Review: x86 Handheld

It’s been a long time since this happened, but I kept going back and forth on whether to write it down. Same as my experience two years ago when I got dragged into a quarantine facility for over ten days — I really wanted to write about it, and it would be meaningful, but it’s easy to get burned, to invite unnecessary trouble.

Last year I really wanted to write it out and vent, but as time passed I forgot about it. Today I happened to remember, and figured I should write it after all — the whole process is not just surreal but ironic, and it also illustrates the state of most of the industry from the side, and it even calls back to some of my experiences from years ago.

To avoid unnecessary trouble, I won’t mention the product’s name in this whole piece, and I won’t comment too heavily on the product’s R&D and management strategy. I’ll try to just lay out the facts plainly and let everyone judge for themselves. At the end I’ll give my short take.

Of course, this is just a long dream I had last year. All the plot comes from the story in the dream, not reality. Please don’t match it to anyone real, and don’t make weird guesses.

1. A revolutionary product and prototype test recruitment

Last year the Win handheld market was brutally competitive. Various no-name factories were fighting tooth and nail, big brands joined in too, and more and more factories said they’d enter. Prices kept going lower. Then one factory said it was going to release a product too, and put out concept art. That part doesn’t matter much — the product was cheap, and it still drew a lot of people. Later this factory announced it was going to cooperate with the Linux community and release an “open-source, open Linux X86 handheld,” and even held a launch event, bringing in people from the Linux community to speak. It made them look very determined. Officially they also said this would be a “Linux First” product, focused on the Linux space.

Then on their social media account I saw a notice for “prototype test recruitment.” The way to sign up was to join a Discord group, fill out a form, and state your reasons for applying.

I honestly went in with a “free toy” mindset — it’s a prototype test, you can return the machine after testing, you don’t have to buy it. And it’s a Linux handheld, maybe it could rival SteamOS? So I wrote a detailed application explaining I was a Linux user, very interested in the handheld, and could provide testing plans and suggestions including logs.

A few days later the prototype tester list came out, and my name was on it. Everyone joined a Discord group for the test, and so the test process began.

2. Delays and waiting

After joining the group, the official side soon posted a manual explaining the test process, blah blah blah. Roughly: every week they’d give a spreadsheet, and after receiving the product you’d test according to the spreadsheet, submit a detailed test process and report every week, and at the end they’d score you based on results and attitude. High scorers would get special discounts and gifts, etc.

At this point I was still confident, because I know exactly what kind of people populate China’s so-called “Win handheld circle” — a guy who plagiarized a foreign community script to make a pirated system image gets packaged as an expert. I figured I could handle the Linux and handheld side.

But soon I noticed the vibe in the group was off. Maybe not 100%, but everyone except me was asking whether the machine could run Windows, preferably shipped with Windows preinstalled for testing. Nobody used Linux, nobody knew how.

“Wait, didn’t the recruitment notice clearly say this is a Linux First device? Do you people not read?”

Then came the traditional “hurry up and ship” phase. I thought, this is a prototype test, not a pre-order — what’s the point of rushing shipping?

Then the official side went silent for a long time. Originally they said testing would start two weeks after the list was announced, but over a month passed and it still hadn’t started. The group complained every day. Finally the person in charge showed up (seemed to be a company manager) and said a contract needed to be signed before testing. They sent a PDF of the contract. Either e-sign it or sign and fax it back. The test machine required a deposit — there were 16GB and 32GB versions. After testing you could get the deposit back or keep the machine, buying the retail product at a better price, etc.

After signing, it was endless waiting again. The official side would go missing for long stretches, then suddenly appear. You could tell this manufacturer had problems with project management and planning, and lacked communication. But considering the brand was relatively well-known, nobody thought they’d run off — just impatient from waiting.

After endless urging from the group, the official side finally released some material photos and said shipping would happen soon. The promised test gift also turned into a so-called “limited edition badge.” Fine, better than nothing.

At first I thought I’d get the machine and install Black Myth: Wukong to play. By the time I’d already reached Chapter 2 on my Steam Deck, the test machine finally arrived.

The same day the test machines were released, the official side announced that because of the long delay and tight schedule, the whole “weekly spreadsheet” process was cancelled. Test period: one month. After one month, either get a refund or keep the machine at a discount.

The way they’d pick outstanding testers would be determined by the official side based on how active you were.

