Hacker News Reader

Comment by PalmPilotProMax | Hacker News Reader

So, to sum up:

First there was Nvidia Gamestream, which was proprietary and later deprecated by Nvidia themselves.

Then came Sunshine and Moonlight, open source server and client implementations of the protocol to allow low latency streaming for all platforms.

Along the way the Game on Whales project started, to allow multi seat streaming plus virtual displays so currently logged in user sessions aren't impacted. Basically this is a whole "host your own Stadia/Geforce Now" solution. But instead of running already setup applications as an existing user, you run these curated containers of Steam, Firefox, etc.

Then someone forked Sunshine and Moonlight into Apollo and Artemis, to make streaming from a virtual display turnkey and OOTB easy. Sadly the virtual display feature of Apollo is in practice Windows only.

And then there's this, Moonshine, in a way doing what Apollo brings to the table but for Linux servers.

Oh, and along the way someone forked Apollo into Vibepollo, an almost fully vibecoded "enhancement". Personally not touching that.

Did I miss anything?

View on HN

The steam link actual.

While the hardware is no longer being sold the software sort of lives on. In an android app among other things.

Tangentially it almost feels like steam could make a stadia like thing work. They have most of the moving parts including crucially, the bit google didn't, an established base of customers. Tomorrow steam could introduce a "play your games anywhere on any hardware" feature and no one would be surprised.

Wouldnt they need racks of GPUs to do that at scale?

Have you heard about the cloud?

The cloud is just someone else's computer. Do you think there's spare GPU compute available during the AI craze?

What if steam users could offer their computers up for credit on steam? Not sure if most consumer internet connections would be up for it, but that would be an interesting model.

I suspect the latency would absolutely kill, plus what happens to your killer in-game progress when the owner of the computer decides to play their own game or shut down the machine? So you'd need some sort of handoff, which seems technical very, very difficult.

I think it’s death by a thousand cuts. I think internet connections are the least of the issue. Plenty of people have gaming PCs wired to 1Gb/s fiber. It may even be better because users can be close together.

Where I see bigger issues is: 1. Economics. GeForce Now for a 5080 is 20ct/h. At retail prices you may get close in electricity.

2. Others want to game when I want to game. What good is offering my PC up at 5am? Who would want to use the service if you can’t reliably get access when you want it. Not to speak of the issue of getting kicked out when the owner wants their PC back.

3. The whole: do you trust someone else’s computer/do you trust others on your computer. Users don’t want their login details stolen or get banned for using a cheaters PC and providers don’t want you to cheat or do other nasty stuff. You can’t really do VMs because consumer GPUs don’t really support virtualization so dual booting

4. You need to dedicate a ton of disk space. Because of 3 you can’t really share and now you have to deal with some system to have the relevant games available.

I agree with all those issues.

Just for discussion though: Economics don’t necessarily dominate these markets when providers aren’t informed - Uber drivers w/ depreciation for example. If your GPU is a sunk cost, and your electricity is cheap enough, it might be attractive (think consumers mining crypto back when).

A floating market rate or “surge pricing” could come into play. Outside of the top 10% of gamers by playtime, there’s plenty of idle compute during prime hours.

A secure environment would definitely be essential. SteamOS could be a great tool for that: dual boot SteamOS and make SteamBucks while you’re not using your computer.

You could only offer your machine for the titles you have downloaded. This again could be addressed by a market: popular titles could get more utilization, but obscure titles could command a premium.

No guarantee any of this would actually work, but it’d be an interesting model to explore if I were at Steam. They have hardware survey data, user activity, etc; to check viability.

I think there's a lot of spare GPU compute. I don't think it'd be made available for this, or if it was, I believe it'd most likely be from a company that won't survive the AI bust. But I think there's a lot of spare compute sitting idle out there.

If there weren't, I think token prices would be higher than they are, downtime would be higher than it is, and there would be far fewer free tiers or more restrictions on free tiers.

There's a pretty fair staging argument of AI for work, Gaming for play, that would keep north american data centers with north american customers running 2/3 of the day.

There's also an alternative which is that token prices are already high, downtimes do exist (semi frequent on Claude) or are managed by serving degraded versions/quants (speculation that has never really be proven afaik) or reducing thinking time ("juice" values from open AI), and that free tiers are not used much at all or are a loss leader.

