Steam Link Anywhere Will Let You Stream Your PC's Games On the Go (pcgamer.com)
Valve is expanding its Steam Link game-streaming feature in a big way with Steam Link Anywhere, a new service that will allow you to stream your Steam games from your computer to anywhere in the world through Steam Link hardware or the Steam Link app. From a report: Steam Link Anywhere is an extension of Steam Link that will enable users to connect to their PCs and play games from anywhere (thus the name), rather than being limited to a local network. It's compatible with both the Steam Link hardware and app, and will be rolled out automatically (and freely) to everyone who owns the hardware with beta firmware installed, the Android app beta, or the Raspberry Pi app. You'll also need to be enrolled in the Steam client beta, and have the latest version installed. Assuming you've got all that covered, you'll see an "Other Computer" option on the screen when searching for computers to connect to via Steam Link. Select that, follow the instructions, and you'll be set. Valve didn't provide specific network requirements but said you'll need "a high upload speed from your computer and strong network connection to your Steam Link device" in order to use it.
This is actually a decent form of shared demos - in the sense that you get access to the full experience, but sort of a artifact-and-lag-laden version of the game that your friends don't have to pay for.
Not so great for strategy games and turn-based RPGs - but a decent additional option.
Kind of inherently precludes any multiplayer as-is though, since you'd still logically need multiple copies, and I don't think they'll let you map multiple controllers to exist on multiple PCs simultaneously - likely won't even let a single copy of any game run at the same time on two systems.
If it works like Steam Link though - you'll automatically see what the outside player is doing - since it's your video card they're rendering with to show the game. Add in automatic voice chat, and you can even guide the guest friend on parts of the game.
Could actually lead to some interesting interactive press-style demos too and the like, depending on what controls they allow.
Ryan Fenton
I agree. I can't see how it could possibly have ANY chance of working as advertised UNLESS the user managed to satisfy ALL of the following requirements:
* 5mbps uplink, absolute bare minimum.
* Expensive router. Most people have NO IDEA that if they have internet connectivity faster than 50mbps down and 5mbps up, their own router/access point is likely to be their network's single biggest chokepoint.
* Good (or better) quality gigabit switches, with proper wiring. The majority of cheap "gigabit" switches can't themselves sustain a true gigabit between two ports, let alone MULTIPLE gigabit connections among their ports. Even IF they can genuinely switch a gigabit from a port, their own back-end fabric probably maxes out at a gigabit too. The internet connectivity might be a tiny fraction of a gigabit per second, but the latency of this is likely to be so bad, even a millisecond of extra latency caused by using 100mbps or less could be the latency that makes it intolerable to use.
* Video card that can do hardware-accelerated h.265... which pretty much eliminates any laptop more than a year or two old, and any desktop video card more than 2-3 years old. ... and even THEN, it would probably still suck. Or at least, end up being an EXTREME disappointment, given the reality that very, very few people can sustain more than 10mbps up, regardless of how fast their nominal downlink speed might be (and because ISPs constrain uplink speed both to allow them to advertise bigger numbers for the download speed AND preserve the distinction between 'residential' and 'business' service by forcing anyone who wants faster upload speeds to pay enormously higher rates to get it).
Netflix can pull off high-quality HD with a 4-6mbps link budget because it does offline non-realtime multi-pass compression. Attempting to pull off a similar stunt in realtime is another matter entirely. It's ~80% of the reason why we didn't get to have 1080p60 in ATSC1.0... back in the 1990s (when ATSC1.0 was finalized), realtime compression of 1080p60 MPEG-2 within the ~19mbps link budget was flat-out IMPOSSIBLE... and for the most part, it still is (at least, as MPEG-2). In retrospect, the decision was kind of short-sighted... we never would have gotten to have things like live sports in 1080p60 using MPEG-2 (at least, not without a 15-60 second time-delay, using commercial breaks as an opportunity to play "catch-up" and re-fill a depleted video compression pipeline), but we most certainly COULD have had things like prime-time TV shows (and pretty much ANYTHING that can be compressed long in advance of airing) in 1080p60, even with mpeg-2
Nevertheless, I think it's absolutely delusional to expect low-latency 1080p60 live video from even a high-end gaming PC connected to fiber or premium cable/vdsl2 within a realistic, achievable present-day link budget. There are just too many things that can (and absolutely WILL) go wrong. It might be good enough for Candy Crush, but I just don't see it as being even REMOTELY viable for things like a multiplayer FPS.
Maybe... MAYBE... if Valve got networks like AT&T and Verizon to colocate high-powered virtual gaming rigs at tower sites (so literally the only constraint was the wireless data link between the phone and tower), it MIGHT not TOTALLY suck.
Yes / No. This isn't such a difficult requirement at all.
a) the venn diagram of people likely to use steam link for gaming on the go from their PC and those who have a modern PC overlap greatly. Any previous gen graphics card can do H.265 encode in hardware so if you have a GTX 1xxx or an RT 5xx or greater you're good to go.
b) people who give a crap about gaming are likely to have a cable / fibre connection anyway, and that doesn't change the fact that this still would work great on WiFi in your own house.
Netflix can pull off high-quality HD with a 4-6mbps link budget because it does offline non-realtime multi-pass compression. Attempting to pull off a similar stunt in realtime is another matter entirely. It's ~80% of the reason why we didn't get to have 1080p60 in ATSC1.0... back in the 1990s
Okay now I take big issue with this comment. Complaining about something in the 90s is silly when what we do now working with real time video compression is actually trivial. Pascal series cards come with a H.265 encoder that happy runs at several hundred frames per second. On my 3 year old graphics card I have no problem at all encoding in real time 1080p, and streaming it to my mobile phone. This is how software like Riftcat or ALVR works to give you access to SteamVR on any modern smartphone with a basic (even cardboard) VR headset. It worked just fine 3 years ago at realtime with lag low enough that gaming was possible.
