Desktop usability worse than in past?

VisionFive 2 （English Forum)
1

The above linked video shows a game, a video, and glmark2 running in a seemingly smooth operation. It is in xfce, and it is one year old.

Why has the current state of software available not reached this level of desktop usability? When I try xfce, it is broken, unusable in the latest release on the local machine. The mouse cursor pointer is missing. Menus do not work, and other setting menus or control panels fail to open completely.

When I try Gnome Wayland, it is usable somewhat, but glmark2 produces very poor results.

When I start xfce4-sesion over and ssh terminal with x forwarding, it works fine!

I read the gpu doesn’t do anything other than 3D acceleration. What is the issue with 2D desktop acceleration? Why does the desktop work less good than in the past?

Is it the HDMI? Configuration? Video out? I am not using 4k monitor. I am just seeking a smooth desktop experience, but I can’t figure out what the actual issue is.

Why have all of the early demo videos of this board used xfce (especially form China), but the engineering release strongly support Gnome and wayland? It makes no sense.

1 Like
2

I’m not speaking on the behalf of Starfive, I’m just a VF2 riscv64 fan boy.

All Linux distros x86_64 and arm64 are migrating towards Wayland. You’re lucky to have any form of X compatibility being provided by xwayland to be honest. You’re not the only one experiencing that sluggishness. I am feeling the same pain too, but I do welcome gnome-on-wayland and eventually it will perform just as well as on x86_64 and aarch64 at some point.

3

As I keep reading this, the elephant in the room becomes clear to me - what I seem to have bought is a development board for a set-top box chip, not a universal risc-v PC in development.

It’s evident because support for everything but wayland and GPU is being set aside. It’s up to “community” to decode the standard GNU stuff and provide the big gnu/linux system sellers. Even if the providers will, in the end, benefit from their product supporting certain “legacy stuff” like X11. Or say, docker and asterisk.

It’s okay, really! Everyone has their needs and wants. But the expectations should have been made a little clearer, because this device is marketed as being officially worked on providing Linux support, not only latest linux. – well this is on me since I kind of assumed my “rpi 3 grade” device will not be able to handle the latest desktop experience (Gnome & KDE) that well, so expected a different developer’s focus. Still that seems to be the case and it caught me by surprise, and caught some other users by surprise as well, it seems.

In the end it’s just misunderstanding … that some of us ossified legacy whiners bought into.

4

I think you’re mistaken. It’s not a set-top box chip. I think it truly is meant for a “universal risc-v pc” as you wanted. It’s still early days for risc-v pc. Lots of kinks to iron out no matter what other people say. I still don’t have a desktop on archlinux VF2 that I can easily just install and run. I still don’t have a fedora linux VF2 image or a fedora silverblue VF2 image although I heard Sophgo has a fedora linux image, but it targets a different riscv chip.

I am starting to realize the riscv ecosystem is going to be just as fragmented as the aarch64 ecosystem. We won’t have one image of linux that we can simply boot and run on any riscv64 board and will run and take advantage of all the chip/gpu/crypto hardware.

It’s still early days, but if the risc-v ecosystem doesn’t unite their SOC’s/ecosystem(get their shit together), i think the entire risc-v ecosystem are going to lose this opportunity like the aarch64 ecosystem has. I would have thrown more money in to risc-v hardware, but it has to have certified, repeatable capabilities before I buy the next time round.

As part of my shopping checkmark criteria from now on, for each claimed supported OS:
-does this board have claimed OS live installer image to simply boot and install on any attached storage device? prove it. At present VF2 does not.
-does this board’s crypto hardware have claimed OS driver support out of the box? prove it. At present VF2 does not.
-does this board’s gpu hardware have software support out of the box? prove it. At present VF2 does not. VF2 gnome on wayland is not easily installed/configured on debian, archlinux, fedora using all the gpu, all the crypto.
-does this board’s network hardware have software support out of the box? prove it.
At present, VF2 networking does seem to work on both ethernet ports.
-Can this board’s claimed OS be easily upgraded using the usual cli claimed OS upgrade tools? i.e sudo apt-get update/upgrade, sudo dnf upgrade, sudo rpm-ostree upgrade, sudo pacman -Syuu. prove it. At present, VF2 does not.

I think if the SBC can do all the above checkmarks, the board is something that all users can work with as expected without any disappointment. New hardware isn’t supposed to waste its users time. It’s supposed to save them time. That’s why I am not a proponent of both aarch64 and riscv64 hardware yet. Ditto for Powerpc. Hardware has to be affordable, but the os software ecosystem has to be there and especially with a live OS installer image to test it on any SBC without it being installed yet. It brings confidence to the user everything he needs will be working before installing said OS without wasting any time. VF2 doesn’t pass this test yet unfortunately.

