GaiKai Beta To Start In Europe "Later This Month"

Alison Beasley sends word that GaiKai, the cloud gaming service being developed by games industry vet Dave Perry, is about to begin beta testing in Europe. (Sign-up page.) GaiKai is a competitor to OnLive, which started beta tests of its own recently. IGN got a chance to try out GaiKai for themselves, and they've posted a video showing how it performed. From Perry's announcement: "Our closed beta has two goals. #1 is to bring our servers to their knees so we can choose the final configuration before we start ordering large quantities of them. (We think we have it worked out, but you can be certain our staff will be swapping cards and testing different processors as each day goes by.) Goal #2 is to test older computers. We've had lots of emails from people describing their computers and 99% of them have ample performance. Remember you don't even need a 3D card to see a 3D game run on our service. I know this is strangely counter to what people expect, but we actually want to get plenty of basic office-grade XP machines testing so we can make sure we can reach the widest audience possible. ... After we choose the hardware configuration in Europe, our next phase will be our USA Nationwide Network Test, that will be using 8 Tier-1 Data Centers, getting hammered by Closed Beta testers. During that process, [we] will be identifying the other data centers we need to include to blanket the USA in a low latency array. Phase 2 of that is Europe, in exactly the same test."

Comments

Casual Gamer

[dintech]

I assume this is going to be a subscirption type service? I'd love this if you could also use as a pay-as-you-go type of thing. I'm the sort of gamer who doesn't actually finish many games and only plays very infrequently.

Hmm... not a good history

[Cheesetrap]

Last time I played cloud games, I got in trouble...

(How was I to know that fog machines set off fire alarms?!) :o

Seriously though... this has potential :)

Re:Streaming games

[slim]

Two comments on this:

1. Gaikai is going for a model where their servers are widely deployed at the "edge" of the Internet. That means negotiating with ISPs to locate servers near the modems. Part of that deal will involve having sufficient bandwidth for those servers and those protocols.

2. This kind of service is going to build customer demand for stable, fast, low latency connections. Presumably market forces will cause ISPs to provide.

Re:Streaming games

[IBBoard]

2. This kind of service is going to build customer demand for stable, fast, low latency connections. Presumably market forces will cause ISPs to provide at extortionate amounts.

There, fixed that for you ;) As a UK broadband customer (albeit one who doesn't need much bandwidth), I can't see the ISPs offering the kind of levels that people expect in the US and other nations for a hell of a long time yet.

Someone commented on digital downloads recently that it was okay because 4GB was only a "small amount" of your average 30GB+ monthly cap. 30GB+? You're probably talking £30+ per month in the UK on top of your £12 per month phone line charge and some contorted "acceptable use policy", not the entry level £10-£15 that most people use.

Tell me this.

[thatkid_2002]

How the bloody hell is this meant to work. I have seen the videos but I still cannot believe it. How can they make it work across the Internet where we cannot even make it work at home on a Gbit LAN. Anybody have an idea of how all of this works? Special graphics driver?

Re:Tell me this.

[Rhaban]

Won't work. Can't work.

If youtube can't start playing a hd video without buffering first, They won't be able to live stream a game with an acceptable resolution. We're talking instant video-compression and streaming to multiple users without visible lag. The technology allowing this just isn't available anywhere right now.

Re:Tell me this.

[Svartalf]

Heh... I say it can't be done because you have to figure your PEAK bandwidth requirements per customer. The moment you oversell something like this in the manner the ISP's have done their bandwidth you're done- you can get away with probably half again more that the math, if you're lucky. If you can't provide snappy service a good 95% of the time, you're not going to get takers. WoW works as well as it does because it's lower bandwidth than this. Ditto most of the other MMOGs.

If you apply the aforementioned guide to how many they can service, unless you get the ISPs to one and all sign up for this and put it fully on the edge (I can tell you that this will be heinously expensive- there's several reasons why epicRealm failed, one of which was being in 50 data centers worldwide to the tune of a $2mil/mo burn rate- and this was with sweetheart co-lo deals...if you don't have the deals, it'll be more painful than that...), the peak numbers without oversell for OnLive, with their stated maximum bandwidth requirements, would be:

30 subscribers on a T3.
103 subscribers on an OC-3.
414 subscribers on an OC-12.
1658 subscribers on an OC-48.

Now, to put the burn rate for this in perspective:

Average cost of an OC-3 is 20,000 USD/mo.
Average cost of an OC-12 is 200,000 USD/mo.
Average cost of an OC-48 is about $400,000 USD/mo.

This doesn't even get into latency issues- either in the framework itself or over the Internet. Most games do "online" because they compensate for lost traffic, delayed delivery of traffic and so forth. As you fill the pipe, packets will be dropped (UDP) or delayed (TCP) as part of the TCP/IP congestion avoidance algorithms when they kick in (they start doing things to you at about 30% or so of the capacity of the pipe...). With so much bandwidth being used compared to the games we've got today, it's going to be difficult for them to accomplish the return end compensation for these issues. Dropped frames won't cut it here- you'll end up with a jarring experience that's different from lag induced issues that we've all seen with online games.

It works in the low-end numbers tests they're running (and they couldn't be running large numbers tests because of the associated burn-rate supporting more than a couple hundred subscribers...) because they're not tripping over peak values overmuch in the local testing or even the remote testing they're doing with GaiKai and OnLive.

As you can see, my disbelief has less to do with the compression and more due to realities of how the Internet and TCP/IP actually work- and they're going to be broken upon the wheel with this stuff. As for claiming that they're lying- I don't think they're knowingly lying. I think they've missed a few tenets that I've laid out in simplistic terms and they are barking up the wrong tree with a neat "what if we..." line of thought that should have been scotched when they did the aforementioned napkin math I did here in this post.