3. The zoo in the test group

Before talking about the machine, let me talk about some of the creatures in the group. Anyone who’s hung around these so-called “handheld groups” knows how high the idiot concentration is. Same for a prototype test group. Here are a few typical types — not referring to specific people, just abstracting a few archetypes:

Number one: Hardware Bro This is someone who never stops calling himself a “hardware engineer.” I’ll call him Hardware Bro. His defining trait is that every sentence needs a prefix: “As a hardware engineer…” or “I’m a hardware engineer…” As if terrified nobody knows his job. From observation, Hardware Bro might have participated in hardware testing and knows some technical stuff, but his understanding of hardware engineering itself is nowhere near deep — a lot of it is bluffing. Hardware Bro was constantly flooding the group, eager to teach, replying to every single conversation, always with the “As a hardware engineer…” prefix. He’d answer the same repeated questions over and over. Every time I saw him posting at full throttle, I couldn’t help thinking of a saying: the more you lack something, the more you keep emphasizing it.

I suspect he’s not a hardware bro at all — just repeating it so others believe he is.

Number two: Angry Bro A more typical case. Basically no technical skill, the type who starts howling at the slightest thing, spamming the group with “what do I do” “what do I do,” getting triggered by any doubt, dumping emotional rants in the group. You’ve probably all seen this type. What stuck with me is that before the machine even arrived for testing, he was fantasizing about playing all kinds of AAA games smoothly on it. I kindly reminded him, based on my experience, that test machines likely have color-shifted screens, build quality problems, poor performance tuning, system bugs, and it’s a Linux system — don’t get your hopes too high, focus on testing.

He got triggered, started ranting that I was someone’s shill, that a test machine couldn’t be that bad, that he’d definitely have a great time with it, no problem. Of course, after he got the machine he got triggered again, found everything I said was true, then turned his anger on the manufacturer, venting emotions. The manufacturer just said “we’ll fix it soon” and he was completely soothed, expecting them to solve the problem — even believing they could fix hardware issues with software… He’d also whine to the manufacturer like a scorned lover, saying he’d sold all his other machines just to wait for this test unit to enjoy, trying to guilt-trip them. Naive as a child.

I have to say, this guy is really easy to fool and easy to manipulate. His emotions kept swinging, he gave no useful suggestions, judged everything by his feelings and subjectivity. Said the machine overheated — couldn’t say why or how hot, didn’t measure. Said games felt laggy — couldn’t say what framerate, just “I feel like it’s laggy.”

Number three: Student Bro This one was relatively normal by group standards — but only this one. Probably a rich international student, didn’t lack money or machines. Not great with Linux either, but could tinker properly and give correct suggestions. Because he was online overseas, he also knew to go to Github and Discord to find Linux community developers and submit issues. But clearly, for him this test was just for fun.

4. The “test machine,” as expected

As I told Angry Bro earlier, I had basically no expectations for test machines from these domestic manufacturers. Most have a terrible screen, buggy systems, and lousy build quality.

Honestly, even retail machines from many domestic brands have trash QC, let alone test units.

What interested me was how the preinstalled Linux system would be — something like HoloISO’s SteamOS? If so, how would they handle TDP? Also, the machine’s exterior imitates the Steam Deck, so could the controls achieve the same effect as the Steam Deck?

After the machine arrived, the exterior was quite pretty. Not surprisingly, it came with a screen so blue it looked like Avatar, low brightness, surreal. Tested color temperature seemed to be over 10000, max brightness felt like no more than 200-something nits, while the official marketing clearly said 500 nits. When I asked the official engineer, they said “firmware optimization” would come later…

I wonder, can screen brightness really be optimized via firmware?

The preinstalled system was Linux, but not like HoloISO — based on another distro, using btrfs snapshot boot: every boot is a btrfs snapshot, a bit like Apple’s macOS. That part was actually interesting, at least ensuring the system is immutable and absolutely safe.

Linux system support was on Discord and overseas forums — clearly not in the Discord group. Communication between the testers and the system side was fragmented. In principle the official side collected feedback and told the overseas side. But the official “feedback lady” knew nothing about Linux or the system, could only act as a relay, and the bugs she reported were unprofessional: “feels laggy,” “some games not compatible with SteamOS,” etc. Some were machine-related, some not, but she couldn’t tell at all, just forwarded everything stiffly, and who knows if it reached the Linux community.

I proactively contacted people in the Linux community to let them know there was a Discord group test in China. The community developers were eager to collect relevant information, and I was willing to be a bridge. But the official side kept disappearing. I had no choice but to be a relay myself, but most people in the group said they couldn’t read English and didn’t want to. The first thing they did after getting the machine was complain about build quality, and the other thing was ask how to install Windows. They had zero interest in Linux, believing only Windows can game properly.