What does this all mean? My x86_64 laptops and x86_64 are still my go-to hardware until these issues are resolved. I am hopeful they will be. Governments world-wide should be encouraging such endeavours because security through hardware diversity is a good thing. Unfortunately we are living in a world filled with crazy geopolitics at the moment. I wish there were a new U.N. resolution stating if any world leader declares war, they should be rightfully given capital punishment to encourage and ensure world peace. The world leaders’ words and actions will be the first to feel their own impact in order for them to ensure they weigh them more carefully. At present, the world leaders are the last ones who feel the impact of their own words and actions unfortunately. The same kind of pattern could be used for other geopolitical policies regarding technologies as well.

I think the fastest way to resolve this is to fly all the world leaders out to Gaza, Ukraine and other world hotspots and have them live in the same levels of quality of life as those currently residing there. Have them eat the same food available as the locals. Constrain them use the same tech as the locals. Constrain their spending budget to the equivalent of what the locals have at their disposal. Constrain them to live in the same kinds of living accommodations at their disposal for those of lowest rank to understand their plight. Unless the world leaders themselves feel the discomfort and pain, the world leaders will do nothing about it and totally ignore it. World leaders need to live with the lowest common denominators of everything in order for them to understand they need to fix all this with expediency.

1 Like
5

Well, I’m going to give it a break for a month or two. I have projects I need to finish. All of this speculation is great, wondering and waiting. But none of that answers my original question - why the shift from a year ago. If xorg and xfce was working as displayed in the videos, why deliver a software set that does not match the video expectations?

I don’t know what I will do with this board at this point; I was very much hoping that risc-v was going to be great.

I bought a LattePanda x86 board. More expensive, but then I know the software will work.

1 Like
6

While I sympathize with you all here, it needs to be pointed out, that starfive is actually putting in an effort to upstream hardware support. There have been way too many RISC-V (and even some arm) SBC that dropped with some custom fork of a Linux distro, seeing no love and being left to rot. Theoretically, once all hardware support is upstreamed, you should be able to just put any off the shelf Linux on it and it should work.
As for GPU support, software support for that is in the hands of Imagination. IIRC, they did say, that they planned to open-source/upstream their software, but I cannot remember where I have heard that.
Another thing that can easily be forgotten is also, that their kernel fork is mostly based on an ancient kernel version (5.15 released in October 2021), which in and of itself can lead to problems with newer packages. I had an annoying issue for example, where compiling such an old kernel with a newer compiler resulted in a neutered kernel that could not load any modules.
Another issue is, that compilers still fail to address RISC-V CPUs properly. While C compilers can somewhat produce acceptable ASM, they still struggle with some RISC-V extensions and have only recently started to be able to vectorize, which is a standard feature we have started to expect by them doing it on arm and x86 for many years now. As for other compilers, they still struggle to even support RISC-V. C# support for example is nowhere near where it needs to be for us to start considering RISC-V as a valid option for desktop-type uses.

All these problems together make the RISC-V and VF2 software landscape the mess that it currently is. It’s still a construction site, we are trying our best to already live in.

While I see, that their custom images are in a regrettable state right now, I believe they should be given some slack, as their main effort is in upstreaming, and once that goes through, custom images should be a thing of the past anyway.

Edit: Fixed my incorrect use of the expression “to give someone slack”

2 Likes
7

I am starting to realize the riscv ecosystem is going to be just as fragmented as the aarch64 ecosystem. We won’t have one image of linux that we can simply boot and run on any riscv64 board and will run and take advantage of all the chip/gpu/crypto hardware.

This is exactly what starfive is not trying to do. The main reason the aarch64 ecosystem is so fragmented, is because every board vendor tries to “provide support for their board” by maintaining their fork of a distro of their choice without actually upstreaming anything. If they were to properly upstream support, they wouldn’t need to maintain their fork anymore.

especially with a live OS installer image to test it on any SBC without it being installed yet

A lot of x86 being able to do that is down to their fleshed out UEFI support, which allows the board to do alot more without an operating system or bootloader. This in turn allows bootloaders like grub that work on a much higher level than bootloaders like U-boot. RISC-V and ARM don’t really have something like that right now, as they both came from the embedded world, where you try to avoid bloated ROMs as much as possible. Luckily, EDK2 support for RISC-V exists, which makes UEFI on RISC-V possible. Now it’s up to board vendors to actually implement it, which is being done (SOPHGO, for example, is trying really hard), so again, it should be a question of time.

