Drake's Weblog

3 minute read

這個 renderer 揚名與世,同時在歐洲、北美與澳洲(這些都是 CG 產業比較發達的地區)都擁有分公司的德國公司。

一家是硬體大廠,一家是渲染器(算圖器)大廠,硬體廠(NVIDIA)併掉了軟體廠(mental images),雖然不如前一陣子的 Autodesk 併掉 Alias 這麼轟動,但大抵來說,也算是 CG 軟體產業的一大新聞(CG 軟體產業,本身很轟動新聞就很少…),於是我們幾個人,隨口聊了一下,各自抒發了看法,帶點嘲弄兼話家常地聊著…

GelatoNVIDIA 在某年,成立了一個 NVIDIA’s Digital Film Group,大名頂頂的 Larry Gritz 加入這個團隊,從裏頭誕生出了 Gelato (這個字,其實來自義大利,指一種冰淇淋,我記得在台電大樓的隔壁有家店吃得到)這個利用 GPU 的 renderer 出來。這個 renderer 雖然利用了 GPU,但並不屬於 real-time 的 renderer,而是比較偏向電影特效使用的,就像 RenderMan 或是 Mental Ray 一樣,所以它一樣有一些比較高階的功能,像是 motion blur、procedural shader、depth-of-field、anti-aliasing、super samplging…。

mental ray 被公認為,是最有可能與 Pixar 的 RenderMan 匹敵的 renderer 了。網路上有關 renderer 的口水戰,最常看到的,也幾乎(可能有 90% 吧)都是這兩個 renderer 的比較。我完全沒有使用過 mental ray 的經驗,也不曉得要說啥,我只知道,李安的綠巨人浩克、俄國兄弟的駭客任務與明天過後,都是使用它來算圖的。又因為 Softimage XSI 裏最常被拿來用的高檔 renderer 就是 mental ray,而 Softimage XSI 又很紅…所以…你知道我的意思的。

Drake:

NVIDIA 的 Gelato,從商業模式來看,算是一個不成功的產品。

怎麼說呢? 首先,它提出了一套有別於其它人的規格來(不是 RenderMan 的 RIB + RSL, 也不是 mental ray 的 mi)。其次,它比較像剛開始發展的 renderman renderer,還沒有很好用(或夠多的 scripts, tools)與其它 3d 工具的整合(ex, RMS, RAT or mental ray in maya)。再者,它的算圖品質還沒獲得大量的證明前,然後它的速度又沒有比全數用 renderman 來得快很多(like one order speed-up!)最重要的是,它不便宜。

不管我是出錢的老闆,或是出腦力的工程師或美術人員,它的投資報酬率都太低了。於是我不可能會在花了很多錢,然後又花了很多工程師的 man-weeks 在上頭,然後替它寫了一堆工具程式與補丁,最後發現算圖的結果或時間比沒有我花的資源來得多。於是我決定,還不如選個 renderman 或 mental ray 來得好。

下頭是信件往來中,我最早寫下的內容:

Dan and I have already discussed this and we got 2 interesting conclusion:

1. NVidia thinks their self developed renderer (Gelato) sucks in technical and business model because it has its own file format, not RIB or other defacto standard which makes it hard to broadcast especially when there is no significant benefit while using it. And one more thing. it’s too expensive for people or companies to devote resource on.

2. NVidia might think of promoting their future hardware products by utilizing GPU’s power to enhance MentalRay’s ray-tracing renderer. In other words, there might be a first famous and broadly used renderer making great usage of GPU. Well, there are lots of possibilities and here is just a simple idea.

公司的 Jeff 如是說:

How about a third possibility?

NVidia will keep developing its Gelato for consumer markets (mainly for games), and will enter professional market (like CG movies) through the acquisition of Mental Images. Moreover, Gelato team can learn or borrow some techniques from Mental Images. In the past, NVidia has recruited some people from Hollywood, and this acquisition just further confirms NVidia’s determination on professional market. Brand is also very important in the professional market, so instead of selling Gelato to filmmakers, NVidia may want to sell “MentalRay powered by NVidia” to filmmakers. Sounds more attractive.

RenderMan 達人兼宅男 Dan 如是說:

My view is that Gelato has failed in the marketplace, and Nvidia bought Mental Images as a way of staying in the high-end graphics market. I think Nvidia will probably stop development on Gelato and integrate the team with Mental Images.

Gelato failed because it is too different from RenderMan and Mental Ray to make it easy for studios to adopt it, and also the performance increase relative to RenderMan is not very large. As Drake says, adding more GPU acceleration to Mental Ray seems like a better strategy.

ooxx ooxx ooxx ooxx ooxx(涉及商業機密,與以銷毀).

遠在它鄉(加拿大)的 Satan(仁瑋)如是說:

總覺得Nvidia還有更大的目標…..

也許只是買到一些 software patent,然後就可以放手去做一些其它更遠大的目標。因為有些東西如果配合硬體的話,做成 patent 會很簡單。algorithm 通常是不能申請 patent 的…. RSA 之前試過…可是那個 10 年一過已經不會在延續…傷害太大。so…配合硬體的話……就變的很簡單。要不然就是你不發表變成 closed source。但是還是有一天別人可能發現其他的用法申請 patent,還不如先下手為强。現在科技的進步就幾乎是 patent 攻防戰,先卡到位置比較重要,所以很多東西產品還沒出之前 patent 就先申請好了,搞定之後才會 announce。

comments powered by Disqus

Recent posts

Categories

About

You're looking at Drake's words or statements. All opinions are my own.