Tuxedo Computers slams lid on Arm Linux laptop after 18 months of pain

German Linux box vendor Tuxedo Computers has canned its long-planned Qualcomm device, citing numerous problems with the state of the Linux-on-Arm art. After a year and a half of work, Tuxedo announced that it is discontinuing work on its planned Arm laptop based on a Qualcomm Snapdragon System-on-Chip. There is some good news …

[phuzz]

There is of course one ARM machine with great linux support: the Raspberry Pi. Not super portable though.

[Liam Proven]

> ARM machine with great linux support

You know what, I cut an entire paragraph about that.

It is now in a new separate article coming Real Soon Now™.

To be very pedantic -- a reason is that the Pis are not Linux machines. They are Broadcom VideoCore machine which happen to have an Arm co-processor.

Linux does not boot the things. Microsoft© ThreadX© running on the VideoCore GPU boots the computer, and then once it's up and running, ThreadX loads Linux from a file on a FAT32 partition, sticks it in RAM, wakes up the Arm core and points it at the code.

ThreadX is FOSS now:

https://www.theregister.com/2023/11/28/microsoft_opens_sources_threadx/

The thing is that the Pi version is heavily modified with proprietary drivers, and you can't build the FOSS version and run it on a Pi.

But that's why the Pi is so simple to boot up with no faffing around with DeviceTrees and so on. It's because Linux doesn't boot it at all.

[ICL1900-G3]

Well... I never knew that. Thanks Liam. Interesting.

[Liam Proven]

> Well... I never knew that. Thanks Liam. Interesting.

Oh my yes.

This is probably why they can keep supporting it so long that 12YO hardware can run the latest OS.

If only RISC OS had been in there, Eben Moglen said he would have put _it_ in the firmware... but it wouldn't have worked _as_ the firmware 'cos it's Arm code.

There are efforts at FOSS Pi firmware but it hasn't got far at all...

https://librerpi.github.io/

... partly because the Videocore is AIUI not very well documented in FOSS materials and there's no existing code or anything they can use.

[Pascal Monett]

Microsoft boots the PIs ?

Now I'm disappointed.

[druck]

Microsoft recently bought the ThreadX RTOS, their only contribution will probably be to kill it.

[phuzz]

Huh, today I learnt a new thing. Thanks :)

Looking forward to the article

[Flocke Kroes]

When I first heard about this laptop here I got as far as Qualcom and expected the project to die from lack of CPU vendor support. First rule of software: never buy based on features promised for the next version. Even for a Pi, up to 4 but not quite 5, up to bookworm but not quite Trixie. If the hardware is not supported before you buy do not expect things to get better without pain and suffering. That goes double if you have already spent money. Double again if you are locked into a specific in CPU by your own hardware design.

[RachelG]

No greener grass

I have Asahi Linux on a M1Pro MacBook Pro and... what works works great: Linux always was faster and more responsive than macOS on the same hardware, and, even with a full-fat gnome/wayland desktop, it's still true in the age of Apple Silicon. It's ridiculously quick and smooth, one might make comparisons to spreadable dairy products...

but what *doesn't* work ... it's kind of shocking that it's almost the same list as Tuxedo reports here: No thunderbolt, USB4 doesn't perform well (and no displayport over it), no hardware video encode or decode, and as for power management... the battery life is like using a PC laptop again: If you're going out for the day with it you'd better bring a charger, and it hardly uses less power when it's supposedly suspended, about half the time. I mostly don't use it to save on battery charge cycles, seeing as the machine itself is four years old now (and not showing its age, but then I'm not mashing the battery every day)...

Asahi kind of have an excuse, I don't think they get zero help from Apple but they don't get much. So you'd think with other chip vendors allegedly being more open and supporting the effort it would be a *bit* better on the other side. I might have been interested, especially if any of them made one that wasn't completely fugly. And/or repairability would be nice too.

[Liam Proven]

Re: No greener grass

> Linux always was faster and more responsive than macOS on the same hardware

Really? You find that?

I have Linux on low-end Mac kit here, e.g. a 2009 MacBook. It runs fine but it's no quicker than macOS. I got given a 2010 MacBook Pro a few years ago: Core 2 Duo, 4GB RAM, and a (dead, non-)spinning hard disk. I bunged an old spare 200GB SSD in it and upped the RAM to 6GB -- all I had that it will use -- and put 10.13 on it. Took about 4 hours, because it downloaded multiple firmware updates, rebooted, continued, did more, rebooted again...

