More on LoTR Special Effects

sushi writes: "Another LoTR article: this one focusing on the technology used at Weta Digital (the CG shop). Interesting that they are undertaking "major" R&D into running more Linux, and that Linux "delivers about two times the price performance compared to systems running proprietary operating systems". I've been lucky enough to have seen inside this place, and it's cool to see a render-wall of linux boxen. Full story from a New Zealand newspaper." We linked to another good article about WETA a month ago.

Comments

Massive

[Alien54]

Weta crowd supervisor Stephen Regelous has created software, dubbed Massive, that creates realistic crowds. Every individual in the crowd moves in response to stimulus such as terrain, and to the actions of others. The battles in The Return of the King will see hundreds of thousands of these intelligent agents in frame at the same time, Mr Labrie says, stretching the software to its limits.

Aside from the impressive technological feat, imagine looking forward to the day when effects like these are availble for Gaming Engines.

Imagine Quake IX out in an open plane of battle with literally hundreds of thousands of soldiers and other things out there all at once.

I am reminded of something similar to the weekend dogfights/lanparties at the Airforce Academy, but with a much larger field of action.

[smile]

Artists, not technology

[SilentChris]

Actually, a far greater tool was used than Linux-boxen in creating Lord of the Rings. Artists.

You could have all the rendering boxes in the world. If you have high-school hacks, you're going to end up doing hack-level work.

Directors at Pixar once said that they hire artists that can use computers, not computer people who claim they are artists. This movie would have been just as impressive if it was made without the technology.

And let's not forget the artistry involved in writing the book, which required millenia-old technology.

Re:I don't understand linux zealots

[sigwinch]

I do a ton of 3D models and animating, for money, even, and I have to tell you that a tool is a tool is a tool.

Well, there are tools, and then there are *TOOLS*. Especially when it comes to industrial-scale work, not all tools are equal. You hardly ever know exactly what you want to do when you start out, and in fact what you think you want often changes several times, so adaptibility and flexibility are highly valuable.

The only consideration for what OS to use for a major company is, "will it run the software I want on the hardware I want, quickly and well?"

Until you discover that you fucked up big time estimating hardware needs and all your boxes really need double the processor speed and RAM, as well as gigabit Ethernet instead of 100megabit Ethernet. And unless you have a crystal ball, such fuck ups *will* occur. (If you could predict future software needs, you wouldn't be lecturing us on animation, you'd be picking your teeth with the bones of IBM.) When requirements change, the only relevant question is "Can we just substitute new boxes?" If your OS can handle all popular hardware, the answer is probably "yes". If you picked a poorly-supported OS that is picky about hardware support, you're doomed, you have to stop and convert everything to a new OS. (And it's possible to pick a bad OS several times in a row.)

In my experience, and based on what I've read on the mailing lists and newsgroups, Linux has by far the widest hardware support. I've personally installed it on a huge range of machines, only finding one machine I could never get it to install on (and that was a BIOS problem). Conversely, the few times I've tried BSDs have been unmitigated disasters.

See, in the real world, it doesn't matter what car you drive, as long as it gets you to work.

On the other hand, a Ford Escort can't haul half a ton of widgets across town to a client's emergency, and even an SUV would have trouble with an emergency sheet of plywood. Fortunately when it comes to operating systems you can, in fact, "drive a cargo truck to work".

The idea that the "cost" of Linux having anything to do with the decisions of the directors to use it is really funny. Does anyone really think that a cost of even $500 per box would make or break this deal?

Sticker price is not the only cost. Adaptation and maintenance dominate, and Linux has significant benefits over many other operating systems. Need Beowulf-style clustering to do your rendering? Linux has excellent tools. Have software written for a 32-processor machine? Use the Mosix patches for Linux to tie together 32 ordinary PCs. Want to be able to rapidly upgrade and reinstall all 1000 machines in a cluster? Network boot Linux using the well-documented tools, and watch the entire cluster reboot in less than 10 minutes. (Try that with Windows sometime.)

FWIW, I use a few different systems in my 3D work. Some use an OS from Redmond. Some don't. It doesn't matter to me in the least.

Big jobs are different in kind from small jobs, not just different in size. A 10% mistake in a small job means you stay late that day. A 10% mistake in a one month, 1000 CPU job means you threw away the company. IMHO, the key to keeping large projects running smoothly is having flexible, adaptible tools, and the Unices come standard with awesome tools of all kinds. Windows, on the other hand, is worthless out of the box and remains inflexible even when you shell out the $$$ for the official tools.