At its Build developer conference Tuesday, Microsoft made a few announcements aimed at bolstering Windows on Arm. The first is Project Volterra, a Microsoft-branded mini-desktop computer powered by an unnamed Qualcomm Snapdragon SoC. More relevant for developers who already have Arm hardware, Volterra will be accompanied by a fully Arm-native suite of developer tools.
According to Microsoft’s blog post, the company will be releasing ARM-native versions of Visual Studio 2022 and VSCode, Visual C++, Modern .NET 6, the classic .NET framework, Windows Terminal, and both the Windows Subsystem for Linux and Windows Subsystem for Android. Arm-native versions of these apps will allow developers to run them without the performance penalty associated with translating x86 code to run on Arm devices—especially helpful given that Arm Windows devices usually don’t have much performance to spare.
I actually wouldn’t mind one of these as an actual product for regular end users. Windows on ARM needs a big push, and while I’m not sure these announcements constitute such a big push, it’s at least something.
Nice,
I might be getting a proper ARM Windows desktop before a Linux one. (I already have Raspberry PI various versions, including 400, but saying those are “underpowered” would be be understatement).
I could not find much information, but I am hoping these machines will support Linux. Either natively, or through WSL. I would be cautiously optimistic on that one.
sukru,
Microsoft locks down all ARM hardware to microsoft secure boot keys and owners are not able to disable/override secure boot on ARM. However I don’t know if microsoft has a program to sign linux distros under their bootloader key on ARM. I couldn’t find information about this, does anyone have a link to clarify this?
I suspect microsoft’s intention is to keep ARM more locked down in the future and make it harder to install alternatives. That sucks for me because my main interest in an ARM computer would be to run linux. Ideally I’d have an ARM computer that I can build my own distro for. This is easy on x86, but ARM is nothing but problems. I personally have no interest in WSL on top of windows either.
> Microsoft locks down all ARM hardware to microsoft secure boot keys and owners are not able to disable/override secure
> boot on ARM. However I don’t know if microsoft has a program to sign linux distros under their bootloader key on ARM. I
> couldn’t find information about this, does anyone have a link to clarify this?
My laptop (with which I’m writing this message) is a Lenovo Yoga C630 (the 13-inch version, there is a 15-inch one with the same model namewhich is x86-based), and it uses Linux. The CPU is a Snapdragon 850. The installer I used is _almost-almost-almost_ Debian’s (this is, it installs a couple of firmware-related packages from Linaro repositories). There are several hardware-related glitches, and some devices are just not recognized/declared on the devicetree, but the computer works quite snappily; it is not a compiling powerhorse, but it serves very good for most tasks I require from it.
You can look for further information in https://github.com/aarch64-laptops/ or in the #aarch64-laptops channel in irc.oftc.net.
gwolf,
That’s interesting. Can you clarify if you are booting using microsoft’s key or are you able to disable secure boot?
I’d interested in an ARM laptop w/linux. Lenovo’s website says these are no longer available new, but probably could find a used one. I do see posts of people having problems running linux on this specific model though, although I don’t know why.
https://www.reddit.com/r/linuxquestions/comments/l2dykm/installing_linux_on_the_lenovo_yoga_c630/
Off topic, but the Yoga’s keyboard is a turn off IMHO. I hate with a passion keyboards with scrunched & overloaded navigation keys that are virtually impossible to use efficiently. I don’t know why so many keyboard designers are inclined to copy aweful keyboard layouts, even when they had space to do things properly and comfortably. For me typing is too important and I won’t even consider a laptop with a bad keyboard. I’ve been there and the inability to touch type was just a miserable experience.
Alfman,
That would be unfortunate if the system is locked into Windows. Maybe it makes sense to wait for early reviews before getting one.
Btw, I have been on the edge for buying an Honeycomb LX2:
https://shop.solid-run.com/product/SRLX216S00D00GE064H09CH/
But it is a very old design (Arm Cortex A72), and with the same budget it is possible to build a 20 node-Raspberry Pi or 15-node-Jetson Nano cluster. Though, they are not in stock anymore
Re: WsL on Windows: It actually runs okay. But they have locked new features to Windows 11, which feels like coercion at this point.
I’ve been looking at the Solid Run boards too. Paired with a video card it should keep up with my i5-4590, but I’d probably use it as a server.
Fedora Magazine had an article on it last years.
https://fedoramagazine.org/fedora-aarch64-on-the-solidrun-honeycomb-lx2k/
There are also the AVA Developer Platforms with the Ampere Altra SoC. ~$5500 though.
https://www.ipi.wiki/pages/com-hpc-altr
https://www.adlinktech.com/Products/Computer_on_Modules/COM-HPC-Server-Module/Ampere_Altra_Developer_Platform?lang=en#tab-ordering
Flatland_Spider,
My “deal breaker” was even at that price level ($800) it was too “bare bones”.
No on board GPU, limited PCIe, and storage (single m.2), but having lots of ethernet / SFP+ ports means it is not directed to the desktop or workstation market, but rather as a network appliance.
I already had a Intel Atom based server doing a similar work, hence could not get myself to do the expense.
But it is very close. There were no updates in the last ~2 years, but after maybe this covid draught, I could see them bring up a more workstation oriented board.
sukru,
It’s supposed to have a pretty good serial console.