I don’t think there is a ton of spare capacity. But even if there was the machines are designed for AI not gaming, e.g. low CPU clocks, the GPUs don’t have RT cores or video encoders, etc. Stadia had custom hardware, GeForce Now uses workstation GPUs, xCloud has custom servers. Gaming also isn’t a good use for spare capacity. Demand is very spiky and people don’t have a lot of time flexibility.

Seems like there is considering that Meta is currently selling their excess AI capacity.

https://www.reuters.com/business/meta-sell-excess-ai-computi...

They're selling it for AI. Unless retro-nostalgia explodes and gamers are overwhelmed by a desire to pay for tokens (more than ai users), that doesn't change anything.

> While the hardware is no longer being sold the software sort of lives on. In an android app among other things.

It is built-in on every Steam. You can basically press two buttons and stream from your desktop to laptop if both have Steam with same accounts logged in.

> Tangentially it almost feels like steam could make a stadia like thing work. They have most of the moving parts including crucially, the bit google didn't, an established base of customers. Tomorrow steam could introduce a "play your games anywhere on any hardware" feature and no one would be surprised.

Already possible. While not so optimized as Games-on-whales.

The thing I found limiting when I tried to run Steam remotely from my lounge room was the need to actually interact with the host PC at certain points, not just during the initial setup. I found this is the same on a lot of similar platforms, and it’s made worse when RDP clobbers the graphics settings.

A truly headless solution with a web ui would get me really interested.

Valve has truly headless runner (gamescope) for Linux, e.g. you can start it manually from command-line over ssh

  gamescope \
    --backend headless \
    -W 1920 -H 1080 \
    -w 3840 -h 2160 \
    -r 60 \
    --cursor-scale-height 1080 \
    --force-grab-cursor \
    -- steam -gamepadui -pipewire-dmabuf

You still need the initial setup for pairing.

But it has many issues compared to Games on Whales, for example, which is much more efficient and scales cursor better.

https://github.com/ValveSoftware/gamescope

Furthermore, it will log in and unlock your host pc. If you're across town at a coffee shop and your 4 year old kids are near the host pc and see it unlock and start playing a game, you better believe they're gonna wreck your save file.

Not to mention greater security concerns.

Yep, it's really only an option for a dedicated gaming PC (or a VM with the the GPU passed into it).

That's not the same thing at all. Stadia was never about streaming from your own dedicated pc, it was about cloud computing. Something Valve does not have. In order for Steam Link to become Stadia, Valve requires the same data center capacity to serve games.

Steam Link has saved my ass more times than I can count.

VPN having issues, can't connect to any of my stack? Atleast I have a desktop on my LAN I can remote into...from my phone...

Basically anytime I have network issues while I'm away, reliable Steam Link connects me back to my infra ~97% of the time.

Wait. Go on? Is this with or without the device?

For the record, I always wondered why I couldn't do something like this with whatever magic is happening under the hood with Syncthing.

I've been doing networking related things for over 20 years, and it's still wild what doesn't click for me...

Steam Link creates an RDP session (I assume their own protocol but for lack of a better term for it) that lets me control my Desktop from my phone.

Once I'm remote on the desktop, I can access my home-lab locally instead of via VPN. Before I moved to Tailscale, my self-hosted Wireguard setup was pretty shaky, and was running on an old Pi that had seen better days.

It's basically just using Steam to jump-box to my LAN, and from there you can do whatever you desktop can do on the LAN. (Steam Link has some oddities around UAC prompts, usually requiring input from the host and refusing from the client.)

You could probably use anything....Parsec, Moonlight, RDP, etc. But Steam is always running on my Desktop, and the phone-app helps a lot. I've used Parsec as well, but sometimes if using a commercial VPN Parsec is unable to connect. Steam tends to still create a session in these scenarios, but YMMV...I'm sure there's some config issues in my setup

Aha, I'm getting back into this sort of thing, and I've always been a fan of Tinc for things like this, but it is showing its age.

Do you leave your gaming rig switched on at home at all times?

They also have datacenters in most major cities, so latency would be minimal. They're one of the only services that allows me to actually saturate my home internet bandwidth. But hosting game data and actually running the games is a lot different.