But to my great surprise it got there and it works _great_. It is fast, smooth, and responsive. Fancy animations and effects in macOS work fine. On a Core 2 Duo! OK, yes, it has an nVidia dedicated GPU, but still, it's amazing.

In my experience OS X on a fully supported Core 2 Duo stomps all over Linux on a comparable machine, e.g. my Thinkpad W500 with an AMD GPU. That can just about creak along with a lightweight distro like Crunchbang++ with Openbox. It's quite quick with Alpine but Alpine is quick on anything, because it uses as much disk as Crunchbang uses RAM, and about as much RAM as Ubuntu 4.10 did in 2004. Maybe less in fact.

I tried Asashi on my M1 MBA:

https://www.theregister.com/2024/12/20/fedora_asahi_41_out/

I think I wrote about it before but I can't quickly find it.

It is fairly quick but to my eyes Plasma is so retina-scarringly fugly that seeing it on my MBA is heartbreaking.

It has all the grace, elegance, style, and sheer panache of Emile LeRay's motorbike:

https://thekneeslider.com/images/2022/01/leray-citroen-motorcycle-museum-display.jpg

LeRay is an absolute hero and his story is inspiring, BTW:

https://sahara-overland.com/2017/08/05/the-2cv-motorcycle-survival-story/

But pretty it ain't. And KDE on a MacBook felt like that to me.

[david 12]

Re: No greener grass

Myth-busters tried the motor bike and couldn't get it to work. Myth-busters has technical backup that is supposed to do the analysis, but in this case it was obvious that they had not considered the correct design of the steering geometry, which is pretty simple -- see any pushbike -- but absolutely critical to getting the thing to run without falling over.

[Liam Proven]

Re: No greener grass

> Myth-busters tried the motor bike and couldn't get it to work.

I do not understand what you are saying here, or are _trying_ to say.

Motorbikes work. I have owned half a dozen of them.

Mythbusters is an American TV show, I think, and I don't watch TV much at all, but I have heard of it. Not heard of this one. I am not sure what they did. Obviously they didn't buy a bike, because, y'know, they exist and work. And they didn't attach a motor to a bicycle, because that works, and you can buy power-assist motors for bikes, or build your own.

So I don't know what they did or why it failed.

I also know a bit about the geometry of rake and trail, but that's because one of my bikes was a custom, and custom builders need to know about that stuff, and I read BSH.

So, no clue on that.

Meta-message: I am guessing...

"motorbikes work, but Mythbusters failed to build one because they didn't know critical info. KDE works [or maybe, Asahi works?] but you couldn't get it to work [?] because you don't know critical info that others know, just like Muthbusters didn't."...?

Odd message. I do know how to build a Linux system, ta. I've been building Unix boxes since 1988 or 1989 and I've found that, while I don't consider myself a senior specialist expert in the field, actually I know quite a bit by now and a lot of people in the industry are basically just bluffing and they know less than me.

I've also discovered that they are perfectly willing to lie about it in attempts to discredit me.

That's why I'm willing to write about it.

But what I said wasn't about that. It was about my personal aesthetic and functional judgement of KDE.

I loved KDE 1.x, thought KDE 2.x was decent and could be shaped into something superb, and thought KDE 3.x was a bloated mess, a bicycle with 11 more wheels on it (in Guy Kewney's wonderful metaphor) -- a file manager that was also a web browser and an FTP client and an email client... guys, WTF were you thinking? *Were* you thinking?

KDE 4 was worse and uglier too, from pointless widgets to floating cashew nuts.

KDE 5 at least was less ugly because it was flat.

KDE 6 is adding the ugly back and it takes about 5-10x the memory of KDE 5 into the bargain.

Not a win.

Someone should go find Xandros' modified KDE 2, resurrect that, port it to Qt 6 -- and render TDE obsolete in the process.

Perhaps Emile Leray knew more about motorbikes than Mythbusters? And perhaps he was motivated by desperation, and they weren't?

Perhaps I know more about the design of desktops and how they fit together better than the KDE team?

[RachelG]

Re: No greener grass

I really find that. But I've always been in the Gnome camp rather than the KDE one, at least since before any of it intersected with me starting to use macs. And I find Gnome very pretty these days, though I do make it like Ubuntu even when it's not. It is very possible my observation is more about the apparent responsiveness of the desktop environment, where straightline compute tasks ought, obviously, to be comparable. eg: a non-hardware-assisted handbrake encode ought to take the same time and if it didn't you'd wonder why (but of course if I have encoding to do I go back to macOS so it can use the media engine...) So it's completely in scope that KDE could give a very different experience.

