LightWave 3D 8.5 x64

Author: Jose “Kalel” Perez & Carlos “Guru” Echenique Reviewed on: January 20 2006 Product cost: $795 Spelling and Grammatical Editor: Ian King and Sean May Manufacturer: NewTekNewTek has released the 64-bit version of their premier 3D modelling rendering software. Emmy Award winning VFX artist Jose Perez and I put this package through it’s paces and the results are pretty astounding.

Introduction
Not too long ago, modelling realistic 3D imagery took a rather esoteric mix of math skills, really expensive software and a room full of Crays. These days, 3D modelling and rendering is used in practically every film Hollywood produces and has allowed filmmakers to achieve a level of realism that was unthinkable 10 years ago. Impossible camera angles, ultrarealistic creatures and effects, and superhuman stunt work are all de rigeur for the 3D VFX artist.

NewTek started out on the Amiga platform with their Video Toaster software. They later branched out into 3D with LightWave. When WIndows NT was released, LightWave was ported to that platform (and others) and then things really began to take off.

We had the opportunity to take LightWave 3D 8.5 x64 out for a spin and put it through its paces. To assist me in this little romp, I contacted an old friend (and Emmy Award winning VFX artist), Jose Perez, to really test it thoroughly. In fact, the 3D models used in recent episodes of Stargate SG-1, Stargate Atlantis, and Battlestar Galactica  were all textured and test rendered (approval shots) in LightWave3D 8.5 x64 on a 64-bit box provided to Jose by PlanetX64.

Description
The latest version of LightWave (8.5) is a free ugrade to registered users. However, this release is not short of new features and enhancements to existing workflow. To see a list of these, you should visit NewTek’s website . For this review, we focused heavily on the 64 bit aspects and speed improvements.

Benefits of 64-bit vs. 32-bit
Jose: As an Artist, I have no real desire to delve into the technical aspects of how things work in the background of a 3D application. In this case however, what goes on in the background in 64-bit vs 32-bit is astounding for people in the Visual FX industry. Among the most notable workflow improvements is the rendering (mathematical computations of lightrays bouncing off of 3D geometry to make up a CG image). Speed is improved dramatically. Each scene is different, but I have seen improvements of 2x, 3x,4x, and even higher in some cases. Needless to say, if a frame under 32-bit takes you 1 hour to render, and in a 64-bit environment the same frame renders in 15 minutes, the results speak for themselves. Time, after all, is money.

A silent, more notable advent in this arena, is the ability to address memory of 4 Gigs and upward. To the average home user this might not seem like something drastic. But if your system can adress 16 gigs or 32 gigs of ram, or higher, not only can more complex scenes be loaded into memory, But the response of the workstation seems more realtime as well. Hence more power to the average pixel pusher.

Carlos: As a Gearfreak, I respect NewTek’s decison to bring LightWave out as a fully 64-bit application. The enlarged memory address space and larger registers allow for the complex mathematics of raytracing to be carried out at speeds unheard of as little as a year ago. Renderfarms (large groups of computers used to crank out 3D animations) composed of 64-bit machines will provide higher quality images in a fraction of the time and cost. This is truly desktop supercomputing.

Tests
To test this new technology, I decided to throw the most complex CG elements I was working on at the time, which happened to be complex spaceships and entirely CG constructed cities, from shows like Stargate SG-1, Stargate Atlantis and Battlestar Galactica. For an added comparison, I loaded up my Mini Cooper Robot from the the viral Ad Mini Cooper/BMW had me build for Zoic Studios.

Aside from the robot, which was completely modelled and textured in an earlier version of LightWave, the rest of the elements were all textured, lit, and rendered in LightWave 8.5 64 bit (beta at the time).

