[misterasset]

Okay, so I'm getting into more and more arguments with my coworkers about this...

I'm working on a model for a client (I'd post example pictures but we signed a non-disclosure agreement). It's a gigantic two-story control-room type space with conference rooms, incident room, release management rooms, show-rooms, stairs with intricate guardrails, balconies to match and an overall foot print of 11,000 sq. ft. I put my model into MAX, we model in AutoCAD, and the walls, flooring, all that stuff minus furniture comes in at 300,000 faces.

To me this isn't unreasonable, but we keep arguing over "making it as efficient as possible." Plus they're using radiosity and it keeps crashing because with the furniture (26 consoles, 136 lcd monitors, 38 keyboards, 38 mice, 38 chairs, conference tables, so forth) it's a little over 1,000,000 faces. So they're blaming my model and wanting to make it "more efficient" by using face modeling instead of solids modeling.

Now maybe it's just my pride, but I don't think my part of the model is the problem. I put this into Maxwell and it runs just fine. Plus with his faces light leaks into the model in a few spots but he's going to put opposite faces in there to block the light.

Is this just an issue because they're stuck in the 10 year old world of scanline and radiosity and I'm right that face modeling is dead (it's not even in AutoCAD 2007) and that if they used Maxwell and advance renders solids aren't a problem?

[michaelplogue]

For my two cents, a million faces is really not that much - I've rendered much larger (using Maxwell, Brazil, and Mental Ray). I personally prefer to model using primitives as opposed to face modelling - so much easier and faster in my opinion - especially when it comes to mapping. It's also much easier to modify individual primitive objects as opposed to modifing a unified face modelled object.

You may take a bit of a speed hit with block modelling as there may be some "gaps" between objects that will be taken into account when calculating light bounces. I'm not sure, but I'd think that any program - to include Maxwell - also has to take into account 'internal' surfaces - which could also slow things down.

I wouldn't go as far as saying that face modelling is dead, as there are still some advantages it has over block modelling. However, with judicious use of booleans and prior planning with texturing and map channels, you can easily make you block objects as "efficient" as face models. I couldn't say that this would drastically improve rendering times, but if you are going to be using the geometry for a lot of renderings (for an animation for example), then every little bit counts. If it's just for "one-off" scenes, then I'd stick with the primitive objects for efficiencies sake (modelling efficiency as opposed to rendering efficiency). I guess it all depends on which is more time consuming - Modelling or Rendering........

[Frances]

Tell me you guys don't load up the entire model for interior shots.

[Tim Ellis]

Heh, tell them I haven't rendered a car model with less than 1,000,000 faces in my whole 4 year modeling history.

Current Enzo project is 1.25 million faces before it's in the building scene.
Pimpala project is just under 2.5 million faces.

Radiosity & high face counts will kill most computers at render, if not before at the radiosity calculation point.

If face modeling is dead then I'm flogging a dead horse. I can't model any other way to the same detail & control I can with face modeling.

Face modeling isn't dead, radiosity & scanline renderers are.

Tim.

[misterasset]

Yes, it's single shots, maybe about 6 different views. And I understand that "every little bit counts," but for me face modeling is slower because you have to think about the fact that in Scanline (not in Maxwell) light comes through the negative side of the face. To me a solid is perfect, you build the room, it's SOLID!!! So you may have more faces, there's no worry about light going in the wrong spot. And it's not enough to make a huge difference. When he makes the model efficient he's going to waste several hours taking it from 300,000 to 200,000 and still be stuck with a 900,000 face model instead of 1,000,000. He's still going to exclude the electronics and furniture anyways, what's the point. Scanline can handle 300,000.

And yes, we do have to load up the entire model. The space is meant to be "transparent" to the customers. The client wants to be able to walk clients in and see all their employees working and consulting and multi-tasking no matter where they are in the building. Every room has basically a glass wall looking onto the central command area.

It's just so fustrating working with someone who is fantastic at a piece of software that isn't the cutting edge anymore and he refuses to let go. I don't want to be taught to be proficient at something that's 5 years out of date.

[misterasset]

Tim, your post came in after my rant.

I guess my point is, if he would just use a "modern" rendering program and get out of the past this wouldn't be an issue. Sorry to use the forum as my own personal place to vent/rant, but I definately can't do it at work to him.