2 Likes
8

I think you meant that you will “cut them some slack”. (picture a slack or loose rope and a taut or stretched rope)

9

Yes, I excuse. English is not my first language.

1 Like
10

I think it’s obvious this board’s primary customers are not your everyday computer users, and hardly anyone here bought it to specifically use for a project now. So nobody’s at a loss.

For my part, I’ve got plenty of computer power, so it’s not something I’d need working; I bought this because three years ago I moved most of my stuff to arm64 through sweat and tears, so I wanted to do it all over again on an even cooler tech, to put it blutly.
And boy, it’s WAY EASIER to get stuff working in rv64 than arm64/gnu for some reason.

That’s just it! My board met my expectations! It makes risc-v look cool and new, all it needs now is an army of people better than me. Which it has little chance of getting if everyone is so discouraged from using it because “it runs slow at this point.”

Guys. Look around the forum. How many people are enjyoing their VF2 in full graphic mode?

I mean, I appreciate the effort in explaining exactly why it needs more time to be usable, but in the meantime I’ve had it run useful stuff like you would run on a normal PC and honestly had great fun with it until … well, until I activated the official graphic stack. Now it’s almost useless and I realize why it gets so little publicity outside this forum and the initial push for rv tech.

This board cost me 110 euro, and I could have used that money to get my son an oPi 5 instead. I chose to support this board and this developer, and I like said, I am at no loss, kids can just play the minecrafts on laptops, so with that in mind:
Are we not allowed to voice opinions on anything…?
There is a legitimate question here. Why is the desktop not working as it used to?

I’ve read these walls of texts that boil down to ‘cut them some slack’ before. And’ve agreed to them, so can we go pass this?

So the question now becomes technical – why is the desktop not working? Can we help?

Honesly, this is a developer board in development, anyone who bought it can be at this point thought of as a developer. I bought this for the openness, among other reasons. So how can I help? Specifically, I want to help fix the modesetting driver. Where do I go?

3 Likes
11

I agree with everything written here, but my question still stands:
Either the videos released more than 1 year ago were factually showing the VF2 board running an X11 graphics stack with multiple OpenGL/CL/Mesa windows playing, or they were emulated.

If the former, why did we loose functionality, what did we gain.
If the latter, why make a video that is untrustworthy?

My question is pretty simple I feel.

1 Like
12

I will respond here, as I feel like your response was somewhat prompted by my earlier wall of reasons on why it’s currently not up to snuff.

Giving reasons on “why is all this stuff taking so long” was not really the point. My primary goal was more in the vein of “the fact that it’s taking so long is a good thing, and we should be applauding it”, as I feel we are at a somewhat crucial point in RISC-V’s history. Right now, RISC-V is at the point where it can be a Raspberry Pi replacement. But we are also at the point where we decide if it will ever be more than a Raspberry Pi replacement. With all the growth RISC-V is seeing, I believe it’s clear, that if RISC-V grows up correctly, it could actually be capable of dethroning x86 (which I see as something worth striving for). So now is the time to pour our time into the projects of the RISC-V sphere, that actually want more out of it than “just another Raspberry Pi”. From what I’m seeing, the VF2 is one such project that makes an effort to be more than some cash grab hip RISC-V SBC.

As for what my VF2s are doing: I got my RX 580 working by simply recompiling the kernel with it enabled (something ARM boards typically cannot) and now I’m trying to get Gentoo working as well as possible. Have the VF2 be a first class citizen to it. Have it compile its own kernel and so on. As such, I have actually done a “boot from live ISO and partition a drive” process almost straight out of the Gentoo AMD64 handbook with slight adjustments to accommodate the oddities of the VF2
From this it’s clear to me, that ARM SBC are what they are “because that’s all they can be”, but the VF2 is what it is “because that’s what we’ve managed so far”.

This is exactly what I’m trying to say. It has seen barely any work, and it already runs so well. What could become of it once it sees actual work put into it?

As for where to go to help with the modesetting driver? I would assume you’d have to go to Xorg. But considering Xorg will die, and RISC-V needs so many more people working on it as it has currently, I’d honestly feel it a waste to invest the little work it’s getting into something that has its expiration date already set in stone.

TLDR:

While this board can be “just another Raspberry Pi”, please let it not be “just another Raspberry Pi”, because it can be so much more. I’m afraid we are losing the bigger picture.