But for me it started with a G4 Mac Mini, which I hadn't had very long before Apple stopped making macOS for PowerPC, so it spent most of its life running Linux, and very well. Long enough anyway to be completely slaughtered by a Raspberry Pi (2B I think), both running Mint. Which kind of woke me up to how old that poor G4 had got; why was I still spinning rust and fan to find a use for it when a not-even-current-then RPi completely outclasses it?

Various other macs in-between got the linux treatment, my feeling being that it's just what you do with an old mac Apple don't make operating systems for any more. Currently in that category is a 2019 Retina 5K iMac, now running Ubuntu doing light server duties. The desktop does run fine on it though, except that Linux can only drive it to 4K (and internal speakers don't work, and wifi and bluetooth...) but that's why, despite having bought several over the years, I kind of hate the iMac: display and computer have different lifecycles, and here's a lovely 5K screen being wasted, forever bound to an old computer that now can't drive it properly.

[Bebu sa Ware]

Planned Snapdragon goes puff and disappears

Puff the Smack·Dragon "repatriated" to Honalee ?

[Liam Proven]

Re: Planned Snapdragon goes puff and disappears

> Puff the Smack·Dragon "repatriated" to Honalee ?

I can neither confirm nor deny that at this time.

But I'm very glad someone got the reference. ;-)

[Nerf Herder]

Re: Planned Snapdragon goes puff and disappears

The puff-the-magic-dragon reference is the only reason I logged in to comment. As well as the broad and detailed IT coverage, it's these little things that make El Reg a pleasure to read.

[Liam Proven]

Re: Planned Snapdragon goes puff and disappears

> The puff-the-magic-dragon reference is the only reason I logged in to comment. As well as the broad and detailed IT coverage, it's these little things that make El Reg a pleasure to read.

Thank you very much.

Comments like this really help me to cope with the galloping impostor syndrome.

[C R Mudgeon]

Re: Planned Snapdragon goes puff and disappears

Yeah, Tuxedo ICEd it.

[MarkMLl]

So what's the meat of the article?

So what are you actually saying here: that they've dropped their custom design based on a Qualcomm chip and instead intend to buy in a laptop based on a Mediatek chip?

[Liam Proven]

Re: So what's the meat of the article?

> So what's the meat of the article?

You surprise me, Mark.

> instead intend to buy in a laptop based on a Mediatek chip?

No, not at all.

TL;DR:

• There will be no Arm-based Linux laptop from Tuxedo in the foreseeable future

• If Tuxedo can't, don't hold your breath for anyone else doing one either.

• You can buy an Arm-based Thinkpad and put Linux on it yourself and it should work reasonably.

• If you buy a cheap one, e.g. Medion, it probably won't work well, if at all. It _might_ work with future kernel versions. You may well be able to run OpenBSD but kiss goodbye to Bluetooth, and possibly to GPU acceleration, hardware-assisted video playback, etc. And that means kiss goodbye to battery life as well.

[MarkMLl]

Re: So what's the meat of the article?

> You surprise me, Mark.

That's been on my resume for 30 years or so :-)

So what relevance does Mediatek have to this article? Are you actually saying "if you want an ARM-based laptop look at this one from Mediatek? Because what you wrote was

"[...]and in a somewhat unusual move, it hinted that the Medion SPRCHRGD 14 was the model it aimed to OEM."

which implied that Tuxedo intended to stay involved with the architecture and form-factor by buying something in.

[Anonymous Coward]

Re: So what relevance does Mediatek have to this article?

Medion, not Mediatek (the latter wasn't even mentioned in the article).

Medion is a brand name of computers and electronics products sold across Europe in chain stores like Aldi, Lidl and a number of consumer electronic stores.

Mediatek is a manufacturer of SoCs, some of which use ARM IP.

[MarkMLl]

Re: So what relevance does Mediatek have to this article?

> Medion, not Mediatek (the latter wasn't even mentioned in the article).

>

> Medion is a brand name of computers and electronics products sold across Europe in chain stores like Aldi, Lidl and a number of consumer electronic stores.

>

> Mediatek is a manufacturer of SoCs, some of which use ARM IP.

My apologies to Liam and everybody else for confusing the two.

Mea maxima culpa.

[Liam Proven]

Re: So what's the meat of the article?

> So what relevance does Mediatek have to this article?

None! I never mentioned it. I went back to check.

The bald unvarnished and uncharitable and ugly version:

Anyone can build a desktop PC from bits. I can do it and my knowledge of electronics is stone-aged. Small PC vendors have no problems. All the interconnections are standardised. It is like Lego.

