正在加载视频...

视频加载失败

Fun fact: It would take the average PC roughly 138,069,000 hours or about 15,761 years to render Avatar: The Way of Water. Here’s how I worked it out: The average PC in 2024 has 6 cores. The average blade on WetaFXs renderfarm has between 16-64 cores. There is roughly...

37,819 次观看 • 1 年前 •via X (Twitter)

11 条评论

lee 的头像
lee1 年前

I worked there on Avatar 2 :) but left before could finish :( Amazing team, everyone amazing truely... Ive been saying 200k years but i forget my working :D True that its lots of rerenders to get the best everything (water/character/sfx/emotions/mocap etc) and done the way a chain of sr artists up to the director wants it too

Rassoul Edji 的头像
Rassoul Edji1 年前

Yeah if you work out absolutely all the rendering that went on for the show including iterations etc. it would be several orders of magnitude more than my calculations above. My calculations are just for a single pass of everything using averages so it’s very conservative. Great work on the film by the way!

pokedstudio ∞ 的头像
pokedstudio ∞1 年前

so my PC is not going to make it?

Rassoul Edji 的头像
Rassoul Edji1 年前

Probably not haha

Brian Hanke 的头像
Brian Hanke1 年前

I better get started then! Just share the assets on Dropbox and I'll set everything up. 😛

Rassoul Edji 的头像
Rassoul Edji1 年前

You’re going to need a huge hard drive 😄

g.balaji 的头像
g.balaji1 年前

you missed its an 3D film so its double Left and Right Eye render for each frame so its double the what you put as facts...

Rassoul Edji 的头像
Rassoul Edji1 年前

My calculations are for the 2D film but yes you’re right, the 3D version is even more computationally intensive!

MrExo3D 🇨🇦 的头像
MrExo3D 🇨🇦1 年前

That’s the final run time but prob 2-3x was rendered and a lot of left on the cutting room floor

Wycliffe Ndiba 的头像
Wycliffe Ndiba1 年前

You may be the perfect person to ask, could they top this? Could Wētā (incredible studio, love their work to bits) have more impressive papers and algorithms on pyro simulations cooking in their labs for their next movie to top this one in visual fidelity?

Rassoul Edji 的头像
Rassoul Edji1 年前

Absolutely, technology is ever evolving and rendering is no different. Weta is always working on more accurate and efficient rendering solutions and algorithms. This coupled with advances in hardware will allow them to do better work.

相关视频

Flip a coin. Heads, your account goes up 50 percent. Tails, it goes down 40 percent. Expected value is plus 5 percent a flip, so you take the bet a hundred times. Expected value says your $10,000 becomes $1.3 million. The most likely path leaves you with $52. An MIT professor explains the entire gap in one sentence, in a free undergraduate lecture, then moves on like it was nothing. His name is John Tsitsiklis. He teaches undergraduate probability at MIT. He also proved in 1994 that Q-learning converges, the result that says the algorithm under modern reinforcement learning does not merely happen to work, it has to. INFORMS gave him the von Neumann Theory Prize for that line of work in 2018. He runs the lecture on the students for an hour. First he takes the average apart. A random variable is not a number, it is a function. A bar graph of probabilities is a PMF. Expectation is the center of gravity of that bar graph, the single point where you slide a pen underneath and the thing balances. He is slow and patient about it. By minute 35 you trust the average completely. Then he stops and says he wants to give "one general word of caution." "The average of a function of a random variable, in general, is not the same as the function of the average... in general, you can not reason on the average." Everything before that sentence was the trap. Go back to the coin. Compounding is not addition. Up 50 then down 40 is not plus 10. It is 1.5 times 0.6, which is 0.9. You are down 10 percent. Do that 50 times each way and you have 0.9 to the fiftieth power. Fifty-two dollars. So where did the $1.3 million go? It is real. It is parked at the very top of the distribution. Run the hundred flips and only about one path in seven finishes above where you started. Only about one in a hundred ever reaches that $1.3 million. Those few runs are gigantic, and they carry the average for everybody else. You will not be in them. In February 2018 that trade had a ticker. XIV, short volatility, $1.9 billion in it. It had paid on the average day for seven years. On February 5 the VIX rose 115.6 percent, the biggest one-day jump ever recorded. XIV went from $115.55 to $4.22 overnight. Credit Suisse shut the note two weeks later. Nobody in it was wrong about the average. They were wrong about which path they were standing on. The usable version: your compound return is your average return minus roughly half your variance. A system averaging 20 percent a year at 40 percent vol does not compound at 20. It compounds at 12. That missing 8 is not fees or slippage. Tsitsiklis delivers the most expensive sentence in finance, finishes the variance section, and ends with "see you on Wednesday." The lecture is free. The average is free. Knowing which path you are standing on is the trade.

Veles

628,984 次观看 • 24 天前