3 Likes
13

I would assume, the original video was a cobbled together system held together with sweat tape and prayers, intended as a proof of concept to show that the board can do it. I would assume, they did that to show the goal, and then tried to get it to do that through proper implementations and through a system that will not fall apart after 1 reboot.
Such proof of concepts are actually quite common.

But as I have not done much research into that video, I cannot guarantee that’s what’s going on.

2 Likes
14

This is exactly what I’m trying to say. It has seen barely any work, and it already runs so well. What could become of it once it sees actual work put into it?

Less! More work put in by the dev made it less useful in my case.

As for where to go to help with the modesetting driver? I would assume you’d have to go > to Xorg. But considering Xorg will die, and RISC-V needs so many more people working > on it as it has currently, I’d honestly feel it a waste to invest the little work it’s getting into > something that has its expiration date already set in stone.

This awful, patronising argument needs to be dropped. This is not Windows where things have expiration dates set by the developer.

X11 has had its expiry date set in stone for the past 10 years. By choosing this path, this device is going down with Wayland, userbase and public be damned. This is not the GNU way.

TLDR:

While this board can be “just another Raspberry Pi”, please let it not be “just another > Raspberry Pi”, because it can be so much more. I’m afraid we are losing the bigger
picture.

See now this is EXACTLY what I fear will happen with developer leaning heavily on a proprietary GPU in a graphical system that is about as transparent as a brick at this moment.

Edit:

Either the videos released more than 1 year ago were factually showing the VF2 board running an X11 graphics stack with multiple OpenGL/CL/Mesa windows playing, or they were emulated.

If the former, why did we loose functionality, what did we gain.
If the latter, why make a video that is untrustworthy?

First question, the cursor is smooth now, even though it’s glitchy. Some stuff now functions that didn’t before.
Second question, it seems standard advertising at this point to show your SBC doing stuff it won’t be able to right now, or for a while.

What’s funny to me is that someone will eventually have to work on xwayland, which will just bring along X support anyway.

2 Likes
15

Most of the maintenance work on X11 was done by Red Hat developers as they needed it for RHEL. Red Hat has officially dropped X11 support and will likely stop maintaining it once it leaves their LTS phase. As X11 is a huge monolithic project that does quite a lot, it needs quite a lot of maintenance for it to remain functional. Considering Red Hat has carried that burden mostly by themselves up until now (not very GNU of the community, BTW), it is very unlikely, that the community will step up and continue support for X11. I know my argument was somewhat patronizing. I would much rather people work on X11 RISC-V support than not work on RISC-V support at all, I agree. But I would also much rather have people working on RISC-V support in dotnet than X11. Considering I have no say, however, I can only hope, and that argument will remain mostly a thing of opinion.

1 Like
16

The latest release was X11R7.7 (2012-06-06), I would call it nearly dead, it is still on life support right now. I have set of X11 Release 5 programming manuals, so I really do love X11. I wish that Wayland was not the replacement for X11, but …

1 Like
17

Most of the maintenance work on X11 was done by Red Hat developers as they needed it for RHEL. Red Hat has officially dropped X11 support and will likely stop maintaining it once it leaves their LTS phase. As X11 is a huge monolithic project that does quite a lot, it needs quite a lot of maintenance for it to remain functional. Considering Red Hat has carried that burden mostly by themselves up until now (not very GNU of the community, BTW), it is very unlikely, that the community will step up and continue support for X11. I know my argument was somewhat patronizing. I would much rather people work on X11 RISC-V support than not work on RISC-V support at all, I agree. But I would also much rather have people working on RISC-V support in dotnet than X11. Considering I have no say, however, I can only hope, and that argument will remain mostly a thing of opinion.

Great, so now we get into Linux politics at this point – we can go all the way to xinit vs. systemd and OSI vs FSF. RHEL developers are official guys. They get paid to do their jobs, and thus they will follow certain directions of their employer or financer. Both RHEL and the big business benefitted from the GNU way and Stallman’s vision, and they are using free software because they have to – it’s simply better than anything they could muster up themselves. They are the ones who approached GNU, not the other way around.

Why are people talking like RHEL are the only Linux developer out there? Like they made all the code from the ground up, and something as complex and delicate (“=bloated”) as X11 was like a gift from them to humankind and not in fact based on countless hours of several hundred “enthusiasts” who contributed bugfixes, configurations, userspace drivers, window managers, toolkits all in order to make X11, both in this case and in particular, the working system that it is today? That’s right, a working system, something they themselves are unable to produce even when sponsored by bigwigs and given free reign …? I’m sorry, but this is quite triggering for me as GNU birthed and sold both Linux and RISC-V only to be thrown out.

