Akamai CEO: Playing games from the cloud? Seems too expensive to be viable right now Akamai CEO Tom Leighton – an expert in distributed computing with 50 patents under his belt – has thrown shade on Google’s plans to launch a cloud-based gaming platform. Leighton seemed skeptical such a subscription service would be cost-effective, and he ought to know: for two decades now, Akamai has operated a global network …
First of all: I don't game, so the blind is speculating on colours here.
But: I assume a decent gaming rig will cost over 1000,- (Euros, Dollars, Pounds doesn't really matter) - let's say 1200 to ease up the calculations. This means, that with a subscription cost of 50,- the hardware alone will cost two years worth of it, not taking into account the power it uses at your end (a.o.t. on Google's end). Most hardcore gamers I know will consider a 2 year old rig hopelessly outdated at least concerning the GPU. Add to this the cost (in expertise and time) of running Windows 10 and outsorcing all that to bigger players (pun intended) and only owning what is basically a smart-TV-sans-TV might actually be a good value proposition for the less tech-savvy.
I'm a gamer and I have a quite powerful DIY rig.
The economics has changed a bit for the last 5 years, due to the stalling of HW evolution.
I think a modern medium class rig can last 5-6 years for gaming so that would 1500 USD depreciated over 6 years, I think. Of course I'm talking outside of the crypto-nvidia price craze of now, more of the average we've seen over the previous years.
Do the numbers compute vs. Cloud gaming, maybe not, indeed.
There are two gaming markets:
1) Casual gamers (like me). Have plenty of free games that play on mid-spec PCs and often just a phone.
2) Hard core gamers who will spend ££££s, but are concerned about every millisecond of lag.
I don't see how streaming games appeals to either of these unless it's free or nearly free.
1000 for a gaming rig? I've built gaming systems for a lot less than that, though it's really down to personal choice which is why I don't think this is aiming at PC gamers (at the moment). The gamers I know (myself included) like owning their own rigs, they like tweeking the hardware and software. Most start with home PCs that they upgrade and spread the cost over years. They might buy a new system at some point, but they'll probably recycle parts from their current system. Most people not into PCs and want to get into gaming buy a console.
Streaming works for video and music because you can buffer, you can't do this with a game
Re "They might buy a new system at some point, but they'll probably recycle parts from their current system"
True.
I am a PC Gamer, but not a hardcore one that spends thousands a year on a "rig". In fact, I don't even call it a "rig".
I usually go for decent mid-range hardware, as I find very little software gets anywhere near fully utilising high end hardware and by the time it does, the mid range hardware does things almost as well.
But, the last time I bought a full PC was probably when I started my degree in 1994. I've just been upgrading bits since then, and the PC I own currently bears no resemblance to that PC, so it *is* a different PC. Beyond a new GFX card, most of the PC I have at the moment is over 4 years old, and I need a fairly powerful PC that is separate to my main PC to do some stuff with Virtual Machines on, so I am looking at building a 2nd PC now, but that's the first time since 1994.
Latency is the big issue I see here. Every single mouse movement, key press etc. has to be sent via your ISP across the Internet to the streaming host, the rendered graphics adapt accordingly, that video stream then has to be compressed and squirted back over the Internet to your PC and displayed. Even with good kit, I foresee a latency of at least 50 to 100ms here and this is going to put off a lot of hardcore gamers for whom low latency is key.
It'll also require a consistent and fast connection as well. Current video streaming often includes a small buffer so that a brief stutter or delay in your internet connection doesn't cause the video to skip. Such buffering won't be possible for a streamed game - it'll have to be properly real time.
In short, I'm skeptical. Very skeptical.
We can use Steam in home streaming as a yard-stick here.
That is, one PC playing the game, and another sat in front of the gamer.
I have a Steam-link, and while it's indistinguishable from a full PC, lets use a second PC as the client machine, so we can benefit from having the Steam-link's limits removed (e.g. it only has 100mbs wired networking).
In this setup, we have two PCs, both connected to the internet network via wifi, but connected together with a 1GBs Ethernet connection, which can be forced into use for the gaming stream.
