A Look At The Data Center That Crunched Avatar
from the behind-the-scenes dept
You don’t often get to see the details of a massive data center. The companies that run them tend to keep things pretty quiet, as they view the datacenter as a competitive advantage. Thus, what happens in Google’s datacenters remains mostly a mystery. And yet, it seems that the folks at Weta Digital, famous for providing the computing horsepower behind major Hollywood blockbusters like Lord of the Rings and now Avatar are apparently willing to open up a bit and provide some details about its setup. What struck me as interesting wasn’t so much the hardware specifics, but how they had to switch from the industry standard cooling system of raised floors and air-cooling, because the machines were too close together to get the necessary bandwidth. So, instead, they went with water-cooled racks. Water-cooled data centers have been increasingly common over the past few years (and were typical with many old mainframes), but they’re still a technology that not all data center operators are comfortable with, and which many still think create more problems than they solve. So it’s always interesting to see another one in action.
At the same time, as neat as it is to read about Weta Digital’s massive computing power (which apparently represents one of the 200 largest “super computers”) in the world, I’m still left wondering if the trend — even for amazing movie effects — isn’t moving away from such massive data centers. We’re seeing more and more what can be done on the cheap. And, no, it doesn’t come close to matching the stunning effects found in the blockbuster movies that Weta works on, but it does have all the symptoms of a classic innovator’s dilemma scenario, where the new stuff isn’t “as good” as the old stuff, but is improving at a faster rate, and quickly reaching a point where it’s “good enough” at significantly lower price points.
Given the regular discussions around here concerning movie budgets, where do people think the technology is headed for movie special effects? Will it always be run in giant datacenters, or is there a place for making high quality (even blockbuster-type) films on cheaper hardware?
Filed Under: avatar, data centers, movies, special effects
Companies: weta digital
Comments on “A Look At The Data Center That Crunched Avatar”
worst. opinion. ever.
you really like to overemphasize the fringe minorities don’t you?
the trend has clearly been to move towards massive data centers that specialize in crunching, which then rent out server time to heavily financed ventures (like james cameron blockbusters). maybe it’s just me, but the BILLIONS of dollars speak loads against your opinion.
“We’re seeing more and more what can be done on the cheap.”
I’ve talked to someone who does video editing and they said that this guy is an amateur. Said you can probably download the robots off the Internet and even the walking motions. It’s an awesome video though (ie: it has a great storyline) but I suspected that whoever did the video editing is an amateur just by watching the walking motion of the robots. Then again I don’t know anything about video editing myself.
Re: Anonymous Coward SUCKS!
You really, really do not understand this. Your ‘animator friend’ obviously doesnt know much about ANIMATION! Wanking perhaps!
The walking motions of the robots is actually quite sophisticated. The swagger of the hips, etc.
However, what needs to be said is that this is bare basic, UGLY millitary technology that the humans are using. It shouldn’t be pretty, because the robots aren’t
What? You mean Avatar’s data center wasn’t built into the back of semi truck containers set strategically around the country to idle behind the nation’s favorite retail chains such as Walmart?
Both
I think there will continue to be both types of special effects: huge, supercomputer effects and lower-budget, still-great-looking effects.
Alien, of course, is an example of a low-budget film with special effects that lack nothing at all. A more recent example is Duncan Jones’ Moon. A really low budget upcoming (Finnish, I believe?) movie that appears set to have at least competent special effects is Iron Sky, with a budget of just over $7 mil. At the risk of going on for way too long about this, I also have to mention the upcoming web series The Mercury Men, which is another small, soon-to-be-released project with special effects that, from what has been released, look entirely competent, convincing, and even awesome.
The thing is, if the director is talented enough, he/she can make really convincing effects on a low budget. On the flipside, though, you have to be good to use “big” special effects, too. Just see the new Transformers movie, which is plagued with so much erratic cutting that even though the effects look great, you can rarely tell what’s going on in the more special-effects-heavy scenes.
Really, special effects can be done both ways. James Cameron has demonstrated a masterful hand with big-budget effects. Other directors seem to thrive off limitations. From my experience (as a film critic), it all just depends heavily on director talent.
Re: Both
Some directors, you just know how the movie will go in terms of special effects. When Michael Bay announces his next project, it’s time to invest in dynamite futures because that market is going to explode (no pun).
I’ve just seen the previews for the remake of Clash of the Titans. Now that’s going to be an interesting one. I’m not sure I’d’ve been brave enough to pit modern CGI special effects against Ray Harryhousen’s stop-motion animation. Not to mention the story. I mean, it’s a remake; you’re really limited in how much you can deviate from the original. They have a huge hurdle to leap and it’s a long drop if they don’t clear it.
typo?
“but how they had to switch from the industry standard cooling system of raised floors and air-cooling, because the machines were too close together to get the necessary bandwidth.”
This didn’t make sense at first so I read the article on their site – I think you mean they were too close for air cooling to be effective, but that they needed to be close to get enough bandwidth.
story first
imho the huge servers required to render huge blockbusters will continue to have a market for their services as long as the majors in hollywood keep trying to produce huge blockbusters. (the foreseeable future that is and long enough for the server farms to recoup their costs)
i.e. filmed entertainment wherein the technology and not the story is paramount (no pun intented).
anecdotal story: i begged my two daughters (10 & 13)to go see avatar with me, they refused: “who wants to see a horrible story about 12′ blue aliens?” me, i wanted to see the technology in action having never been to a “3D” movie, the trailers told me the story was run-of-the-mill at best.
btw, what’s the r.o.i. on paranormal activity?