Test Rig Specs
Jose’s Rig:
Dual Opteron 850’s
OCZ 4GB DDR400 ECC Reg SDRAM
Tyan Tiger K8W Motherboard
Custom CM Wavemater case from Performance-PCs.com
LSI MegaRAID 150 SATA controller
3x 250GB Seagate Barracuda 7200.8 SATA HD’s
3DLabs Wildcat Realizm 200 video card
Windows XP Professional x64

Carlos’ Rig:
Dual Opteron 875’s
OCZ 8GB DDR400 ECC Reg SDRAM
Tyan Thunder K8WE
Custom CM Stacker
Seagate Barracuda  7200.8 400GB
Dual eVGA 6800 Ultra Video cards in SLI
Windows XP Professional x64

Carlos’ rig was used for renderfarm style tests while Jose’s was used for modelling, design and test renders.

Analysis

Jose: Lets begin with the Battlestar Pegasus from the second season of Battlestar Galactica.  After I completed the model in its full, high resolution assembly, the polygon count was at around 3.1 million. Of course this was including fully detailed guns and other details to handle closups.  Normally we don’t load the full model up for shots, only the neccessary pieces. That being said, she is bigger than the Galactica in physical size, but not in polygons. Galactica herself has more detail.  However 3.1 million poly’s in 32-bit Lightwave is nothing to sneeze at either.

I textured the Pegasus and ran test renders in the 64-bit version of LightWave, for approval of the look and feel of the ship.  My first impression was “This workstation handles the geometry in OpenGL much faster than in 32-bit”. So this was a pleasant surprise, as I found myself spinning the ship and and checking textures at almost realtime, compared to a choppy slow movement in the 32-bit version. Hence I was more productive and found myself getting things done faster on a ship of this magnitude.

The ringer was when I started to run test renders with the ship.  In 32-bit LightWave, a full frame shot was rendering in about 30 minutes, with a full light kit in there. In 64-bit however,  the same identical frame flew by at around 7 minutes, give or take a few seconds – this was jaw dropping. For a single Freelance artist like myself, to wait around 30 minutes for a test render twiddling thumbs, as opposed to 7 minutes, just enough time to go get a cup of coffee and move on.

Now I’m no math genius, nor would I pretend to know what fraction of time that shaved off.  But from 30 minutes to 7 minutes for the same exact High definition frame with a decent amount of antialiasing, I would say that’s a vast improvement.  Granted, NewTek has also sped up LightWave’s rendering technology. This 64-bit workflow just lends itself to so many more possibilities from an artist standpoint, not to mention a studio’s deadlines on high profile projects. One machine is impressive, a farm of 250 or more 64-bit nodes is just incredible.

Carlos: Fortunately for Jose, my kung fu (in mathematics) is much stronger than his. His test renders with the Battlestar Pegasus were clocking in at just over 300% performance increase (328.6% to be exact). The 64-bit platform’s ability to manipulate very large numbers and access to far larger amounts of storage vs. 32-bit are what gives it the edge.

For my next tests, I loaded up a few models I had just modelled for Stargate SG1 & Atlantis.  They included the new Daedalus battlecruiser , new F302 fighters, and a new version of the City of Atlantis which I modelled for Season 2.  Again Texturing them in 64-bit Open GL was a joy, nothing held me back.  Just some stats here to give you an idea of how these models fared. The Hero version of the Daedalus weighed in at around 1.3 million Polygons without the full rez guns in there. Guns added another 300-400k polygons. Once done texturing, the test renders began for approvals from the studio.  On my 32-bit workstation, the Daedalus was running 14-15 minutes for a full frame, high definition shot.  When the same frame was tested on the 64-bit box, it came in at around 4-5 minutes. These tests can vary a little since different models use different amounts and sizes of  texture maps, but over all, I’m seeing a boost in all rendering, making the performance alone worth upgrading for. Again, time is money.

Carlos: That’s 200 – 250% speed increase for those of you keeping score.

Carlos: That’s a 500% speed increase, ladies and gents!
A much smaller model, the F302 fighters that are housed inside the Daedalus, weighed in at around 200-300k polygons. Not a big deal in 32-bit Open GL. But once the rendering begins, it is apparent once more that the new render speeds alone blow the 32-bit box out of the water.  The 32-bit box rendered fighters at 1.5-2 minutes depending on the angle and lighting. Whereas the same frames on the 64-bit box rendered at 15-20 seconds.

