Software vs. Hardware?

[JDgraphX]

Hi, newbie here. I know this topic is still under debate, and perhaps being beat to death, but I must share my opinion and also get other people's...

 

I am preparing to start my own business in Architectural CG, and as a one man show with a limited budget, I can't afford to simply go out and purchase Brazil, Final Render, and VRay, just to see which one works best for me. I will be doing interiors and exteriors, as well as "fly-through" animations. I had an opportunity to try a demo of Brazil, and was very impressed. The speed and quality of Max's scanline renderer aren't even close, as you all well know. I am looking for more insight, from an all-around architectural point of view, which one might be best for me. As in an earlier post, I'm also looking for a comprehensive comparison w/ the same computer, scene, model, etc. between the three mentioned renderers. However, I also wanted to add to the mix... has anyone tried the ART PURE card? It advertises as being far superior in speed, however the only example I've seen of Architecture, is not at all impressive to me. Thanks for your opinions.

[nuno]

u can start with the search utility in this forum. there are some topics regarding your question....

btw, welcome to the comunity

 

nuno

[andymcnair]

Hi and welcome - your best option is to try all the plugins before you buy them and see which one you prefer in terms of ease of use. Theres also a free limited version of Vray and if you speak nicely to Cebas they may offer a time limited copy of FR for testing. They all have their good points and then its down to personal taste. I don't know about the Pure card but a dealer would probably let you see how it works.

[Jennifer]

Hi,

 

I have an article on the PURE card on my site at render.4dArtists.com that should answer a lot of your questions on that product.

[JDgraphX]

Thanks everyone for your input!

 

-Jenni, I did read your article, and it was very informative. I still don't quite understand a couple of things, maybe you can help... My understanding to this point is the PURE card does not calculate radiosity or photon mapping? Also, it seems as though certain scenes will take LONGER to render with the PURE card than with the scanline rederer? GI is important to me because I use it in EVERY scene. In my opinion, it's night and day compared to images with fakeosity. Even on the PURE website, while the manufactured product images look very good, the architectural ones, mostly the interiors, don't even come close to those from Brazil (again I have yet to use fR or Vray). Speed is important, I understand, esp. for animations, however, until someone can show me some architectural images from this product that match up to the quality of the software renderers, personally, I'm not sold.

Hopefully, someone will soon do a thorough test with the same machine, same scene, using the three different softwares as well as the PURE card, and include the render time and details of the setup between each one. This would be EXTREMELY helpful for me and countless others in making this difficult decision.

 

Thanks again, everyone, for your input.

 

-Jim

[Greg Hess]

Hopefully, someone will soon do a thorough test with the same machine, same scene, using the three different softwares as well as the PURE card, and include the render time and details of the setup between each one.

 

I think you'll find that although you can approximate the performance of a specific software renderering package...you can't really tell which one is faster then another.

 

Each software, (or hardware if the PURE card), will have its own set of parameters for optimizing a particular scene.

 

These optimizations can greatly effect the speed of the scene, as well as influence the overall quality. Only an expert within each particular software package can fully utilize a particular package's optimizations for performance.

 

Secondly, a scene rendered as a still can have an entirely different look when its put into an animation. Artifacts which are invisible in a still, can suddenly become extremely apparent when movement is added to the scene.

 

When it comes down to it, usability and features tend to have more of a bearing then raw performance.

 

One of the things that can't be ignored is the fact that most of the software renderers now have support for DR (distributed rendering). This allows for multiple machines to work on the same scene simultanously...even if its only a single frame.

 

http://www.3dluvr.com/crossbow/incoming/benchmarko.jpg

 

http://www.3dluvr.com/crossbow/incoming/vraybench1.jpg

 

 

This is just some shots of one of the vray benchmark scenes I'm using to test the Dual Xeon vs Dual Opteron.

[xgarcia]

Greg - the four buckets are b/c of hyperthreaded dual cpu's right? man...it looks so damn cool. four buckets!

 

Xavier

[JDgraphX]

Greg - nice work. Is that the Sponza Atrium?lol. When you're finished testing those cards, post the results here if you don't mind. Good point about needing to be able to fully understand and optimize each of the renderers to know which one is best for the job, and to achieve satisfactory results. Thanks again.