Re: story first
Avatar’s story isn’t horrible, just done a million times (Dances with Wolves, Shogun, The Last Samurai, etc.). Although, if your daughters are young enough, they may not be familiar with the whole white-enemy-goes-native story.
Also, why the heck are you begging your daughters? Tell them you’re going to the movie and they don’t really have any choice but to come with you 😀
I imagine that if you add more fluoride to the water, the heat transfer may be higher, leading to a datacenter that remains cooler.
Where it is going?
I think the game industry will engulf the film industry.
People will have interactive stories to play and thousands of different endings and each time you see(play) the movie it will be a new experience with little differences.
I remember being scared playing the first resident evil(biohazard in japan), or being amazed at how Castle Wolfenstein(the apple version) had such great graphics and now you have blue people being done entirely in CGI on the screen.
NTT was demonstrating such possibilities already using a 3D immersed technology to showcase virtual walks on EDO era streets.
And going even further people could even smell not only hear things if the frog mask concept comes to reality in one form or another, and people are working very hard to make it happen, people like Pranav Mistry that demonstrated how a kind of sixthsense can be done using of the shelve products that can be combined to create a reality overlay.
All those things will come together in the future I believe and will with more things that can create a unique experience individualy or in group.
That is my vision of the future. I would like to things to be interactive and connected.
I imagine teens grouping together to see a film and one says “hey look” this would be cool and then they show it to each other.
Now who holds the copyrights for those stories? LoL
When artistic materials are used as props on peoples on visions of the stories how is that going to sell more if people can’t use it?
Re: Where it is going?
Looking at how much money something like Modern Warfare made, even compared to Avatar, perhaps movies will become free promotional content for the games!
Re: Re: Where it is going?
Looking at how much money something like Modern Warfare made, even compared to Avatar, perhaps movies will become free promotional content for the games!
in general, video games based on a movie franchise tend to suck miserably. also, movies based on a game franchise doubly so.
i’m not saying that will always be, just that it’s what always has been.
there are exceptions: butcher bay and arkham asylum were two movie based games that surprised much of the community by not sucking.
What it ought to cost
Computer animation ought to be cheaper than live action. The fact that it isn’t says a lot about the way the film industry works.
Doubtless “Toy Story” could now be rendered on a standard PC in a reasonable time – but if they made a new “Toy Story” they wouldn’t be happy just to reproduce the effects of the original.
Likewise, unless Moore’s law hits its thermal limit in the meantime, Avatar will probably be renderable on the PC I will have 10 years from now.
Also the “SETI at home” approach would allow an amateur group to crowdsource the computing required to render “Avatar” if they could marshal the public interest. After all the Microsoft search puppy probably consumes more computer power per day than all the supercomputers in the world put together….
There are two main reasons why computer animation is expensive at present.
1. It is bought alongside other expensive items (star actors voices and all the “grunt effort” required to create 3D assets do mo-cap etc etc) and so there is little incentive to cut costs.
2. They always want to be “better than the last one” even if the result is scarcely visible on the screen – let alone important to the story. The slow rendered frames that end up on the screen will be using ray tracing, radiosity etc etc – techniques that can absorb unlimited computation if you are lazy – but the direct rendered (real time) versions that were used to edit the sequences would probably be perfectly adequate to tell the story.
A cheap computer rendered film will arrive when some bright spark with no money has an idea that requires computer animation and just goes for it. Maybe we will see the “Blair Witch” of computer animation and then it will all change.
Re: What it ought to cost
Also the “SETI at home” approach would allow an amateur group to crowdsource the computing required to render “Avatar” if they could marshal the public interest. After all the Microsoft search puppy probably consumes more computer power per day than all the supercomputers in the world put together….
one that is here and now is called the Big Ugly Rendering Project, and it uses BOINC as the framework:
http://burp.renderfarming.net/intro.php
Time's are a changin'
After reading this article, it appealed to the inner-geek and possible budding video/movie producer type. All these high budget special effects movies will be able to be produced on systems that are starting to fall in price way down into the budget of a home-enthusiast’s pricerange. Supercomputers are coming down in price to where you can not only build them cheaply, but also buy them for a reasonable price. The GPUs are dropping in price while geometrically increasing in capability for production. All this leads to hardware tools for the home movie producer to use. Of course, the area that usually lags behind the hardware is the software, but even this is starting to catch up.
The software will pretty soon be out there, if not already, for home users to download and use. Then for the rendering, all one will have to do is build, or buy, a machine that can do the rendering while the home user goes about his/her way.
Eventually, there will most likely start to be crowdsourced/group collaborated movies where you will find users out there working with others who have developed specialties like sound effects/visual effects/characters/3D environments getting together to make or tell the stories in movie format.
I also agree with the person who commented about games because that is where the whole “movie” perspective is going anyway too. Interactive environments for the potential viewer.
on a side note
It’s funny how the right wing nuts do not like the movie