Unfortunate it is still essentially impossible to share consumer GPUs with multiple virtualized hosts, which limits the utility. My dream is to build one beefy server and play old games like COD via connected thin clients

intels gpu supports vgpu with no extra licence

It's possible to run vgpu with modded drivers up to, I think, RTX 2xxx series.

Thanks to the comment above I discovered Games on Whales and just read up on the possibilities. Isn’t that exactly your usecase? It seems you can stream to multiple underpowered systems provided you have network, RAM, GPU and enough cores. It even streams to Apple TV, Chromecast etc.

No, I mean to have multiple concurrent sessions, e.g. to have friends over on laptops playing games from a centralized device

That's exactly what Game on Whales offers

It's multi seat.

Yes, because it uses containers. It’s not using VMs which would probably be required e.g. for many windows games

Not sure if LXCs would help in that regard as I haven't used it for gaming but you can share a GPU between LXC guests and the host. I have an unprivileged LXC running in Proxmox that uses the GPU for transcoding video.

AFAIK there is no way to share a GPU between a host and multiple VMs without commercial hardware

Thats the thing, LXC isn't a VM, its a container, so they all share the same kernel. I think its worth a shot, so I may test this.

FWIW Incus is really good here. I don't quite have it fully working for headless gaming containers sharing a single GPU, but I'm very confident it will work at this point.

OK, so to clarify, it's Sunshine (game streaming server) but headless for linux.

So if I used sunshine, and streamed a game, the display of the box running sunshine would be showing the game? and this moonshine allows us to do away with that, to be able to run headless, or to run multiple games at once or something?

Is that correct?

But Sunshine _is_ headless, I would know if it weren't, I'm using it as such (albeit with a dummy plug, but that's just a hardware limitation).

The selling point of Moonshine (and some of the Sunshine forks/descendants) is that it's headless without the need of a physical dummy plug.

Nvidia GPUs refuse to accelerate 3D w/o an output connected; if I had a Radeon, I wouldn't need the plug, it's just an Nvidia thing I believe.

Afaict, Moonshine relies on a separate graphical session running on your monitor and that's how it gets around that limitation of Nvidia GPUs. If I had a monitor plugged in to my box, Sunshine would work just as well.

It’s possible to make a fake display to avoid that; both Linux and Windows have this ability. No dummy plug is required (though it is simpler than software solutions).

I don't use the dummy plug for display, it's only that consumer nvidia gpus won't have working 3D acceleration if there is nothing physically connected to the card. My dummy plug doesn't support the resolutions I run games at anyway.

EDIT: Ok, correction, I bothered to go unplug the dummy plug and it still works. So I was wrong, Sunshine does not need a dummy plug at all, not even on Nvidia. Now it remains a mystery why I have one in the first place if it's useless after all…

Sunshine was significantly easier for me to get working using a dummy plug. I'm currently exploring other options but that's definitely the happy path.

ok, weird. Yeah, I saw this and thought "what's new, isn't that what moonlight/sunshine already is".

Is it just a new implementation? is that it?

[EDIT] all answered here: https://news.ycombinator.com/item?id=48976789

You're technically right, and I appreciate your comment, but it's funny to me to be like 'It is headless! It just needs a fake head!'

Now you try working with a fake in your head's place! :þ But yeah, real head (a monitor; doesn't have to be on) also works just as well, and even no head at all works, just won't accelerate 3D. Arbitrary resolutions also work. My only complaint with sunshine is that the webui authentication is really broken. Fortunately I don't need to go there much.

I setup a fake display on my fedora instance (inside proxmox) at boot time (boot/kernel args) that mimics the specs of my real display. Sunshine sees it as a real display so I can unplug everything from the box except power and lan cable. No hdmi dummy dongle or virtual display driver needed. Using linux as a gaming server works fine!

I've been doing almost the same exact thing for a while, except on Arch and with a different EDID than my actual display drive my handheld that I stream to is 16:10 rather than 16:9 like my actual monitor. After a couple of months something changed (maybe a kernel update, although I never fully tracked it down) where the kernel no longer respects the parameter about which display to use for the initial LUKS decrypt and tries to use the fake display no matter that, which prevents me from unlocking it until after a timeout of a few minutes. I've made this somewhat less annoying by setting up unlocking via ssh on a local IP so that I can just initiate a reboot and then do other things around the house without having to wait around next to it until it times out and falls back to the real display, but it would be really nice to not need to wait in the times when I don't have any need to do anything else for a few minutes. Did you ever happen to run into this timeout issue, and if so, do you know how to fix it?