When the community developer asked me how many people in the Discord group used Linux, I could only say, probably no more than 3, because Linux is hard to use in China. There seemed to be one Arch user in the group, a Linux user, but he barely spoke, and didn’t seem very interested in gaming on this device.

The developer offered to provide a Chinese mirror, modify flathub’s Chinese mirror, even help Chinese users add default language support. But the people in the test group didn’t care at all — they just wanted to delete Linux and install Windows ASAP… Many already had.

When the community heard this, they could only say “Sad…” It’s Linux First, but nobody cares, and the device manufacturer had completely vanished.

5. The first serious problem

The reason most people tried to install Windows after getting the test machine wasn’t just that they disliked Linux — the machine itself had issues. The test unit’s Linux system showed obvious performance drops and lag when plugged in.

Angry Bro was the first to break after getting his machine. While saying he didn’t know Linux, he spammed high-intensity screenshots complaining the machine was trash, performance awful, so laggy, so unbearable, etc. I tried to get him to provide logs to see why it was laggy. He ignored it, saying “what’s the use of logs, this machine is garbage, it’s frozen to death.”

Hardware Bro in the group was the first to rebel, saying he couldn’t stand Linux’s lag and was going to install Windows. Just then the official side released Windows drivers. Hardware Bro immediately deleted Linux and installed Windows. Then Hardware Bro reported victory again: under Windows the lag was gone, and after installing a so-called “motion assistant” for TDP adjustment, 25W could run any game smoothly.

“See, Linux is trash.” Angry Bro immediately echoed.

“I’m a hardware engineer, but nobody at our place uses Linux. For gaming, it’s gotta be Windows.” Hardware Bro also gloated. “Although I’m a hardware engineer, I can confirm this is a Linux bug. I’ve already reported it to the official side, told them to fix it fast. Windows just has fewer bugs!”

For a while everyone in the group was asking Hardware Bro how to install Windows. Hardware Bro excitedly replied one by one: “I’m a hardware engineer, trust my judgment, Linux is just trash. Let me teach you how to flash the Windows firmware!”

I clearly didn’t believe Linux was causing the lag. I asked Hardware Bro and Angry Bro in the group if they could provide logs and help test. I got a hilarious reply.

“I’m a hardware engineer, this is a Linux problem, what logs? Installing Windows already solved it! I suggest the retail version just ship with Windows preinstalled! I’m a hardware engineer, I don’t use Linux, leave this problem to someone else.”

This is supposed to be a Linux First device, this is a prototype test group — if you don’t solve the problem, what kind of test is it? Just blame “a system bug” and dump it on the developers? And under Hardware Bro’s call, almost everyone in the group except me and Student Bro had switched away from the original Linux system.

With Hardware Bro and Angry Bro flooding the group at high intensity, the crowd quickly reached a consensus: Linux system problem caused the lag, installing Windows solves it.

Is this the right attitude and solution for a prototype test?

6. Investigation and fix

After I got the machine I also found the performance drop. I immediately communicated with the Linux community. The developer was baffled, said he couldn’t reproduce it on his device at all, didn’t know why the test machine would lag.

I tried to reproduce the lag and found that as long as the machine was plugged in, after a short while serious performance drops occurred. Opening the OSD to check CPU frequency, it was throttled to 800 MHz, and the GPU was also at minimum. Looking at the system’s built-in TDP plugin, it was set to 35W — clearly too high.

I started lowering TDP and found that below 30W the lag instantly disappeared, CPU and GPU frequencies went back to normal. But I also found that on battery, 35W TDP ran fine; plugged in, it throttled.

After nearly a whole night of repeated testing, I had a hypothesis:

The device has a power delivery problem. When plugged in, TDP above 30W causes both CPU and GPU to throttle. On battery, no problem. This is probably unrelated to the OS.

Hardware Bro’s crowd insisted Windows had no problem because Windows by default has no TDP control and likely runs at 25W TDP.

And the official Linux system had a TDP plugin that, unfortunately, defaulted to: auto-switch to 35W TDP when plugged in.

I tried to contact the Linux community to report. The community person told me his machine had no problem. I asked what config his machine was — turned out his was an 8840U machine, while our test units were 7840U. Maybe his power delivery design was different, which would explain the issue.

If my hypothesis was right, those who flashed Windows and plugged in with TDP above 30W would have the same lag. I asked in the group if anyone with Windows could help test, but nobody responded. Hardware Bro just kept repeating “I told you it’s a Linux problem, install Windows and enjoy~” I got annoyed and snapped at him, and only then did he weakly say “I don’t understand Linux, but I know installing Windows solves the problem.” Damn, is this the testing attitude of an engineer?