[Tim Ellis]

That's what we're all here for, in our nice little community.

Is he a boss, or one of your employees?

Perhaps it's time to get someone who can work with you at the same level and with the same set of tools as you.

Tim.

[michaelplogue]

It's just so fustrating working with someone who is fantastic at a piece of software that isn't the cutting edge anymore and he refuses to let go. I don't want to be taught to be proficient at something that's 5 years out of date.

Don't feel too bad. This is becoming more prevelant in the computer industry.

We still have senior System Administrators who got their start using Unix, and haven't a clue on how to administer even MS NT servers.

We also use a piece of software in the government that's been around quite a while. And while it's been regularly updated, they've still been using the programming language it was originaly written in (Fortran or Cobol, cant remember which). They've kept hiring top young programers with knowledge of the newest languages, but they have to basically re-learn the older language. Since the original language didn't make use of dll's or any other external libraries, everything was included in the single executable file. It's finally gotten to the point that it's so bloated, they finally have rebuild it from scratch using a newer programming language.

[Frances]

And yes, we do have to load up the entire model. The space is meant to be "transparent" to the customers. The client wants to be able to walk clients in and see all their employees working and consulting and multi-tasking no matter where they are in the building. Every room has basically a glass wall looking onto the central command area.

Looks like it's unavoidable.

It's just so fustrating working with someone who is fantastic at a piece of software that isn't the cutting edge anymore and he refuses to let go. I don't want to be taught to be proficient at something that's 5 years out of date.

Bad modeling yields bad renders no matter what engine you use. For light leaks with radiosity, I would suggest solid slabs for walls and ceilings that are booleaned together. Also decrease the shadow bias for your lights to something like 0.01. Never use "Force 2-sided" for objects or 2-sided materials - they seriously slow things down. It's okay to use 2-sided shadows though. With Max 7 and up, radiosity has adaptive subdivision. When used properly, you get good detail where you need it and less where you don't. The mesh is what causes the resource hit.

[michaelplogue]

The space is meant to be "transparent" to the customers. The client wants to be able to walk clients in and see all their employees working and consulting and multi-tasking no matter where they are in the building. Every room has basically a glass wall looking onto the central command area.

Gads! I'd hate to have to work there! Nothing like being under the constant eye of micro-managing supervisors with a god complex......

[Frances]

...micro-managing supervisors with a god complex......

You've just described my 2-year-old.

[michaelplogue]

...micro-managing supervisors with a god complex......

You've just described my 2-year-old.

Then he must work in my office....

[misterasset]

Ha, I'll have to show the space once it's finished. Don't know when I'll be allowed to show it, maybe after they see the renderings, maybe after it's built. If that's the case it'll be a year or two until I can show it.

How do y'all handle presenting renderings to clients with the timeframe of Maxwell? My boss's (he's the one I'm arguing with) biggest problem with Maxwell is the time it takes to create a pictue. Now I haven't had time to test the new network rendering (I hope it is as nice as they say it is) but I've brought in extra machines to try to help out with the timing. His big thing is that if we do 6 views and the client makes a small change we can't turn around and give them new renderings right away like you can with a stored solution in Radiosity. One time to calculate, 6 pictures. Maxwell is 6 times to calculate 6 pictures.

I really want our office to start using the best out there, but I have to be able to shoot down his arguments if I'm going to be able to do that. In my head it already makes sense but I need people with more industry experience backing me up.

[michaelplogue]

Just tell your boss there's two choices: Quality or Speed.

There's room for both if he plays it right. Speed for test renders to show the clients during the initial phases of the work, and then the quality "money shots" for the final output.

He can even set up a different price scale for each type, depending on the client's needs.

Being able to do both can only help the business. The Maxwell renders will draw in new clients. Once you've 'hooked' them, you can explain to them that the M~R images are a bit more expensive and take longer to produce, but are well worth it.

It's all about strategy. If your boss has any business savy, he can really exploit this to his advantage.

[Tim Ellis]

This is why RS0 is included in Maxwell, so fast less accurate renders can be shown over the slower higher quality renders of RS1.

At least that's the theory.

Tim.