快捷导航

[渲染] VRAY渲染提速的40个方法

[复制链接]
2014-6-5 19:34 / 查看: 17783 / 回复: 18
V5CG提示您:本文由saundy发表,如需商业用途或转载请与saundy联系授权,谢谢配合!
saundy
积分:27
最新注册
architectural-visualization-luxury-house-2.jpg

VRAY渲染提速的40个方法,求大神们给翻译吧,不翻译,我实在看不懂啊。
  • Limit your polycount to minimum as much as possible. The more polycount, the slower the rendering.
  • If you are using Vray, always proxy the ones with high polycount.
  • Don’t put too many subdivisions on your shadows (as much as possible).
  • Remove unwanted objects from your scene
  • Don’t use giant resolution textures. I.e. If you render your final image at 3500px you don't need more than 4000px image anywhere.
  • Material wise: avoid using too much glossiness and highlights. A simple diffusion, bump and reflection(with .95 – .75) will do. Try to use your own materials rather than prepared. You will have better control on them.
  • Use high polygon objects only when they are close to your camera.
  • Know the difference between copy and instance… Simple but effective
  • Get rid of the undo window, especially commands on the list that eat up memory…
  • I always shutoff my antivirus while rendering (or don’t install one at all).
  • Be wary of iteration levels when making curves/mesh smoothing.
  • Caustics are a time killer. It's up to you if you want to drink and eat and sleep until the rendering finish.
  • Be wary of glossy reflection and frosted materials..(case to case basis)
  • Instead of using dof in physical cam, Photoshop can do the trick with zbuffer channel or plugins like Alien Skin’s DoF and DoF Generator PRO by Richard Rosenman.
  • Post process could  be short  the rendering workflow, professional composition programs like Autodesk’s Combustion, Photoshop, After Effects, Nuke, Fusion… could help you a lot.
  • Irradiance and Light Cache (save to file), skip the computation when using the same sets and scene over and over again.
  • Before you place all the shaders, first try to override the materials in the rendering parameters (global switches) to make sure all polygons are modeled correctly for test rendering so that you know the types of shaders you placed individually and you can trace easily which materials could possibly cause the rendering longer.
  • Too many lights can also cause overkill of the rendering.
  • I prefer ADAPTIVE DMC rather than Adaptive Subdivision…
  • Check your Raycast Parameters too...Like render region division, region sequence etc.
  • If you’re using Vray displacement mode, try to limit the area of 2d mapping and setup for it eats up a lot of ram.
  • Use 64bit Max
  • Attach those objects – 3ds Max is much more efficient working with 9000 10k polygon objects than 200,000 1k polygon objects. Attaching all the components that are not being individually animated adds a significant amount of speed and flexibility to the scene you’re working on.
  • Geometry Proxies – Ah yes, my new favorite tool! A geometry proxy is an optimized piece of geometry that is designed to load and render much more efficiently in specific render engines such V-Ray.
  • Bitmap Proxies – Bitmap proxies are a great way to minimize the amount of RAM you expend on a scene.
  • Work Locally and Incremental Saves – You must wonder why… Well, as you save to the network share, you can run into some traffic issues which can occasionally cause crashing and corrupted documents.
  • You might want to check the multi-threading option if you use a dual processor.
  • Use Rendering Region: render only what interests you. From time to time check your shaders, verify the little differences and quickly find the result you’re looking for.
  • Do not add Glossy effects. Add it only when you think that the scene works fine.
  • Geometry – scenes with lots of objects and/or triangle counts require more memory to render. There are several ways to reduce this amount: -Adjust the raycaster settings in the System rollout (reduce Max. levels, increase Min. leaf size, increase Face/level coefficient and switch from Static to Dynamic Default Geometry).
  • Displacement mapping – objects displaced with the 2d displacement mapping method may require a lot of RAM to render, especially with large displacement maps. If this is the case, use the 3d displacement mapping method. Also, if you have several distinct displacement modifiers with the same displacement map, it is better to replace them with one modifier applied to all the necessary objects. This is because each modifier will take RAM for the displacement map, separately from other modifiers, even if they have the same map.
  • Bitmap filtering – Summed area filtering uses much more memory than Pyramidal filtering. Use summed-area filtering only for smaller bitmaps.
  • Shadow maps – these may also take up significant amounts of RAM. Again, these are managed by 3dsmax and VRay has no direct control over their memory usage. To reduce memory usage, you can switch to raytraced VrayShadows instead.
  • Image buffer – large output resolutions require a significant amount of RAM to store the final image. Additional G-Buffer channels increase that amount. There are several ways to reduce this amount: -Use the 3dsmax Bitmap pager, if you render to the 3dsmax default VFB. -If you use VRay’s own VFB, use the Render to VRay raw image file option and then use the VRay raw image file viewer to convert the final file to a different format.
  • Image samplers (AA) – the image sampling algorithms of VRay require some amount of RAM to hold all the data for the sampled image. This amount can be quite large, depending on the chosen bucket size and sampling rate. To reduce that amount: Reduce the bucket size. -Switch to a different image sampler – for example, the Adaptive DMC sampler uses less RAM than the Adaptive subdivision sampler.
  • Even though you select the VRAY VFB as your output, the 3dmax VFB is still created and thus takes additional memory. If you want to reduce that memory, you need to uncheck the “GET RESULUTION FROM MAX” option. Set the 3dmax resolution to lower value like 100 x 100 and then choose your real output resolution in the VRAY VFB option.
  • Inspect your scene -were there lots of unused polys? Were there alot of models that only clatter the scene but are not viewable on camera? Were there lots of unused materials on the editor? I mean, sometimes we think that our render setup is hampering the rendertime, but if we just inspect our scene, the things mentioned above are factors that affect a lot.
  • Sometimes our model in autocad was located far from the axis origin; this happens quite often. Then once we link it up to our rendering program we just let it as it is. This also prolongs the rendertime when executed. Bring it back in 0,0 axis position while still in autocad.
  • Overlapping of Models and meshes. – A scene with many of these will take longer to render. You need to tweak your setting into higher so as to cover up those splotches.
  • Rendering Large Images – rendering a large output image takes longer when rendered as a whole. Use alternatives like split rendering or some render plugins like Super Render – where, in the scene, it will be subdivided into bucket window and then automatically combine them after the last bucket.
加入土豪吧,爽!
anzyu -2014-6-6 11:56:59 | 显示全部楼层
粗略的看了下,能力水平有限,有兴趣的凑合着看吧!欢迎指正和补充。轻点喷~~~