If you have a tpm (or virtual tpm), you can enroll your passphrase for auto unlock. When sealing the tpm slot, you can decide how much should be sealed: kernel version, drivers, firmware updates, hardware changes etc.. meaning a kernel update can break the seal and ask for passphrase if you included kernel version in the seal. You can set it so the kernel/driver updates won't break your tpm seals. It's usually just one command to reseal the tpm values and the auto-unlock should work again, but it wastes time like you experienced. I use systemd-cryptenroll and a couple boot args.

If you have claude code installed (or ssh access for it), you can ask it to investigate your entire boot chain. It can find and fix it in 5 minutes.

I'm not sure if I'm misunderstanding what you're suggesting, but my issue isn't that I want unlocking to be even more streamlined (since I'm fine with what I have and don't want fully auto unlock); it's just that I don't want the timeout that somehow started happening before it's even letting me manually unlock.

And yeah, I did try out various LLMs (including Claude) when it first started happening, but none of them were able to point me towards anything definitive about the cause, let alone the fix. In general I think it might be worth giving someone asking a question about something like this the benefit of the doubt about whether they've tried it, especially if you have outsized assumptions about how likely they are to solve problems. My experience is that when someone mentions having an issue, telling them "just ask Claude/Codex/etc." doesn't tend to end up being a very valuable contribution to discussions; people who are open to using LLMs for things will almost definitely have already done so, and people who haven't already are a lot more likely no to have done so for other reasons than the idea not having occurred to them. It's pretty hard to go more than even a few minutes into any sort of discussion about tech nowadays without someone bringing them up, so I think it's a pretty safe assumption that most people are well aware of them and considered them for whatever problem they're having.

To be clear, I'm not trying to berate you or anything but genuinely give feedback that might be helpful for you when trying to give advice on things like this in the future. Trying to give advice in a way that people will be receptive to can be difficult, especially when they're strangers on the internet, but I also think it's a a valuable way to spend time compared to a lot of other ways to spend time online, so I think it's worthwhile to invest time and effort into figuring out the best ways of doing it.

Vibepollo is damn great, best solution on Windows by far. Totally painless and damn smooth, don't diss it because of its vibecoding.

Agreed. It's awesome.

I would love "normal linux user remote desktop session suitable for gaming using the host's GPU".

Nothing seems to do exactly that, though this (moonshine) seems like it might be the closest yet.

Can it run a desktop environment as the "Application" that it's streaming?

I've been running Sunshine on my Linux desktop for a while now and connecting to it with a SteamOS handheld and my laptop for mostly gaming for a while now. From what I can tell, the biggest difference with Sunshine and Moonshine when running on Linux is that Sunshine doesn't use a separate desktop instance per stream; by default, the streams will be mirrored to the clients, although there's a way to work around that somewhat by setting up a fake display output device with the boot parameters and a custom EDID with the specs you want (there's at least once repo where someone has a large set of them dumped from thousands of different monitor brands/models) and then setting up Sunshine to output when mirroring, which effectively means that only the remote client will actually have anything displayed.

It took a fair bit of iteration to get my setup to the point where it mostly works the way I like, but it's been pretty stable for months now. I game exclusively on my handheld streamed from my desktop, and on my local network the input lag is pretty consistently low enough that I don't notice any difference from when I had tried to play things locally on SteamOS, only I can easily max the graphics settings on most games I play (and on the rare ones that do struggle, there's often not a ton of downside to lowering the resolution a bit due to the handheld only being 1200p).

I sometimes ssh into my box with a port forward and starting wayvnc, then connect via tigervnc, I only played a few older/simpler games like this on Steam for a short test, but seemed to work. The server was Arch with Sway and client Fedora with Sway.

It seems like Sunshine and Moonlight are now active again, and have a release more recent then Apollo and Artemis.

On windows, is it possible to use the server at the same time as someone is remote gaming? Most virtual display solutions I found just turn the server display off.