Building a laptop is hard. Very few companies can design and build laptops. There is no standardisation of parts. You can't even put a 2010 Lenovo battery into a 2011 Lenovo. FFS you can't put a 2011 X series Lenovo battery into a 2011 T series, or a 2011 W series. They're all different.

These bloody idiots never watched _Apollo 13_ and learned about the air filters.

Anyway. What this means:

Small and medium PC vendors do not build laptops. They say they do but they don't.

Dell does. Apple does. Almost nobody else does.

They buy ready-made laptops from Taiwan and Shenzen. They order them from Compal or Clevo or a handful of other companies.

Small _Linux_ vendors do not design or build Linux-compatible laptops.

What they do is buy some cheap OEM ones from Clevo with all the bits on offer, test what bits work with Linux, and order a container of that spec.

They get them and put on stickers with their name on them.

The bigger resellers put them in their own boxes.

That is what you get when you buy a System76 or a Tuxedo or a StarLabs or whatever.

Here is my reconstruction of what happened:

Tuxedo found there was a cheap Arm Windows laptop. (Not a Chromebook.) Medion, the Aldi budget computer brand, sell them.

Tuxedo found who makes the Medion, and they ordered some.

They set to work to get Linux to boot with the existing Snapdragon X Elite code.

They got it booting and thinking that was 90% of the battle, they announced it was coming.

Remember the FOSS motto's rephrasing of JFK.

"We do these things, not because they are easy, but because we _thought_ they were easy."

They then discovered that it booted but the battery life was crap, the power management was crap, the ports ran slow, there was no way to update the firmware, and that it had various other bells and whistles that only work in Windows and they work because of proprietary Windows BLOBs that they can't use anyway.

Just like the bad old days of Wifi support and the NDISwrapper.

They then wasted a year and a half trying to fix it. In the immortal words of Douglas Adams...

"And, in the end, they grumpily announced that such a machine was virtually impossible."

And here we are. There will be no Linux Arm laptop from Tuxedo, and that means there won't be from System76 'cos it's busy with COSMIC and anyway I suspect they have sold very very few of their $3000 Arm desktop, because I could count how many people need it *and do not have AWS accounts* on my fingers.

Mediatek, like the plumage, don't enter into it. It's bleedin' demised.

[MarkMLl]

Re: So what's the meat of the article?

OK So what you're saying- if I understand you properly this time round- is that Tuxedo tried to put Linux on laptops which other people brand as Medion, but found it wasn't feasible and pulled out.

[Liam Proven]

Re: So what's the meat of the article?

> found it wasn't feasible and pulled out.

Yes. Finally. ;-)

And the kernel devs don't want the code Tuxedo is trying to upstream.

https://www.phoronix.com/news/TUXEDO-X1E-Cancelled-DT

[jake]

Re: So what's the meat of the article?

"Remember the FOSS motto's rephrasing of JFK."

Also, management world-wide would do well to learn about and internalize the Pareto Principle.

_{Prediction: They won't.}

[Mockup1974]

ARM is a dead-end. Long live x86.

[Roland6]

A viable computer platform is more than just the CPU, ARM and friends need to develop the full motherboard and effectively open source that like IBM did with the PC.

[Liam Proven]

This is wrong in 3 separate and independent ways. Good work.

1.

> ARM and friends

Nope. Arm doesn't do that. Arm sells designs for cores. It doesn't care what you do with them. It will sell you a matching GPU if you want. Lots don't -- RasPi doesn't, for example.

> need to develop the full motherboard

Nope. The entire point about this story is that there is no "motherboard" here. It's an SoC. The whole computer on a chip. In some cases now including RAM and in Apple's case with the Flash on a BGA pad next door to it on the same substrate.

_That_ is why...

Positives:

• they're so quick: very high level integration

• they're so cheap: very high level integration

Negatives:

• you can't upgrade them

• you can't fix them

• you can't swap bits in and out: it means designing, testing, and fabbing a new chip

That's why:

• they're all different

• they're all incompatible

• every SoC needs a new port, and a new kernel, and a new bootloader...

> and effectively open source that like IBM did with the PC

No it didn't, not at all.

Eagle and Compaq and a few others clean-room reverse-engineered it and cloned it. That's why they were called "PC clones".

[heyrick]

"• they're all different • they're all incompatible"

To my mind this is one of the main problems of the ARM devices. With a "PC" [*], it boots as a dumb x86 with some bog standard ports and a VGA style display. There are known ways to probe to find out what is actually inside, but at the very least you can get something that feels like a DOS box.