[Jennifer]

Hi Jim,

 

I'm working on an article discussing max's Advanced Lighting and the PURE card, so have been playing with this a lot.

 

You can use max's radiosity solution when rendering to the PURE card, but it does not use the mesh tessellation feature of the radiosity solution. You need to manually tessellate objects that need additional indirect illumination.

 

PURE does not use Jensen's Photon Mapping technique, which is the reverse of raytracing, where photons are cast from the light rather than a ray being cast from a render pixel. PURE's Selective Secondary Illumination gives results similar to renderings I've seen that used the Monte Carlo method for generating indirect illumination, and can be noisy and produce hotspots in the image if the scene isn't lit properly to begin with.

 

Simple scenes will take longer on the PURE, so if you make a box and compare render times for that, scanline is faster. When your scene gets more complex, and you have reflections, refractions, raytraced lighting, and so forth, the PURE card can plow through that much faster than software.

 

ART VPS has an interesting news article that talks about the speed advantage of using PURE/RenderDrive with max's radiosity: LINK

 

I've downloaded the VRay Free version and started working with that to see what it can do, and render some comparisons. It didn't come with any benchmark scenes, though.

 

If you are looking for Photon Mapping, then VRay and Brazil seem to be the ones to look into. If you are just getting started and have a limited budget, then just using 3ds max R5/R6 will keep you busy learning for quite a while, and can produce nice results. If you store direct and indirect lighting in your scene with Radiosity, it will render pretty fast.

 

VRay, Brazil and PURE all come with their learning curves, on top of those with 3ds or VIZ. Photom Mapping purports to be fast, but I've read articles that say that it just adds another level of complexity on top of the other rendering time. I need to do some tests....

 

Somtimes is not the size of the magic wand; it's the magic in the magician, ya know?

[Greg Hess]

JD,

 

Its not testing cards...but systems. Video cards don't effect rendering times...unless your using Maya 5's GPU renderer.

 

Or you using the PURE system...but thats not really a video card .

 

Yes thats Sponza. Part of my vray tests include those two scenes. (Both the cathedral and the atrium). They make excellent test subjects due to the amount of time it takes to render them at film resolutions.

 

Xgar,

 

Ya its due to Hyperthreading on a Dual Xeon system under Windows XP.

 

I was trying to use it to demonstrate the effects of DR however.

[JDgraphX]

Hi Jenni,

Appreciate the info. I'm getting started w/ my own business, however I've been using Viz4, on and off, for just over a year. I got the chance to do learn it and do a few projects at my current job. I also recently finished demo'ing Brazil, and got somewhat familiar w/ the photon mapping.

 

The PURE card seems very interesting to me for the speed, however if you have to generate the radiosity solution first, that would add to the total render time. Most of the work I've done with Viz, the radiosity takes nearly as long as the scanline render itself.

 

It would be nice to have the card for the speed, and have the 3rd party software renderer's features.

 

Thanks Greg, for clearing that up. I'd still like to hear your thoughts between the two systems? I'll have that decision to make soon, too.

 

Thanks again everyone!!

[Jennifer]

Hi,

 

Any time you add features to your rendering pipeline then you can expect it to add time, also, to the rendering. All things come with cost, some more than others. Photon mapping shows promise of being fast, but you do need to cast enough photons to get a good result. You can't get radiosity, or irradiance, or whatever, without it doing the calculations first, whether you use PURE, VRay or scanline to render. The nice thing about radiosity is that it is stored in the mesh, and it calculates it only once. So an animation doesn't take any extra time, the solution is reused. VRay can store the illumination info, too, if I understand that correctly from their web site.

 

If you are comfortable with Brazil and the techniques needed to get the effect you want, then you may want to stay with that. If you have a scene that you like the look of, then render out a few seconds of animation with Brazil and make sure that it fits your needs time-wise.

 

As Greg mentioned, some renderings may look great as still-images, but animating them can leave artifacts and noise that can only be seen during playback. Usually you can get rid of this noise, but the time-cost to do that may be extreme. What you can do for a beauty-shot is not always what you can afford to do in an animation.

 

And, just because you have a tool doesn't mean you have to use it for everything.