In the end Student Bro stepped up and tested with me, confirming the problem indeed came from power delivery.

So we recapped:

The device’s default Linux system preinstalled a plugin: 20W TDP on battery, auto-jumps to 35W when plugged in — like different settings for different scenarios.

Test machine users initially booted unplugged, so it felt fine. Once the battery ran low and they plugged in, the power delivery issue triggered immediately, causing both CPU and GPU to throttle and severe lag.

Those who claimed Windows had no problem — the reason was simply that Windows had no TDP setting and ran at default 25W TDP.

Having identified the problem, I described and recapped it in the group. Hardware Bro and Angry Bro said they didn’t understand. But Hardware Bro saw the words “power delivery problem” and started high-intensity repeating in the group that there was a power delivery problem, as if he’d discovered it.

Actually he quickly took credit for it, saying as a hardware engineer he’d known about the power delivery problem all along and had promptly reported it to the official side so they could fix it fast.

Later the official side had an engineer contact me. I repeatedly communicated and confirmed the reproduction method with the engineer. The engineer finally understood the cause and said an EC update would be released soon to fix it.

A few days later, Hardware Bro said that through his efforts as a hardware engineer reporting the bug, the official side would soon give him an EC to test first, but he decided to selflessly share it with everyone. When the fix was released, the hardware problem was finally fixed.

However, this would be the only meaningful problem fixed in the entire test.

7. Problems never fixed

Once the test machine could run stably at 35W TDP, the whole machine went from “unusable” to “usable.” I started using it to continue Black Myth: Wukong as a test.

This machine is indeed more powerful than the Deck, provided you completely give up battery life. But the screen is terrible. The sticks feel okay, but the white body quickly turned yellow. The sticks also started producing worn powder — after only one month.

The body uses too much glass, so it’s heavy. The D-pad feels awful, no feedback at all.

People reported these issues more or less, but the official side never gave a clear answer about fixing them, just said they’d collected them and would fix them, without saying how. As someone who’s participated in multiple domestic handheld crowdfunding campaigns, I basically understand: if they don’t give a specific plan, they’re not fixing it.

The biggest controversy is that to imitate the Deck, they added two touchpads, but these two can’t operate independently — both left and right emulate the mouse. Not to mention emulating directional keys. These two things not only feel bad but don’t support haptic click. I have no idea what they’re for.

I made several constructive suggestions: could we use Linux’s freedom to write our own driver, use left and right separately, or add vibration motors to simulate clicks, making these two touchpads useful, and the machine could be released as a “SteamDeck compatible machine.”

But the official side didn’t respond. The community said it would require official hardware-level open support, otherwise impossible. Meanwhile the group atmosphere was wrecked by Hardware Bro and Angry Bro, completely preventing constructive discussion. Whatever you said, he’d start with “As a hardware engineer, what you’re saying is just pointless fiddling, just use Windows…” Totally irrelevant.

People from the Linux community also tried hard to register Discord and join the group to collect Linux bugs, but after joining found it was all Windows users. The only two Linux users were his acquaintances — we’d already talked a lot on Discord. Later these foreign friends stopped coming to the group.

And Hardware Bro never missed a chance to show off — he started saying in English “As a hardware engineer…”

This guy annoyed me so much that I muted them all, and only then did the group content become slightly normal. What I hated most wasn’t just the daily repetition of being a hardware engineer, but the constant interrupting, pretending to understand, disrupting the discussion atmosphere. Whatever anyone said, he had to chime in — a complete idiot. A technical test group, and he’s acting like the group pet?

The official staff were very unenthusiastic about fixing bugs, and there was no visible launch date for the “retail machine.” For a Linux First device, the official side’s later work even shifted to “releasing a Windows front-end UI.”

Of course the UI eventually came out, but it was alpha at best, many features unfinished. Hardware Bro said to wait for the future, but the month was almost over — not much future left to wait for.

8. The beta ends: awards, then the exit

The month-long beta went by fast. I put deep hours into one AAA game on this handheld — Black Myth: Wukong — so I’ve got some ground to stand on:

The hardware looks fine. Chasing Linux-first does attract a lot of foreign geeks, and the overseas Linux community is a decent bunch; the guy in charge kept posting weekly progress updates like he meant it. The Chinese factory side, meanwhile — the managers and engineers — went dark at the drop of a hat. No clue what they were busy with, or whether the project was even alive.