ARM devices, on the other hand, don't even have a standardised way of booting. You might need start.elf, you might need MLO and it being in a certain place in the directory tree. You might even boot from raw gibberish held in a serial flash chip...

* - I'm mostly talking about BIOS systems here. UEFI is "similar but more complicated". At any rate, the PC starts up identifiable as a PC...

[hmv]

> clean-room reverse-engineered it and cloned it. That's why they were called "PC clones".

Well, not quite; they weren't clones because of the clean-room reverse engineering; they were clones because they worked just like IBM PCs. The clean-room reverse engineering after the clone manufacturers discovered that IBM keept a pack of ferocious attack lawyers and would quite happily unleash them in the face of those who just copied the BIOS.

[Liam Proven]

> those who just copied the BIOS.

TBH sounds completely fair to me.

You can't just copy someone else's code and then try to compete with them by selling something cheaper using the copied stuff.

That is my point here: this stuff WAS NOT FOSS.

[Dan 55]

If ARM is a dead end it's because everyone depends on support from Qualcomm. The same Qualcomm who managed to screw up their devkits. As they can barely manage to support mobile phones, don't hold your breath for PCs.

And as Qualcomm aren't short of money, one can only assume that's the way they want it to be.

[Liam Proven]

> ARM is a dead-end.

That's what everyone said in 1985 or so when Acorn announced it was doing its own CPU.

They were wrong then and they are *way more* wrong now.

There are something like 10x more Arm cores out there than all x86 cores sold since Intel released the 8086 in 1978, including NEC, AMD, Harris, VIA, Centaur/IDT, all of them put together.

These days I believe I've seen numbers showing that every year Arm chips outsell x86 something closer to 2 orders of magnitude: not around 10x but more like 100x over.

Now, on the desktop, Apple ones outperform Intel far enough you can run Intel apps using _emulation of x86 code_ and in the server where if it's FOSS it doesn't matter -- it's been recompiled with all the Gentoo-guy optimisations turned on.

At the rate it is going, RISC-V might prove to be a dead end yet, but Arm? Arm is going great, and Intel is in slow collapse...

[Richard 12]

If we're talking about applications-CPU architectures, Intel x86 is completely dead, it's ARM vs AMD64 now. RISC-V is extremely unlikely to be anything other than an MCU in the next decade.

AMD64 has the huge advantage of standardisation. I can buy any AMD64 laptop, desktop, or server, and know that my software stack will run on it and be able to probe what's there. It might not be able to use all the peripherals, but it will work.

ARM does not, and that's a massive problem for anyone less vertically-integrated than Apple.

[Wolfclaw]

Said once, say it again, too many distro's and not enough resources to develop them, time for some to swallow pride and come together for the killer Linux distro to rule them all, then fork your own flavour.

[Liam Proven]

Oh don't be silly.

> the killer Linux distro to rule them all

Yeah, it already happened. If you knew this subject as well as you think you do, you'd know that: it's Android and its little brother ChromeOS. They Outsell Macs and Windows put together every year for 15 odd years now.

The penguin taliban won't even admit they're Linux distros.

[Richard 12]

It's just Android now

ChromeOS is being eaten by its parent.

Not that many users will even notice, as the Android version will feel basically identical. Might get a new logo, I guess.

[Czrly]

Don't forget the Steam Frame!

Isn't Valve's new-fangled Steam Frame going to be Arm? That's going to run a build of SteamOS – as sure as eggs aren't spherical – and, so, at least *someone* has a card with "Linux portable device on Arm" up their sleeve. Of course, Valve will be targeting a single reference SoC of their choice and likely not a completely new part because they will want to have production running before their order-books open in the new year. And Valve are a behemoth compared to TUXEDO (I'm a fan of both companies but just being a realist) so I could well believe that Valve have resources that just might compensate for Qualcomm's ambivalence to the cause.

But, then again, Valve are also promising not only X86 emulation but also Windows-X86-on-Linux and promising that at gaming-grade performance and latency.

If Valve pull that off, they could change the landscape of user-side computing significantly. If Valve pull off this coup and Arm settles on any kind of stability in their platform so that Valve's work remains relevant for futures Arm SoCs – or variants of whatever comes in the Steam Frame – I wouldn't be surprised if TUXEDO revisit Arm laptops again in a few years – as would other vendors.

[Liam Proven]

Re: Don't forget the Steam Frame!

> Don't forget the Steam Frame!