So yeah the network is a barrier to overcome, but the video encode/decode is most definitely not.
So games with all of the fun turned off?
...except on iOS.
Incipiamus, fratres, servire Domino Deo, quia hucusque vix vel parum in nullo profecimus.
This is good news. It may or may not work for SuperUltraMegaTwitchBlast'Em'Up but maybe you fancy a relaxed game of Civ, or perhaps an MMORPG, or....well, you get the idea. Plenty of good uses cases, and hey - you're gaining a new facility anyway.
Steam Link hardware went on sale for around £3 the other month. They're not asking for much here. I'm happy with the announcement, whether I intend to make drastic use of it or not.
A gave a shot to the steam link at home. And essentially, you can forget it over wifi. So good luck getting it to work on the go; where network connectivity is typically terrible.
In the future, I could imagine that steam would be bundled on smart TVs that you find in hotel rooms and if they have decent internet and are wired to the network, that it could work.
A gave a shot to the steam link at home. And essentially, you can forget it over wifi.
I tried steam link at home, and it worked fine on WiFi. Maybe you've done something wrong. It's not going to work on g, but it works fine on n (it worked for me), and should work fine on a.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Maybe you should fix your horrible WiFi network. I use it just fine. No discernable lag as far as I can see on either of my networks, even the room with some questionable coverage is still playable.
I don't see this working out. There's no way that the latency would allow for a playable game under any conditions.
It probably will not work for the twitch-style games, but there's plenty of more casual games, turn-based, RPGs, simple platformers, retro-remakes, etc. that latency would be manageable. These are also the sort of games that do well in mobile markets already. If you check Steam's library you'll find many games that are also available on iOS/Android stores and customers have to choose one or the other platform. Now Steam is both platforms. At the very least it helps retain customers in the Steam sphere, who might move out to other platforms.
Also, game developers may prefer designing games for PC/Steam to be streamed to mobile, instead of designing parallel iOS and Android versions of their games.
The problem with realtime compression via ANY codec is that you lose your ability to use B and P frames unless you add at least 16.67ms of additional latency to delay transmission by at least one frame. After all, you can't predict the content of a frame that doesn't actually exist yet. Any time there's a radical scene change that can't be delta'ed from one or more earlier frames, you're going to end up with a mangled frame or ten, because ultimately, if your link rate is 6mbps, you're transmitting 60fps, the limit of what you'll tolerate is a delay of one frame, and the actual compression is instantaneous, that STILL limits you to 100k (6mbps divided by 60fps) to somehow convey the state of somewhere between 2 million (1920x1080) and 3 million (2160x1440) pixels.
Do the math. Unless you're trying to replicate the look and feel of 'Asteroids' on a vector graphics CRT, I don't care if your compression algorithm is downright miraculous... fsck'ing 320x200 256-color VGA needed 64 kilobytes. 100 kilobits works out to approximately 12 kilobytes. Let that sink in... you're trying to somehow send an I-frame with 2-3 million pixels using approximately 20% of the ram needed to display a 320x200 VGA bitmap.
Absent some magic compression algorithm that uses something like Markov chains to pre-share a humongous dictionary of graphic data specific to your game, there's no way on god's green earth you're going to somehow convey 2-3 million pixels' worth of data in single-bit black and white, let alone 32-bit alpha-blended color, within the constraints of a 6mbps link budget, 60fps, and zero-frame pre-buffering delay (because even if you manage to compress a frame instantaneously and begin transmission immediately, your absolute data limit before you fall a frame behind on the client end is the amount of data you can physically transmit within the time it takes to display a single frame of video).
Repeat after me: the ability of h.265 to massively compress high-quality video is inversely related to the number of frames' worth of latency you'll tolerate. And even a single 16.67ms 60fps frame is enough of a delay for your controls to feel perceptibly sloppy. Two frames @ 33ms, and you're dying for reasons that aren't even your own fault. Three frames @ 57ms lag, and you might as well be wading through wet concrete.
Don't believe me? There was a story right here on Slashdot a few days ago about a study that confirmed that in a multi-player networked game, players running at 120-144fps consistently beat players running at 60fps. Latency is EVERYTHING.
One of the hardest aspects of programming a fast-action responsive multi-player networked game is dealing with conflicts where one computer sees a legitimately-landed deathblow, while another sees that same player dodge the blow because he was able to react in time. That's one of the things that prevented realtime-networked peer-to-peer Soul Calibur-like fighting games for so long... games like that literally hinged on single-pixel hits or misses within the span of a single frame. Without a server to authoritatively serve as referee, a game like that would have been unplayable even on a 100mbps LAN, let alone over the internet. And even fast DOCSIS3, VDSL2, or fiber that passes through at least one peering point has too much latency for games like that to work unless you can come up with some excuse to factor hair-trigger single-pixel hit/miss logic out of the game (and it's why someone who played games like Soul Calibur 20 years ago would view hand to hand 'combat' in today's PvP networked games as a cruel, sad, sloppy joke).
So, yeah. 5-10 frames of pre-buffering latency might be tolerable for an interactive-fiction adventure game, and 1 or 2 frames might be semi-acceptable for a PvP game with sloppy high-level excuses for 'combat', but it's just not going to cut it for games with intense action and hair-trigger reflex action. And if you're going to settle for a game where latency is tolerable, you might as well just run it natively on the phone instead of screwing around trying to remotely run it on a PC and virtualize its display over the internet.