The screen is terrible. Cooling is so-so but gets the job done. There are two trackpads and they’re completely useless, and nobody wants to develop a use for them. You get the feeling nobody knew what this machine was for when they designed it.

Officially, the retail version will come in an 8840U model, and if I don’t return the device I keep my discount eligibility. But I plainly don’t believe they’ll get a retail version out anytime soon. Judging by how they communicated during the beta, and the habit of playing dead for days at a time, the retail version is probably still being dragged out at a snail’s pace.

I figured that with the power-supply bug I’d tested and analyzed myself, in that chaotic group chat, I’d at least rank near the top in contributions — maybe get some souvenir. Then on the last day of the beta, they announced the “Outstanding Tester List.” The top two were Hardware Guy and Angry Guy. Me and Student Guy weren’t on it at all.

Mind you, the only people communicating with the community were me and Student Guy — pushing Linux bug fixes, raising suggestions. The Linux community told the company this multiple times. And their “outstanding testers chosen by level of participation” turned out to be based solely on… the number of messages in the Discord group…

In other words: talk the most in the group, and you’re an outstanding tester with perks.

At that point I had zero desire left to continue the beta. It made me sick, honestly. I ended it on the spot, got my deposit back, and shipped the machine out.

I also wrote a letter to the Linux community explaining everything that happened during the beta, and that I was disappointed in the product and the company and would not be taking part further — a goodbye letter. The community people appreciated my work and said they were sorry to see it end. After that, probably nobody will be talking to the community at all.

Then I left every group, contact, and account tied to the beta. No more ties. Seeing those people makes me sick. Though sometimes I console myself: at least I got a free month with a 7840U handheld, right?

9. Reflections

This wasn’t my first product beta. In college I did a batch beta for a “SmartQ” device, and even then the company’s level baffled me — you’d report bugs seriously and get brushed off; you’d make suggestions and nobody listened.

Later I grew up and understood: once a product reaches the beta stage, big changes are basically impossible. All that’s left is patching little software-layer things…

That “SmartQ” I complained about back then didn’t die, by the way — it later became the OEM for Xiaomi’s smart bands.

Back to this handheld. Looking at it from start to finish, the whole beta was absurd:

• Absurd company: no communication, no efficiency, frequently missing, utterly ignoring timelines. The June beta slipped to August; the September-to-October launch date kept slipping too.
• Absurd group members: it’s a Linux device, and almost none of the recruited testers knew Linux. They treated it as a “pre-order.” No idea what use that was to a beta — and in fact there was none.
• Absurd operations: outstanding testers were judged by Discord message count. That alone tells me the product can’t succeed. Rules changed on a whim, the original beta process abandoned — short-staffed, was it?
• And finally, those assholes in the group. I really want to kick that “hardware engineer” twat’s balls in.

One small side story that shows the company’s competence: shortly after the beta started, the test device showed up on Xianyu (the secondhand marketplace). Lots of people complained to the company, and the company’s response was to call on the seller to delist it… Two weeks later the guy was still selling it. Their warning did nothing. What use is this company?

The “beta group” was more like an after-sales group. The members were all pre-order buyers, and the vibe was exactly like every other x86 handheld community: everyone bragging, everyone claiming to be an engineer, everyone saying they’d bought it for productivity — when of course they’d bought it to play games. And a real engineer, in my experience, does not spend all day marinating in a group chat, let alone constantly announce “I’m an engineer” — it makes you look like a total idiot.

As of now, nearly a year has passed since that beta. So did the handheld I beta-tested ever get a retail release?

No.

Yep — the thing slipped another year. Mainstream handheld SoCs have moved on to stuff like the HX370, and the 7840U is already an antique, while this thing still isn’t out, and there’s no telling when it will be. The latest news was half a month ago: someone asked the Linux community why the machine still wasn’t out, and the community said the company was still optimizing…

Which makes me sigh with relief that I bailed early. At least I got a month of play out of it (not even a good month). As for the suckers who kept their devices after the beta hoping to buy the retail model at a discount — who knows if they’re still in that damn Discord group, bragging about things nobody cares about.

Update: I never would have guessed this thing would stay unreleased all the way into 2026, through the memory price spike and shortage — to the point that it got frozen and cancelled. Hilarious: https://www.pcguide.com/news/linux-gaming-handheld-put-on-hold-due-to-ram-and-ssd-prices-as-creator-waits-for-a-good-time-to-a-launch/

Looks like demanding a refund back then was the right call.