1. It's custom hardware. This does not compare.

2. It's a pair of bloody goggles for playing games. It is a literal *toy*. It's a fancy Viewmaster.

(For the annoyingly young: https://en.wikipedia.org/wiki/View-Master )

Nobody is going to complain that when they plug in a USB4 SSD to their hi-tech Viewmaster thing that it doesn't deliver good enough bandwidth because *it's not a computer* -- it is a TOY. It is designed and built to be played with.

So the baseline for "works well enough" is quite a lot different for a $300 toy and a $1000 laptop you should reasonably expect to do your job on.

[blcollier]

Re: Don't forget the Steam Frame!

Liam, I rarely have cause to take issue with any of your comments on articles, but I'm afraid I do with this one.

Yes, the Steam Frame is a device intended for playing video games. And let me diverge from the main point for a moment to address 'It is a literal *toy*. It's a fancy Viewmaster.', which comes across as rather disparaging and dismissive.

My desktop computer can play games as well as do 'real work'. The Steam Deck handheld can play games as well as do 'real work'. It's clearly a handheld that's designed for playing games, but it can also be hooked up to external peripherals. It can also switch to a 'desktop mode', where your primary interface is no longer a gaming-oriented UI, but a much more familiar 'desktop' interface. External peripherals aren't a requirement for that desktop mode either, it works handheld.

If you're _truly_ sadistic, you can even run Windows on the Steam Deck. (Although please make sure I'm not around when you do, because it's likely to make me want to vomit.)

So at what point does a 'computer' become a 'toy'? Where is that dividing line, the 'market segment' the device is targeted at, or the capabilities of the device itself?

The Steam Frame is a head-mounted display for playing games, yes, but it has built-in compute and isn't dependent on a video feed from another computer. You _can_ feed video from another machine if you wish, but it can also natively run those games on the built-in hardware. It runs a version of SteamOS, the same Linux-based OS used by the Steam Deck, and it will be similarly capable of running Windows-native games on that Linux-based OS. It'll also have a single-lane PCIe Gen4 port, albeit with what will likely be a custom/proprietary connector.

And it'll do this on an on an ARM Snapdragon 8 Gen 3 SoC.

No, they're not _intended_ for you to write articles or code on. But they probably _could_ do that, and to disparagingly dismiss it as a 'toy' seems very short-sighted, especially coming from you.

The real reason the Steam Deck has enjoyed the (admittedly moderate) success it has is not the hardware itself, but the software - specifically the Proton compatibility layer. Without Proton, the Deck would be a total flop. In fact Valve tried 'Steam Machines' with a Linux-based OS some 10 years ago, and the initiative was indeed a total flop - there were few Linux-native games available at the time and Proton did not yet exist.

Once again, it's software that will be critical if Valve is to make a success of an ARM-based device running a Linux-based OS that can play Windows-native games. This time the software in question is the 'FEX' x86-to-ARM translation layer - that's the basket in which Valve is placing its 'make the Steam Frame a success' eggs. It's progress on this front that I'm personally most looking forward to: just as Proton has driven gaming on x86 Linux, FEX has the _potential_ to drive greater adoption of ARM. And I believe that was the same point that Czrly was making.

For what it's worth... you can bet your sweet bippy that people _will_ complain if storage bandwidth on the Steam Frame is not sufficient - not for USB4, but for MicroSD cards. One of the 'design goals' (for want of a better term) is to be able to install games to a MicroSD card and swap that same card between: Steam Deck, (2026) Steam Machine, and Steam Frame. You can already install to MicroSD card on the Steam Deck, and in most cases the difference in load times compared to an SSD is quite small (at least at the resolutions and detail settings that the Deck is capable of).

[doublelayer]

Re: Don't forget the Steam Frame!

When related to the original dream of Valve somehow making ARM-based Linux more generally functional, the Steam Frame is very unlikely to help at all. We already have plenty of devices that run Linux quite well on Snapdragon 8 gen 3s. Specifically, we have all of these. And what have those meant for other Linux running on other, or even the same, SoC? Nothing. They've got custom kernels that don't get updated and have optimizations for specific hardware. The steam Frame will have exactly the same.

The original comment was breathlessly hoping for miraculous things like ARM changing how their chips work for Valve's benefit. ARM wouldn't even if Valve did have to do extra work, but Valve doesn't need and thus won't ask for architectural changes. Valve is not in the business of making ARM laptops work. A platform they run on a smartphone chipset isn't going to change anything that benefits mainstream computing use cases. If they need to, they might make Proton on ARM more powerful, which for anyone who doesn't intend to run Proton on their ARM laptops is irrelevant and anyone who does just a tantalizing prospect of what might be nice if the more important parts worked.