This is essentially the best possible setup for streaming a game across a network - if a single cable can be considered a network.
I haven't done any particular in-depth tests with this, aiming only to satisfy my desire to play casual games from the sofa, older games were chosen to reduce the workload so as to not bottleneck the encoding and streaming:
Picture quality - pretty good overall. At times, you can notice artefacting, which appears similar to turning the sharpness up too high on a TV, or jpg compression on images.
Delivery quality - Over a direct wired gigabit connection, the stream is rock-solid, with maybe a little hiccup maybe once an hour.
My current setup, with a 100mbs steam-link connected to PC through two switches is just as solid.
Over Wifi, it's a different story - even on 5Ghz uncontested link, hiccups occur, at a rate of 2-3 a minute, lasting a second or two each time, which is't terrible from a performance state, but is immersion breaking. Reducing the image quality reduces the frequency and duration of these hiccups, but they still occur and you have a much more noticeably compressed image.
Delivery latency - One of the games I tested was Mirror's Edge (2008) Constant latency is distinctly noticeable, even with the direct wired connection. It's perfectly possible to play the game, but you have to anticipate all actions. Most notable is when performing a roll when landing - playing locally, you pull the trigger just before hitting the ground - playing remotely, you have to pull the trigger prior to landing - when about 2m in the air. The same latency is noticeable when engaging in combat, making shooting harder (smooth movements are required) and disarms of AR armed enemies pretty much requires slow-mo due to the short period given to perform the move. I have a fairly sensitive NFS Shift configuration set-up with almost no dead-zone on my wheel. The latency made racing effectively impossible, being almost impossible NOT to over-correct for even minor deviations from the racing line. It was like trying to drive in one of those drunk-driving simulators.
From memory, the reported latency was reported by steam-link as about 25-30ms but I'd take that with a grain of salt. I'm pretty sure there are additional sources of latency that are not being accounted for in this value, and this is a local connection. Over the web, I would expect latency to be twice that.
For "casual" games, it's great - and I'm using casual in terms of "not requiring reflexes" here - games I can relax when playing. Something like XCOM-2 works great, but local streaming on anything requiring fast reactions can be anywhere from having annoying latency to unplayable.
Lets just say that I'm not optimistic about web game-streaming providing an experience that suits anything other than "casual" gameplay.
That sort of game can typically survive being run on lower powered (aka cheaper) hardware locally, albeit at "cinematic" frame rates without causing issues to the gamer. This is where I think the competition will be, and the cost equation is a lot different between a £300 machine and a £1200+ machine when compared to the costs of an online gaming subscription.
I'm not sure the lack of a physical game box etc will be much of a barrier. People are getting very used to not owning a physical copy of copyright material. Look at music downloads, film subscriptions etc. Even Steam for a more relevant example.
I guess it'll work more like Amazon's Prime model. You buy in to the system (initial outlay) and you'll get a certain amount of inclusive content with that. Then you can pay extra to rent or 'buy' more content. You'll never actually own it.
As an owner of an XBox I know that even owning physical devices and copies of games doesn't mean instant access. Regularly the games will need significant updates before they will play. Younger gamers will probably be delighted at having no delays like that. Parents may like not having to accomodate large boxes of hardware in the house.
Bandwith and latency might be much more of an issue the same as it is now with multiplayer games. Single player games this won't matter too much.
But I can see this all happening quite quickly. The next generation of consoles might be the last.
The problem with Cloud Gaming is that it will always be a worse experience than a local piece of hardware, unlike TV and Film where as long as a consistent framerate and audio sync is maintained, it is fine; Cloud Gaming REQUIRES an extremely low latency. If the round trip is 50ms, that is 1/20th of a second delay between performing actions and seeing actions occur on screen which is a jarring experience. More so there are other things also causing delay such as screen delay too.
