. Still, what you do get are relative performance differentials between all tested scenarios.
At around an eight to 11 percent boost comparing Ryzen 5 to Ryzen 7, there is meaningful scalability from the higher-end mobile chip in the 12-15W range, while once again the lower-end offering delivers higher frame-rates in the extremely power-constrained 8W range. I’d definitely be looking at these results using the ‘percentage differential’ mode rather than the frame-rate mode though, because as stated, frame-rates in effects-heavy combat are lower.
Within the video at the top of the page, you’ll see a fair amount of Control played at 720p on low settings using the standard AyaNeo Air with the Ryzen 5 5560U and the 12W power limit. This one is very challenging for the hardware on a number of fronts in terms of actual gameplay: firstly, there is no 30fps cap, meaning a very inconsistent experience from start to finish. Secondly, without the 30fps cap, a lot of the stutter comes from 16ms frame-times you don’t really want – and you’re paying for all of those extra generated frames with battery life. And finally, performance can dip into the mid-20s. Ryzen 7 at 15W will be a better experience for sure, but still subject to the drawbacks of unlocked performance.
Control: 720p, DX12, Low, TAA
The benchmarks reveal that the Steam Deck’s RDNA 2 graphics core is capable of far better performance at the same power level, but it is important to note that Valve’s decision to go for a quad-core CPU cluster does result in diminished performance in some scenarios, where the AyaNeo operates with an improved CPU architecture and more cores. It won’t affect every game, where the GPU ceiling is prevalent but stats offered up by Forza and Tomb Raider benchmarks show some interesting results. Both benchmarks offer a readout at the end that highlight the performance level of the CPU in processing game logic and render logic, the game simulation and the time taken to prepare commands for the GPU.
Forza Horizon 5 shows fairly consistent averages in processing game logic, but with more CPU cores available, lowest scores improve significantly – meaning better results for the AyaNeo Air products. Meanwhile, on the rendering side there’s a clear pecking order in performance moving from Deck at the lowest to Air in the middle to Pro at the top, though lowest frame-rates are similar between Air and Deck.
Shadow of the Tomb Raider is more illuminating – in terms of game logic, average frame-rates on the Air and Pro are significantly faster than the Deck, and lowest frame-rate is correspondingly lower on Deck too. Looking at the game logic side of things, more of a gap opens up between the three CPU clusters though it’s clear that Steam Deck is considerably slower. As I said, most of these results don’t impact much of the actual game because the GPU is the primary limit but certainly in the case of Shadow of the Tomb Raider, we have good data here on why we’re getting stutters on Steam Deck at certain places in the benchmark that we aren’t getting on the Ayaneo handhelds.
While not especially relevant to the games I played, I can absolutely see scenarios where the AyaNeo Air may outperform Steam Deck. For example, looking at game emulators, typically you’ll find that CPU performance is, on balance, more important than graphics. I’d recommend taking a look at MVG’s video on AyaNeo emulation for some insight on this.
|Forza Horizon 5 CPU Bench||Game Logic Min FPS||Game Logic Avg FPS||Render Time Min FPS||Render Time Avg FPS|
|Steam Deck (4C8T)||70.9||92.7||59.8||69.7|
|AyaNeo Air (6C12T)||76.9||91.3||59.8||75.3|
|AyaNeo Air Pro (8C16T)||81.7||92.2||62.4||79.0|
|Shadow of the Tomb Raider CPU Bench||Game Logic Min FPS||Game Logic Avg FPS||Render Time Min FPS||Render Time Avg FPS|
|Steam Deck (4C8T)||33.0||49.0||54.0||75.0|
|AyaNeo Air (6C12T)||43.0||60.0||58.0||90.0|
|AyaNeo Air Pro (8C16T)||43.0||68.0||64.0||99.0|