Whether or not the Steam Frame is a toy or a computer, if your goal is a useful ARM Linux machine and not another thing to run games on, it's not going to help. Valve is not trying to nor will it obtain anyway any of the things Tuxedo failed to achieve.

[blcollier]

Re: Don't forget the Steam Frame!

I think you're missing the point somewhat.

For starters, I specifically mean a _desktop_ Linux distribution, i.e. not Android. Out of all those devices you linked to, I'd be willing to bet that only a single digit percentage could run a distro such as Debian or Arch.

The Steam Frame will run SteamOS, the same Arch-based OS (ARM vs x86-64 compilation notwithstanding) that runs on the Steam Deck. Not Android. SteamOS can currently be installed on any x86-64 machine with an AMD CPU & GPU.

I think the OP misunderstands the fundamental nature of ARM, given the 'if [...] Arm settles on any kind of stability in their platform' comment. ARM _is_ stable, from the point of view of the instruction set and core design, there is no fragmentation or 'instability'. But I don't think it's fair to say that this is 'breathlessly hoping for miraculous things like ARM changing how their chips work'.

But that really is nit-picking, because the wider point the OP and I are making is that: the effort on the software that the Steam Frame will rely on has a great deal of _potential_ to drive wider ARM adoption.

I brought up Proton in my last reply because I still think that was the real boon we got from Valve's latest hardware efforts, and not the actual hardware itself (although it _is_ really good hardware). But in this context Proton is irrelevant - it'll be one part of the Steam Frame's software stack, yes, but it's not really what we're talking about.

The potential advance in this case is FEX, an open source x86-to-ARM translation/emulation layer, and Valve are putting a lot of effort behind it in order to make the Steam Frame run as intended. But unlike Proton, FEX is _not_ Valve's software - it has existed since at least 2021 - and nor is it _strictly_ specific to gaming. Running games _was_ the original goal, but the very first post I found on the project's blog states that '[it] can already run full fledged applications like GIMP or clang'.

Yes, application compatibility is only one piece of the puzzle for wider ARM adoption; yes, there is already a great deal of Linux software available for ARM; and yes, there are still many hurdles to overcome.

But a seamless and performant x86-to-ARM translation/emulation layer is absolutely not a meaningless endeavour.

[doublelayer]

Re: Don't forget the Steam Frame!

"For starters, I specifically mean a _desktop_ Linux distribution, i.e. not Android."

Of course, and my pointing out of Android was meant to indicate why Valve doesn't have to do what Tuxedo did. Tuxedo needed to make improvements which they ended up not being able to entirely complete, in order to make a laptop with an ARM chip have proper power management. Valve doesn't. Why don't they? Because Qualcomm already did that for them in the custom kernel they'll be running on their hardware, likely the same or almost the same as the Android phones run under theirs. What will that mean for other computers running other chips? Nothing at all. What will it mean for laptops specifically running the 8 gen 3? They might have something they could use, but there aren't any and, if one were to be built, you probably couldn't use any kernel but that one which means it will fall behind like all the ARM machines tend to. This is a problem that affects nearly every ARM board running Linux. If it can't run mainline, and it usually can't and retain the features it needs, then it quickly drifts behind. I think this is what the original post might be referring to when they hope for ARM improving "stability", but they're not going to get it.

And application compatibility is not the problem for the same reasons you said. I can already compile most of the software that runs on Linux for ARM. If we had fast and working hardware, those exceptions might eventually become the limiting factor, but they're not now because the hardware is bad. That's especially true for most of the people who would have considered buying the laptop from the article who would almost certainly be running software that has existed in ARM compilations for a decade or more. If FEX becomes perfect and can run literally any X64 software faster than the original, the hardware will still be bad and that will still be the biggest problem. Since Valve isn't going to fix that problem, it's going to stay one.

[Liam Proven]

Re: Don't forget the Steam Frame!

I think you're so angry about the side-issue you are missing the big point,

> let me diverge from the main point for a moment to address 'It is a literal *toy*.

It is. It is a 100% leisure device for recreational use. So is the Steam Deck.

_Those are not bad things_. Just because _I_ don't want it doesn't meant I am saying they're bad. I am not.

Would it be a problem for its intended use if the Steam Deck in desktop mode had poor bandwidth to/from external storage? Would it harm sales? No. Because that doesn't matter for its intended purpose.

It's a handheld computer but it can be used for productivity stuff if you need that.