It’s slightly more then some mATX server boards, but not too bad, and it’s similarly equipped to other boards in the same category. ASRock Rack has a Ryzen board I’m interested in, and an equivalent with 10G. I think the 10G Ryzen board is more expensive, with less networking, then the Solid Run board.
It is definitely a network dev board. There is an updated version with x8 1G ports.
I don’t have an equivalent at the moment, so I might as well. If I had a 16-core Denverton, I might pause, but I don’t.
What do you mean by proper, those people with abandoned Windows RT based Surface and Surface 2 might ask what when wrong then and why now?
I never had those RT devices, and I don’t think Microsoft was very serious with them either. They did not receive a proper version of Windows nor they had good x86 emulation. Fortunately that it all in the past.
It was a “proper” version of Windows, fully capable. The problem was it lacking native software because at the time MS didn’t want anyone to write real software for ARM, only that stupid “Metro” crap that made writing decent apps impossible. However, some open source stuff such as Paint.Net has been ported natively to Windows RT.
I know that they had some limited apps. But Win RT API did not take off.
And unlike Apple’s M1, it never received 3rd party productivity applications. If I recall correctly, it did not even have alternate web browsers, and limited to IE/Edge only.
Anyway, today Win RT evolved into UWP, and even that is evolving.
I think (could be wrong on this) that the reason Linux seems weak on the RPi 4 and newer is mostly due to no GPU acceleration. I think they are still not fully open. Pretty sure my pentium 3 1.1ghz with a Matrox G400 in it actually ran Gnome smoother than I would ever seen my RPi 400. But the CPUs should be faster than that ancient thing!
The GPU is partially open source with Vulkan:
https://www.raspberrypi.com/news/vulkan-update-version-1-1-conformance-for-raspberry-pi-4/
However it is a very old GPU core design, and shows its age on modern displays. For one reason or another Broadcom and Raspberry foundation did not upgrade to a better GPU.
sukru,
I think it might come down to RPI having to choose between a GPU the community has put a lot of work into and is already well supported versus switching to new GPU and having to start over with new proprietary drivers. I think it’s a tough call for RPI because everything they do has to compromise for problems created by others. While users understandably complain about older GPUs, they’d also complain about new GPUs with regressed support. So IMHO it makes sense to hold things back for the sake of quality. Eventually RPI should get a new GPU, but realistically it may take time for the community to support it as well as the old GPU.
It’s probably good enough for the target market. Until it needs to drive 8K monitors they’ll probably keep the GPU.
Flatland_Spider,
I think so as well. I don’t expect it to run intensive games, that’s not the target. That said though it would be nice to have a high end GPU for opencl & machine learning.
Alas, I think software support is the bigger bigger hurdle right now.
https://forums.raspberrypi.com/viewtopic.php?p=1875150
https://github.com/doe300/VC4CL/issues/86
Flatland_Spider, Alfman,
I would say the culprit is video.
The GPU is good enough for “Windows Me / Gnome 2.0” level basic desktop. However modern desktop has become dependent on modern GPU features.
Even ignoring “composition”, simple things like h264 video play requires going though hoops:
https://lemariva.com/blog/2020/08/raspberry-pi-4-video-acceleration-decode-chromium
And add in video calls, which now use ML based systems for background blur.
… and even RPi 4 becomes encumbered.
Anyway, given open source support it is still one of the best options out there.
Alfman,
It would be nice. I have a workload that would work really well with an RPi + performant GPU.
sukru,
Yes, they have. The experience is really bad if the video drivers have problems.
That’s a good point. I haven’t been able to get an 8GB RPi to really test everything out. I would probably pick something like Sway instead of Gnome though.
The Chromium hardware acceleration problem isn’t specific to the RPi. A special version of Chromium has to be installed on x86 Linux to get hardware accel too.
Yeah, Gnome 3 is heavily dependent on video drivers with hardware acceleration. It was originally pretty crappy until the Linux video drivers leveled out.
I’m not expecting much in terms of horsepower, but it might be good value for people who just need an average desktop. I’m always a fan of more competition and of all the companies microsoft is probably in the best position to bring standards to ARM….but if they lock down hardware and deny owner control, that’s a major detractor. Personally I want to support the vendors with the most open hardware, it’s just that with ARM they’ve all been bad
I wouldn’t touch one of these with a 10 foot pole, remember Windows 8 on ARM? Say what ya want about X86 but if it gets abandoned by whomever makes it then its pretty damn simple to just slap some version of *nix on it, with ARM so much of the hardware is black box that it can be hell finding an updated supported OS that will run on it that isn’t the OEM.
Pretty sure these are more in line with the developers need something so they don’t have to wait until the heat death of the universe to finish compiling. No one else should by one as once that M1 competitor shows up, this will likely be garbage no one wants at all. Interestingly, there are already a few Windows 11 ARM desktops that snuck by most of the press, but their performance is abysmal.
The real problem here is to believe on MS commitment for that.
Even worse, on the commitment by Qualcomm to actually compete on the desktop market. Their offerings so far on that arena has been dismal, only for portables (laptop and convertibles) and looks to be the only ARM manufacturer (that is, the only one that is the business of selling SoCs) actually trying .
Is Qualcomm trying? I mean is engineering trying; the sales team is obviously trying. XD
Given that their offerings are retrofits from smartphone processors, no, their engineering don’t give a damn. And that means that a lot of folks above the engineering neither do as well.