ZBrushCentral

is 4096 the max for a map?

ZBrush Usage Questions
#1

I again am a new user, sorry to irritate.

But I cant see any option to make a map (normal / disp) larger then 4k.

Surely a stunning prog (i’m seriously impressed with it so far) can do more than 4k??

Mental Ray can handle larger than this.

And maya will let you make them bigger than 4k and handle > 4k too.

Thanks Zbrushers :smiley:

#2

ZBrush can create textures of up to 8000 in size. To create a normal map of this size, you would first create a blank texture that is the size you want and then enter ZMapper. ZMapper will create its map at the same size as the currently-selected texture.

However, just because you CAN use one this size doesn’t mean that you should. There are two big reasons why:


  1. Computers move data in blocks. An 8k texture is four times the memory requirement of a 4k texture. Any time you use that texture, your computer must shuffle the data for that texture as one big chunk. This is unwieldly, and bogs down performance. On the other hand, you could use four 4k textures to get the same amount of texture information with considerably better performance the part of your computer.
  2. An 8k texture has 64 million pixels. Does your model have that many points? If not, then there is absolutely no point in trying to create a displacement or normal map of 8k in size. Nor is there any point in transferring polypaint to a texture of that size. When you’re creating your maps, you want them to have about the same number of useable pixels as your model has polygons. Any bigger, and you’re just wasting system resources needlessly. As a rule of thumb, a 1 million poly model is equivalent to a 1k texture. A 4 million poly model is equivalent to a 2k texture, and 12-16 million poly model is equivalent to a 4k texture. (A texture always has some unused space.)
The bottom line: bigger does not always mean better. Look at the actual needs of your project before making a snap judgement that an 8k texture is the be all and end all of CG work.

#3

Auric,

thank you so much for the comments they are really useful to me.

Do we need this size? Well actually we do and maybe bigger, the models we invisage would be 200m polys or more. (once the HD modeling can be normal/displaced) mapped.

The reason for such size is if we pull in the camera in extreme closeup the texture / normal / displace will then hold.

Example if you have a car and dont have the map’s big enough and you pull into the wheel nut, the map/texure will not hold if the final output is itself 4/8k res, but the object will look fine from a distance. And thats just one of the things we would like to do.

But again thanks for you answer, it help me find the current “limits” of Zbrush, which is what i am doing with these “newbie” questions. :smiley:

#4

You’re still better off breaking the mapping up into smaller UV regions. As I said before, four 4096 textures have the same amount of information as a single 8k texture – but your computer will be able to work with these textures much more efficiently because it has more freedom in how to manipulate the data. You’d be better still with 16 textures of 2k in size, which again hold the same amount of texture information.

What’s more, in answer to your comment about extreme close ups: It’s a lot easier for your system to deal with these extreme close ups when it only has to use one small size, yet very highly detailed texture. Let’s put it another way: If your extreme close up shows 5% of the model’s surface, then why does the remaining 95% need to have the same level of detail? It doesn’t! And even if your shot pans across the surface of the model, it’s still more efficient for your computer to load only the highly detailed textures that it actually needs, and even then only when they’re actually in view. With small textures, your computer can offload the parts of the surface that are not currently being rendered. But with a single large texture, your system must hold ALL that data in memory at all times.

#5

aurick,

I see exactly your point, but lets look at the problem another way, and i only say this to add to your obvious knowledge, as a seriously vaible alternative.

By having 16 textures in 2k size, your right the system does not need to hold as much. But and here is the big but, this basically means in pre-production you have to have all the camera moves down. It unfortunately means less flexibility in terms of the director or DP making ad hoc changes (which always happens).

Further to the above it takes more time, which costs more money in terms of as I said the pre-production, and the modeling and texturing of any / all characters, and thats always a large portion of the cost.

Your point about the memory is well true, but the cost of ram vs the cost of employee time, is well insignificant, as is true for hiring (i got recently mailed) a render farm that charges for excess work 70c per ghz/hr for 150 quad core cpus. Compare those costs with the less involved work in generating one large n/d map for the artist, rather than have to generate 16 or so assuming there are no changes. Added to that the time to test the quality of each map generated. You see it can add to the production time and cost.

If I balance these cost of less flexability and employee costs vs a bit more ram a a bit longer rendering time, as i said its really a non-starter.

Now I admit for the majority of people, your methodology works beautifully, so if anyone elese is reading this apart from us :wink: , aurick’s comments are a very valuable.

But if you need ultra flexibility, mutiple pov, ECU’s etc and have the production budget then this (one ultra map) is a viable alternative, as we all know staff costs are a major impact, and I have yet to meet a dp/dir who doesn’t see the oppurtunities that cg gives and likes to test things out.

I would like to add on a personal note, aurick, thank you for your valuable assistance to me, it’s been a welcome pleasure dealing in this forum. Keep up the good work, as for me, it’s well appriciated.