Most of these solutions work by directly forwarding the GPU-rendered image. On Linux, Moonshine uses Wayland for headless rendering, but that approach doesn't work on Windows.

Windows requires an active user session with a connected display(physical or virtual) in order to build a desktop to render something. RDP can create a backgorund session, but on non-Server editions of Windows, you can't have a background RDP session while you are using it. It only allows one active session unless you patched the RDP service.

I don't think that's entirely true with the recent Apollo and Vibepollo developments.

> And then there's this, Moonshine, in a way doing what Apollo brings to the table but for Linux servers.

What was the motivation? Games on Whales seems to work perfectly in my experience.

I've used Moonlight but didn't like having to unlock my Linux home PC to play. If I want to play remotely using my (idle) PC, what's a better solution? Games on Whales (which I'd never heard of before now, and I'm not sure if it works on Linux), or Moonshine?

I have server running in basement all the time. I start Games on Whales over ssh and then launch Moonlight on macOS client. You can do it completely remotely. Only pre-condition is that computer must be turned on. I tried Sunshine as well but it was too much pain and not so effortless.

Nice, thanks! You can't do that with Moonlight in Linux, I always needed to log in to my session first in person.

Games on Whales runs containers, it doesn’t stream your desktop or your steam installation for example. So use case is for those that use the PC directly but also want it available via moonlight without a display connected.

The sum up is pretty good, I didn't know half of the history

About Apollo, someone started a fork of Sunshine to be the equivalent of Apollo (with virtual displays) for linux, it's called Polaris https://github.com/papi-ux/polaris (nb, it's not the apache polaris). It's very Alpha right now and the UI feels wonky but it's promising.

Thanks for the summary!

Does anyone happen to know how to make a Windows machine, the extended display of a Linux machine? Both machines are on the same network.

I'm looking to use a Windows laptop as a mere HDMI monitor for the Linux laptop. I don't need gaming/streaming requirements. Like, can anyone list our options in 2026 as the parent comment did?

I assume one can use this Moonshine on the server (Linux laptop) and run Moonlight on the client (windows laptop)?

My unhelpful answer is: It sounds like what you're describing is logical. Can you go ahead and set that up and report back?

Yes, I do this. Got used to my work laptop being my second monitor under my main but at home use a desktop. I do it with a Mac laptop but moonlight on the laptop in my experience is not the finicky part of the setup. I do it with basic sunshine connecting to a headless display, though. I think moonshine sets up the virtual monitor for you? Never used it.

I don't have the time to provide such a detailed answer, but there are a few options.

At least on Sway (and I imagine other wlroots-based compositors) it's very easy to do: create a virtual output and start a VNC server attached to it. Start your vnc client on windows. A comment I posted 5 years ago: https://news.ycombinator.com/item?id=25892128

The comment linked above is on the topic of Deskreen, which is a tool that does that.

Funilly, I also commented something similar there: https://news.ycombinator.com/item?id=34010970 the article explains how to do it on a lower level (pretending a display is connected to the GPU), works under X11 too.

A few days ago, an update locked me out of my server. I used a cheap USB HDMI capture card to use my laptop as a screen, which worked very well: that could be another option.

Then I guess that there is this option, which is new to me. It feels like this could be a toogle in sunshine, somehow :)

Then there's Steam Link which actually works, even with non-Steam games and full desktop. My computer screen is 21:9, TV is 16:9 and that is too much to handle for the other streaming solutions.

Yes, nothing will be as fast as the Nvidia implementation.

Those implementations don't have access to the low level encoding Nvidia has.

There is a rumour, that they did this because of GeForce now, because there never will an open source implementation/ service competition for GeForce now.

And there's also parsec.app but only for windows hosts through.

Why do we need these? I have sunshine streaming headless to my laptop without issues

Woof, so which one is the “good” one?

Sunshine is the stable, maintained one. That should work fine for most users.

Apollo solves many issues that Sunshine has on Windows. If your host is Windows, a correctly configured Apollo (i.e. where virtual monitor is enabled) will be a way less frustrating experience.

And Vibepollo solves many issues Apollo has

> Did I miss anything?

Steam has had good built-in streaming for a long time.

Reply on news.ycombinator.com