More so, in PC gaming, there is multiple different set-ups that players can use and Cloud Gaming will not cater to them, such as ultra-wide gaming, VR or anything above 4K/60Hz. Most medium spec PCs can do 4K @ 60hz with the same type of graphics that these cloud gaming services provide because they also won't do things like Anti-aliasing, G-sync, triple frame buffering, etc. It'll at best be around the same quality as a PS4 or X-Box one and I wouldn't expect any improves on that for years as to run these services will require significant amounts of hardware and to upgrade will requires significant investments on the suppliers side. As it goes, I doubt Google is going to do a better job than either nVidia, Sony or Microsoft in this field and we already have offerings out from nVidia and Sony...
One other thing to think about - Bandwidth and/or compression. If we assume that the 25Mbs that Netflix says you need for their pre-compressed 4K is accurate then there would be addition latency for compression or you'd need a much bigger pipe.
Can someone in the field confirm my hunch?
Why would you need a bigger pipe? The only difference in the connection will be in the inclusion of user input, even if that included Audio, that isn't going to be a significant increase. The Video stream is still being sent one way. Compression itself is processor bound and would need to happen on the server side.
The issue incurred would be the delay for compression, as it does require CPU to be processed, it would also need to be decompressed on the client side. Now if you were to drop compression entirely, you would need a bigger pipe but you could potentially reduce the latency by doing this.
That's what he means. Compression and decompression necessarily adds latency, especially at higher rates, simply because the algorithms involved aren't optimized for realtime operation. Lots of multiplayer games are timing-sensitive (so-called "twitch" games where even a frame of lag means kill or be killed--a real buzzkill for battle royale-type games where you only get one chance per game). A lag of over 16.7ms means at least one frame of lag assuming 60fps. Granted, this is true of all gaming, especially those spanning long distances where the speed of electricity puts a physical floor on the lag. Games have had to compensate for uneven lag. Perhaps Google and Sony found a way to get right what OnLive and the like got wrong?
Technically the solution works and well. Obviously you need a good low latency connection to a local data-centre however with that in place, you really can deliver a decent 60fps low latency experience which is good enough for the vast majority of games out there. Certainly anything on console is easily playable.
So the question is not whether it technically works. The question is whether it's economically viable. Are there enough people with good enough connectivity to take advantage of the service?
People aren't missing the point, people are questioning the Demand, nothing is economical if there is zero demand in the first place. If the demand is zero because the experience is inferior than that is actually very much part of the point. How do you sell something that people aren't necessarily interested in?
Best you can do is pull an Apple and try to convenience people that something inferior is in fact superior but there is nothing superior about this offering over having the games locally installed on a PC or Console. If you want something that works on the go, the Nintendo Switch is already superior as it isn't reliant on mobile signals and mobile or wi-fi bandwidth.
The $10 a month subscription is for the "pro" package, which gives 5.1 surround sound, 4k video streaming and some other stuff (game discounts and I think free games while paying the subscription, but the only free game at launch will be Destiny 2).
Google has said that in 2020 there will be a free base package that just does stereo sound and 1080p video streaming, where you just pay for games. You also won't need to pay for additional hardware if you already have a controller of some kind and a chrome browser.
The $130 founders edition package includes a Chromecast ultra, controller (which can also be used as a HID wired controller for other devices, from what I've read), and 3 months subscription, plus an extra 3 months to give to someone else (possibly not useful until the base package is available).
Assuming what they've said is true, and the streaming is reliable with very low latency, the prices seem reasonable at the moment.
* If I already own a game, will I have to pay for it again to play it on Stadia
ie would I be able to play games from my Steam library?
* What will the selection be? AAA only, or indie games too?
* What about save games? Will I be able to upload my existing save from my local PC, and download again after?
* Will multiplayer servers be limited to people only in the same datacenter?
Your ping rate affects others, not just you - that extra latency is not fair on people running locally, so is likely to lead to segregation
* Mods - will you be able to upload mods?
* What forms of input will be supported? If I have a HOTAS controller for example for use with Elite and such, or a steering wheel, will I be able to run the config software for it? Will DirectInput controllers be supported, or only XInput? What about things like eye trackers?
* Will you be limited to only the voice comms solution provided with the game (If any)? or will you be able to use Discord etc?
