We just tried out OnLive, and of course it worked perfectly, being a demonstration on the order of 8-10 machines set up by the company itself. The people we talked to were naturally very optimistic, and my concerns over the availability and reliability of multi-megabit connections were waved away. That isn’t the only problem, though. OnLive will need a top-tier computer for every player at all times, and when a new game is introduced, the volume and subsequent will be unsupportable. Add this to the fact that video quality (which looked okay to me) is questionable at the framerates they’ve set forth, and things start looking a bit pie-in-the-sky.
This informative article at Eurogamer is immensely skeptical, and understandably. These are serious questions, and they demand real answers from OnLive, even if they have to be coy about the details. If they want to serve 20,000 people at a time, they’ll need 20,000 computers or the equivalent thereof. Even at the hugely conservative estimate of $1000 per machine, that’s $20 million just to start. Is growth sustainable? Is the revenue sufficient? Once the shine wears off, will we find that OnLive is making promises it just can’t keep? We’ll know soon. Until then, I’m withholding judgment, but leaning towards calling BS.
Know something we should know? E-mail us your tips! We respect anonymity. »
Comments
by
Devin Coldewey
on
March 26, 2009
Responses
Comments 
Leave Comment
Commenting Options
Enter your personal information to the left, or sign in with your Facebook account by clicking the button below.
Alternatively, you can create an avatar that will appear whenever you leave a comment on a Gravatar-enabled blog.
Trackback URL
- Actively Discussed Posts
- Contest: HP MediaSmart Server EX495
587 comments - Hasn't Infinity Ward earned your trust for Modern Warfare 2?
61 comments - Exclusive video of the Litl Webbook
43 comments - Siren.gif: Microsoft COFEE law enforcement tool leaks all over the Internet~!
33 comments - Ford does the unthinkable, announces the $650 car
18 comments
- There are no posts to display.
You know, with the Onlive system only providing a 720p stream (1280×720x60Hz), it’s not actually all that demanding, graphically. A GeForce 9600GT (about 2 generations old now) could easily provide that, at high quality, in most games (Though maybe not Crysis – That thing’s a graphics-card killer). Sure, you couldn’t run it on your Intel IGP system, but you wouldn’t need absolute-most-top-of-the-line to run it either.
In fact, I’m sure you could find some way to run this on an ITX box, or a blade server. Then the scale of cost goes way down.
Now, if it was 1080p they were talking about (1920×1080x60Hz), that might be a different matter.
Just a thought.
If they’re serving 20k people at a time, that means they have at least 100k users. If they can milk $20 out of each user, that’s $20m. With a subscription service like $10 a month, they are banking quite well. And with compression, 720p shouldn’t be a problem over home broadband.
Biggest issue I see is the latency… round trip is probably like 150ms on most broadband, which means you move the controller and there’s a delay noticeable to the eye before something moves on the screen.
well the latency between there sever is 1ms latency i just dont understands how much that will go up after being transfered to you computer
186,000 miles per second – it’s not just a good idea, it’s the law. A 1ms latency is physically impossible for anyone not living in their datacenter.
Interesting how you’re thinking system requirements could be a barrier. I don’t recall them mentioning anything about their servers running Windows.
Yes, it’s far-fetched that they created some kind of hyper-optimized Windows-compatible custom OS, but we’re already being asked to suspend our disbelief that it can work at all.
I agree with Eurogamer that I want this to work, though I’m less skeptical that they’ve truly pulled it off. Then again, I tend to suspend my disbelief when faced with something that I’ve been hoping would happen for years.
The Hardware problem was my main concern as well…
The PS3 has a rough power of 1.8 teraflops, and I am forever hearing how a top end PC today can run rings around a PS3. So lets say you need 2.5 Tflops to run crysis at the stated specs (60fps, 720p, full detail). So to have a 100,000 people playing crysis, you would need 250,000 TFlops of power to render it all.
The best supercomputers today can manage roughly 1 Petaflop, (1000 Teraflops), so you would need roughly 250 of the fastest supercomputers (E.g. RoadRunner, $133 million x 250 = $lots) on the planet today.
That is the part that I find hard to understand. I’m not saying it can’t be done, it’s just that the costs look impossible.
All that stuff is from my layman calculations. If somebody with a bit more knowledge on the subject could show me the error of my ways then I’d be very grateful
Hardware is not a problem. Latency will kill this. Nobody will play a fast paced game with a constant delay between your input and the reaction on the screen.
So, unless they only want to host RPGs and building simulations, I don’t see it working.
Why is hardware not a problem?
they have less then a 1ms latency on there servers to you i dont know how they just did it
1ms is imposable that was there computer set up by them probably 20 ft away from the other one
The lag was reported a 40ms – 50ms and up to 80ms For actual field testing
50 ms is alot fingering how people spend an extra $100 to get a monitor with 2ms responce instead of 6ms or so
although i completely agree with this article saying that it wont work with the masive overhead the company is gonna have to pay
they are doing something totaly revolutionary
buying a little box that puts out 720p over the internet? not only that but there graphicly demanding games its totaly revolutionary thats why it sounds rediculous but i mean its a awsome concept never havong to upgrade your hardware. Games on demand? its just awsome and ill buy one when they come out.
don’t we all remember Sega Channel?
Didn’t sega channel download the entire game into memory first THEN play it?
@Devin
Sorry but $1000/user for hardware as you imply is way off.
Lets assume for a second that the avrage user will need the equvivalent performance of a ps3. such a thing cost $400, and if you start removing all excessive parts such as cabinet, blu-ray etc. we are down to $200 or lower for the core rendering CPU and GPU hardware. Then remember that 1 OnLive user dosn’t equal 1 machine, as people don’t game 24/7. Over time the might be able to have 4+ users per machine – then we are down to $50/user. Such a service will cost around $10/mo + as they are also the retailer of all games sold they will also get around 30% off all those expensive games. Very good business in the long run.
Shure they will have some initial and especially peak performance problems, but if they have spend 7 years developing this I suspect they are in it for the long run.
The latency problem seems more endangering.
Just my 2 cent.
//Kristian
you couldent run all high crysis on a striped down ps3 lol anyways they sell those Way less then what they cost to make
I think sony loses $100 for every one they sell
They would need a hell of a super computer to render crysis 1000 times real time at once
nope the have hit break-even according to the most recent isupply teardown. The actual cost for cell + gpu is even lower that I wrote above.
My point is that the hardware part is: a) not going to be impossibly expensive as Devin described, and b) their biggest problem (that’s latency)
I agree about hardware costs, for the most part. But what’s lacking in the calculation of cost is regular business capital, i.e. warehouse storage, installation, cooling, repair, advertising, upgrading over time, etc…. I don’t think it will add too considerably to the total for the service, but its not negligible either.
As for latency, obviously if it was really bad that would suck, but I for one would still go for it if it were only noticeable and not annoying. I know many wouldn’t, but if it’s worth shelling out the bucks to get/make your own system, you’re welcome to it.
Way too much overhead for this to be successful. $20 million just to start is an absurd number to begin with. Good luck with this, but I see a flop.