Do you think anyone is likely to do that with the Steam Frame? Do you think they will want to plug it into a USB-C hub and connect mouse, keyboard, and monitor and work on it?

Sure, you _could_. You can do that with a smartphone today.

But almost nobody does. This does not hurt phone sales. And as a result they are not very good at it and that doesn't hurt them as products and because it doesn't there is no pressure on them to get better at it. A phone that's a better desktop PC doesn't make it sell better...

And as you can now get a perfectly good phone for under £150 and a perfectly good desktop for under £150 _as well_ then there is no pressure to make one £300 gadget that does both.

Nobody cares if their phone is slow at accessing USB4 storage and nobody will care if their Steam Frame is, either.

It will not hurt Steam Device sales.

But if they buy a £900 Arm-based laptop, you bet they will care!

THAT IS THE POINT HERE.

A phone is a dedicated device. An insanely versatile one but it's not a computer. People don't use them as computers and don't care that they're not very good at it.

This applies even more so to a handheld games console, and even more than that to face-mounted 3D goggles.

Real world parallel: in America the bloody Cybertruck sells tolerably well to testosterone-poisoned MAGAt types. They like trucks. They like 4WD. Here's a cool looking 4WD truck that upsets "the left". So, it's selling. It's a crap truck. It's useless offroad. Doesn't matter.

Look at most 4WD hybrid-offroaders _in the world_. They are never ever driven offroad. Maybe parked on grass occasionally, that's about it. The owners don't care. That's not what they're for. It's a symbol. They want to think they _could._

Laptop owners want a general-purpose computer for all sorts of purposes. An SUV of computers. It's portable and it can play games and you can watch movies _and you can work._

All laptops do that.

Like the vast majority of SUVs never leave the road, the majority of laptops never leave the desk.

But Apple sells Arm ones that have spectacular battery life and great performance and don't run hot and they're super slim and light.

It is natural a Linux vendor might want a piece of that action. So one tried. It went wrong. I tried to explain why.

You were too busy being angry about me being dismissive of gaming to address the fact that issues that are completely irrelevant to a face-mounted gaming device are deal-breakers for a laptop that's 3-4x the price.

And there is another parallel here. Not all cars are SUVs. Cheap little basic boxes still sell.

Not all laptops are flashy all-rounders. Chromebook owners don't care. Chromebook owners drive the cheap economy car, or take public transport or walk. They paid £200 and they don't want an SUV computer. It lets them do their email and homework and shop online and that's all they need. It's a web client that does nothing else, and they sell by the millions per month.

If these issues affected a Chromebook those owners wouldn't care either.

And that is right, fair, reasonable and understandable.

[Dan 55]

Re: Don't forget the Steam Frame!

Many things Valve did and is doing with the Deck found their way upstream. So if the Frame is a product which is built around a Snapdragon SoC, Valve will upstream code which will help running standard Linux on ARM. Hopefully driver/SoC code which Qualcomm isn't upstreaming.

Has Google's Chromebooks helped getting standard Linux to run on ARM? Maybe? But not in a way that I can tell. Google have deigned to allow Chromebook owners to run some Linux commands in a locked-down shell on a locked-down device though. But that's not a real Linux, it hasn't got enough GNU. Android is even more locked down.

[Taliesinawen]

The Qualcomm-Microsoft contract

Microsoft and Qualcomm Technologies collaborate on Windows

[Liam Proven]

Re: The Qualcomm-Microsoft contract

> The Qualcomm-Microsoft contract

Yes indeed. Excellent point.

There's an industry rumour that such a deal has been in place for a decade or so now and it is a major factor crippling Linux on general-purpose Arm computers.

The trouble that I can't find any evidence that I can cite so I can't really write about it.

Tuxedo's problems could indeed just be another aspect of this MS/Q deal.

The brilliance of Valve's move is that it's not trying to do a general-purpose Arm Linux computer so it might be able to dodge the bullet.

[mark l 2]

The failure here lies squarely with Qualcomm, they could easily afford to put some of their devs on a full time project to get the Snapdragon SOCs working well with Linux, but don't seem to want to bother despite them promising to support Linux when they first announced them. Its a shame as when all those Snapdragon laptops become EOL and start to flood the second hand market they could be put to new use running Linux, rather than the bloated AI infested mess that Windows will be.

[Taliesinawen]

The failure here lies squarely with Qualcomm

mark l 2: “The failure here lies squarely with Qualcomm, they could easily afford to put some of their devs on a full time project to get the Snapdragon SOCs working well with Linux ..”

No doubt it's forbidden in the Qualcomm-Microsoft contract terms of service.