Which ties into the point of this topic. Yes big business now “support” Linux, which means they get to decide what comes and goes. This is in fact something a lot of people frown upon (as they vaguely recall a time when the OS was more ‘free’), as big business now acts as if Linux can’t exist and function without them. So they get to say: no more X11! … like this was a Windows thing. Great, so now what?

Shelf the vision five 2, this free and open source computer, until someone official makes it work. No, don’t use it like that. Wait until you can use it like this.

Why did I buy into the RV64 ecosystem again? Wasn’t it for the freedom?

This is where we get back to “why is the desktop not working.” It’s not working because it’s not supposed to until the developer gets it working the way they want it to, no different from an Apple or amd64 system and OS.

1 Like
18

X11

Looking more into it, I cannot find a usable source naming how much of the maintenance of X11 is actually done by Red Hat developers. All I can find is this. Red Hat is a for-profit company. If they wouldn’t need to have so many dedicated developers on X11, they wouldn’t, as developers are expensive. Their profit would improve a lot if the community did more work. If X11 could survive without them “just fine”, why would they spend so much on maintaining it. We will only see how much they really did once they stop working on it. If X11 stays well maintained by the community, I’ll be the first guy to celebrate, but being realistic, I don’t think that will happen. Not because “Red Hat is such an angel, and only they can maintain it”, but because X11 is a dumpster fire. It’s the part of Linux that is giving me the single biggest headache. It’s a Huge pile of botches on top of an archaic model that completely disregards how modern computer setups work. Not only that, but it does not follow the Unix philosophy doing just about everything making it almost impossible to modularly swap out certain components requiring you to go all or nothing and forcing you to build your entire GUI model around their idiotic ideal.

GNU

That’s rich! Linux was not affiliated with GNU back at its inception, and RISC-V has nothing to do with GNU. GNU is a cancer, hijacking the gracious work of open source developers, claiming it as their own and using it to fight a philosophical crusade against companies. Who do they think they are? Claiming to be an advocate of “freedom” yet feeling like they get to tell people what is free and what is not?

Clang is under the Apache license. Apple did not need to upstream or even publish the changes they made to it. Even still, they upstreamed some changes, donating their work to the open source community.
T-Head was under no obligation to share their RISC-V core design. Even still, they published the Verilog source under the Apache license.
Companies have shown time and time again, that they are willing to release their code, for which they have paid their developers a fortune to write under Open Source licenses. They cannot take away your X11. If the open source community is truly capable and willing to maintain X11, then red hat dropping support for it will have no influence, and you have no reason to make them the villain. If that is not the case, then you need to accept, that they carried the project on their backs, and gave their work away for free, and you have no reason to make them the villain. Either way, they are not the villain.

VF2

This is neither what I said, nor the reality. You do not need to use official builds by Starfive. There are quite a few community maintained forks that have a lot more functionality than the official builds. I run a custom Gentoo on my VF2, all the Starfive components of my system have extensive community patches applied. And Starfive working on upstreaming support for their hardware is bad how exactly? If something is not working, why don’t you fix it yourself instead of making Starfive the villain for not donating a free fix to your crusade?

I feel like this discussion is going way off-topic. I have left out a lot of my comments on GNU, as I don’t really think this is the right place to discuss Open Source politics. I feel like all I can contribute to this thread was left in a post above, and I apologize for attacking the people here and begging the question without really having an answer to give the people who might’ve just searched an answer.

3 Likes
19

X is dead, get over it. No point implementing support for legacy cruft (compared to “maintaining” an existing dead implementation).

2 Likes
20

One fundamental thing to keep in mind about X11 is that the core design of the X11 protocol was based around the most advanced hardware available in 1987 (~36 years ago). When it comes to displays, graphics acceleration technology and networking there has been an incredible number of changes since then. You have to give credit to the people who designed X11 to allow it to function as well as it has for so long, but it is long overdue a core redesign, and as much as I would wish for an X12 protocol, ideally released on or before the 40th anniversary of X11, it looks Weyland which requires logind (usually provided by systemd but can be provided by a shim like elogind, to avoid needing systemd), will be the future. Wayland and Compositors are available on FreeBSD already and is soon to land in OpenBSD. It is here until something better (or worse) replaces it.

2 Likes