1、尽可能的限制场景面数,场景面数越多,渲染越慢。
2、如果使用Vray渲染器,面数多的物体使用VrayProxy。
3、阴影细分尽可能的不要给的太高。
4、删除不需要的物体。
5、不要使用高分辨率的贴图,例如你的成图尺寸为3500像素,那么就不要让4000PX的贴图随处可见。
6、避免使用太高的模糊反射和高光,简单的漫反射,凹凸和95-75之间的反射足已,尝试着自己去调材质而不是材质库,你会发现你会控制的很好。
7、精模一般在特写镜头或者离相机特别近的时候使用。
8、要知道实例与复制的不同,虽然简单但非常实用。
9、摆脱撤销窗口,尤其是命令列表,灰常吃内存。
10、渲染的时候关闭杀毒软件或者直接不安装。
11、曲线网格平滑的时候注意迭代级别。(例如涡轮平滑的时候,迭代次数太高会死JJ,控制在两次以内足已)
12、焦散是渲染时间杀手,如果渲染的时候你都在吃喝拉洒睡,这完全没有问题,由你决定。
13、遇到镜面反射和磨砂材质的时候需要特别注意,根据实际情况而定。
14、DOF尽可能不要渲染来完成,因为Photoshop的通道或者像Richard Rosenman开发的Alien Skin's DoF和DoF Generator PRO这两个插件都可以搞定。
15、后期过程或者一些微小的后期效果尽可能不要利用渲染来完成,像Combustion、PS、AE、Nuke、Fusion等专业的后期合成软件对你帮助很大。
16、当你在同一个场景中反复使用同样的设置的时候,发光贴图和灯光缓存信息是可能不会重复计算的。需要注意!
17、在给场景赋材质之前,首先使用全局覆盖材质来测试下看场景模型是否完全没有问题,也方便你可以清楚的知道场景里面都有哪些材质类型和可能需要的渲染时间。
18、灯光数量太多也会导致渲染速度太慢。
19、我更喜欢准蒙特卡罗(Adaptive DMC)而不是自适应细分(Adaptive Subdivision)。
20、渲染一些小的区域或者序列来检查下Raycast参数。(不知道是么东东,也可能是翻译理解错误。)
21、如果你使用Vray Displacement,请用2D mapping.
22、使用64bit Max。
23、减少物体数量,9000个10000个面的物体会比200000个1000个面的物体的运行速度快很多,附加所有的零部件(场景中零散的小物体),极个别有动画的物体的运行速度和灵活性也会好很多。
24、几何体代理,它是一个几何设计负载以至于在像Vray这样特定的渲染引擎里更有效的进行渲染。
25、位图代理,位图代理是将一个场景随机消耗的内存减到最少的一个极好的办法。
26、随时在本地增量保存文件,因为当你保存到网络共享盘的时候,可能会因为一些问题导致你文件损坏。
27、如果你用的是双核处理器,你可能要检查下多进程选项。
28、区域渲染,渲染你想要渲染的地方,不时的检查材质。快速得到一个渲染结果便于验证一些小的差异。
29、少添加光泽(Glossy)效果,除非你的场景真的做的很好。
30、几何体-场景物体或者三角面数量太多会需要大量的内存来进行渲染。这里有几个还不错的办法来尽可能减少你的内存使用数量:在系统卷展栏调整光线追踪参数。(减少Max.Levels,增加Min.Leaf size,Face/level,开启默认静态到动态几何体)
31、置换贴图-被转换的物体使用2D置换模式可能会需要大量的内在来进行渲染,特别是较大的置换贴图。另外,在同一个案例中,如果你有很多个置换修改器并使用同样的置换贴图,最好是用一个修改器应于于所有必要的物体(实例复制修改器)。即使它们有着同样的贴图,但修改器仍是独立计算的!
32、位图过滤器:Summed-area会比Pramidal filting使用更多的内存,Summed-area仅使用于一些像素很低的贴图。
33、阴影贴图:这些也可能会占用大量的内存。此外,3ds max和Vray也没有直接控制它们的内存使用情况。
34、图像缓冲区:输出高分辨率的渲染成图,储存的时候也会花费大量的内存,额外的G-Buffer通道数量增加。这里有几个方法来减少这些数量:如果你使用3ds max默认的帧缓冲区可以使用3ds max pager,如果是Vray帧缓冲的话可以勾选Render to V-Ray raw image file选项然后使用V-Ray raw image file viewer转换成其它的格式即可。
35、图像采样:VRay的图像采样算法需要一定数量的内存来保存所有的数据采样图像。根据所选的采样率来减少这些数量:使用不同的图像采样器,使用准蒙特卡洛就要比自适应细分消耗内存少的多。
36、即便是你使用VrayVFB用来输出成图,默认的帧缓冲区仍然会创建图像并消耗额外的内存,如果你想减少内存消耗的话,就必须关闭get resulution from max选项。将max的输出尺寸设置的很低例如100×100然后在VrayVFB中设置你真正想要输出的成图尺寸。
37、整理场景看是否有多余的多边形、相机视图不可见的模型以及材质编辑器中用不到的材质。我的意思是,很多时间我总是会认为是因为渲染参数设置的问题导致浪费了很多时间,其时只要你仔细检查我上面提到的这些因素会省很多事儿。
38、有时候我们的模型是从AutoCAD中导入进来的,但是它离坐标轴原点太远。这个时有发生,一旦用这样的场景来进行渲染的时候也会花费很多渲染时间。解决办法,把模型归零即可。
39、模型重面或者面与面之间离的很近这也会花费很长的渲染时间,你势必会开很高的参数来解决一些渲染造成的躁点。
40、渲染大图:作为一个整体渲染输出一个大的图像也需要花费很长的时间,我们可以用一些渲染插件或者分层渲染的方式来进行。在一个场景中,渲染的时候会分成一个个Bucket降低内存的需求,最后合并成一个图像!(Bucket可以理解为渲染进行时的渲染格子。最后一句也可以理解为分布式渲染,都差不多,意思表达到就行。)