* Once a game is made available, will it be available for ever, or will even single player games be pulled at some point
I suspect the answer to most, if not all of these questions will not be the one I want
Personally, I think something like Parsec cloud (Where you rent the whole PC, not just access to the game window) will be much better for the consumer
They didn't abandon the idea, they (the publishers) f-cked it right up the corn hole by demanding that every user must have separate hardware and licenses. This is not a model that will work under any fiscal plan. It's what killed the idea and made it not worth it to the customer. It's like if Netflix demanded that you buy every movie you watch on their service, and for every customer they need a separate DVD player. It's not necessary, and quite frankly mind blowingly stupid.
Lets be clear. Game publishers - most of them not just EA - have always loathed their customer base. Why would they support a mechanism that will see more customers spending good money on a subscription bases for their games? I mean, they'd make more money, sure, but customers would be happier and more plentiful, and they can't be having that now...
Games in the cloud? what happens when your internet is down? online games lag enough as it is, and they expect this to be practical?
also with the cloud you don't own your games... I hate the steam DRM for the same reason, and mobile games that need the internet to run really piss me off...
Seems like a paid alpha for a product that will be viable in 10-15 years time when ISPs (hopefully) improve their edge networks. I'm sure it will work soundly for a blessed few - whereas the rest of us on congested, inconsistently performant networks will just experience the same thing as most did with OnLive - noticably laggy, jittery gameplay. I suspect this generation's services will all fail, publicly, but provide Google, MS et al with the data they need to re-launch them when the majority of the world's infrastructure is capable of supporting streamed gaming fairly reliably.
But on a different note, just how was the conclusion that this is economically unviable reached? The traffic point is pretty moot - streaming game video is just the same as streaming Netflix, and that certainly hasn't been a problem. On the hardware front, as a cloud provider Google willl have plenty of spare compute capacity in its DCs, and Stadia will be niche enough to run lots of part-time casual gamer workloads with demand balancing while it works out how far to scale. Spare capacity is wasted capacity, so it may as well be put to some use...
So, first off, I suspect that Akamai doesn't have a network architecture that supports cloud gaming. Since delivering the full cloud gaming experience is nothing like delivering streaming video files. From a CDN perspective, with cloud gaming, caches are useless and the network cannot rely on the video distributor to have an already encoded asset ready for delivery to the target device. There is going to be a tighter integration of the rendering (encoding) with the delivery infrastructure, which means edge compute, and just a very different architecture than any CDN has today. So with that said it's no surprise that Tom would dismiss the "opportunity." He's actually correct, provided he is speaking only from the Akamai perspective.
Now, as for the commercial opportunity of cloud gaming... Consider that several quarters ago, in their earnings call, Reed Hastings stated that Netflix main threat was Fortnite. And I believe he's right. Yes, Fortnite isn't a true cloud gaming experience, but what has driven its massive popularity among serious and casual gamers alike, is the cross device ability to play anywhere you want. Standing inline at the store, and you can whip out your phone to play. Riding in the car, you can play. Need a break at work, you can play. Connected to your fav console on the big TV... etc... This is where, cloud gaming is going to open up massive opportunities to extend the entertainment of game play to a broad cross section of players who till now haven't even known perhaps that they would enjoy the experience. So, say what you want about previous attempts at cloud gaming, but what's different now is that we have ubiquitous power on the device side, with compute and cloud infrastructure and bandwidth (soon to be accelerated with 5G) that can enable technically an experience that is going to be stunningly close to dedicated hardware. Furthermore, compute and connectivity speeds and density are only going to grow/get faster...
Bandwidth isn't going to do squat about latency, which is mostly a function of distance (speed of electricity puts a floor on that) and computational complexity (meaning the latency between feeding a raw video stream into the compressor and sending the end result out to the end user in a bitrate that the Internet can actually handle, even if it's in the middle of nowhere where the only option out there is a wet noodle).
The TL;DR version: Why didn't OnLive take off then? What are Google and the like going to get right what everyone else got wrong about physics?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