评分

参与人数 2节操 +2 威力 +2 收起 理由
深蓝fever + 2 很给力!
admin + 2 很给力!

查看全部评分

寒江雪 -2014-6-6 11:43:52 | 显示全部楼层
看到这帖子真是高兴!
╭♥.Madlife☂. -2014-6-7 13:50:34 | 显示全部楼层
牛啊   又学到知识了
津城帅哥 -2014-6-7 17:22:01 | 显示全部楼层
泡一壹清茶,看着你的作品,慢慢品
basc520 -2014-6-15 09:50:28 | 显示全部楼层
我只是来回帖的,无回帖,不论坛,这才是人道。
討厭哋凊 -2014-6-15 13:26:15 | 显示全部楼层
这不是高帅富玩的吧~
小笨 -2014-6-16 13:34:48 | 显示全部楼层
其实这种应该不是很容易实现吧,我也不懂...求指导
powergraf1 -2014-6-17 21:41:01 | 显示全部楼层
整理的蛮好的
淡然 -2014-7-2 14:14:53 | 显示全部楼层
好东西     谢谢楼主分享  同时也谢谢 翻译那位朋友
hahaxxx -2014-7-11 13:36:03 | 显示全部楼层
好东西  学习学习
lianqin8888 -2014-8-2 15:23:43 | 显示全部楼层
泡一壹清茶,看着你的作品,慢慢品
tjabcb -2014-8-7 19:53:48 | 显示全部楼层
想下载看看!~
今儿个真高兴
fezhfu@qq.com -2014-10-22 22:26:31 | 显示全部楼层
收到的的地地道道的
82872779 -2017-3-3 10:12:48 | 显示全部楼层
学习学习学习学习学习
lianqin8888 -2017-3-9 13:14:30 | 显示全部楼层
看到这帖子真是高兴!
您需要登录后才可以回帖 登录 | 立即注册

本版积分规则