I followed up with my last model on Atlantis. This was a HUGE one – well it sure felt huge to me, in scale at least .  A new highly detalied version of the city of Atlantis for season 2 on the Sci-fi Channel. Total count with all medium rez buildings and piers loaded in was 2 million Polygons.  Textured it in 64-bit Layout, and threw a test light rig in there for some nice test renders.  On the 32-bit machine, the city rendered in about 20-25 minutes with fog, and BG radiosity in there, also medium antialising.  On the 64-bit box, the same High Definition frame came in at a whopping 7 minutes.  That sealed the deal for me. 64-bit LightWave is the way to go if you can afford the hardware upgrade, and the memory bump up.  Fortunately, Planet X64 took care of that for me.

Carlos: I’d like to thank Tyan, AMD, 3DLabs (Creative Labs) & eVGA for their participation here as well. Oh, and by the way, 257% speed increase.

Before I packed it up and said I’m done testing, I figured I’d load up an oldie of mine. My Mini Cooper from the viral ad campaign BMW ran a few years ago.  I remember specifically having a difficult time spinning the model in OpenGL, and handling the full model in Modeller. Why?  Well, 2 years ago 1.2 million polygons was a bit high for 3D apps to handle, and with less memory in your system – well you can imagine the rest.  Especially running LightWave 7 or 7.5.  The model was chock full of details, down to the engine parts, and a full dash under the chest, which was covered by the famous Mini Cooper hood.  It had to hold up in all sorts of mediums from magazine ads, to web content, to full animated video, and even a 37 foot tall billboard in Times Square, New York.

I can’t say I was surprised anymore, given the results I had just experienced with the other, bigger models, but a pleasant smile came over me when I loaded the Mini up in 64-bit, and just handled it, smoothly and easily. Finally the big test. I remember rendering all sorts of huge odd sizes and render formats for printable versions of this guy. So much so, that renders at one point were taking around 12 hours with full area lighting, assisted radiosity, all the bells and whistles. So I cranked up one of the big 1.2 million polygon scenes, set with all the good lighting, and full details, at around 6k rez. I had marked this at around 12.5 hours on my old box.  The new 32-bit box did it at 7 hours. On the 64-bit box, the same huge frame rendered in 55 minutes, just a little under 1 hour.

Carlos: I loaded up the same scene on my quad core Opteron system to simulate renderfarm performance. LightWave 3D x64 is multithreaded from the ground up and does take advantage of multiple CPU’s with only a couple of setting changes in the scene file. The quad core system shaved the render down to 32 minutes, 16 seconds. Here’s a table of the speed increases:

Machine Type	Render Time	Delta
Original 32-bit worksation1	12.5 hrs	–
Current 32-bit workstation2	7 hrs	78.5% performance increase
64-bit SMP workstation (Jose’s rig)	0.92 hrs	1,258.7% performance increase
64-bit quad core workstation (Carlos’ rig)	0.5375 hrs	2,225.6% performance increase

1) The original 32-bit machine used for the Mini Cooper Robot had the following specs (as best as Jose can remember):

• Dual Athlon MP 2000+
• 4GB RAM
• GeForce4 Video
• 500GB HD
  

2) Jose’s current 32-bit workstation:

• P4 3.02 Ghz (single CPU)
• Intel 875p Chipset from MSI motherboard
• 2 GB Ram
• 500Gb HD Sata drives
• Nvidia Geforce Ultra 6800

Conclusion
Jose: Performance, productivity, & workflow. Those are the keys to doing good work, on time and on budget. Sure there are many new features in NewTek’s latest version of Lightwave, and I urge you to check out their site (www.newtek.com ), but to me as an artist, the workflow and performance boost is what this is all about.  And Newtek delivers as always in a field constantly driven to do much more, and look better, in a shorter time frame. I definitely recommend the upgrade.

Carlos: NewTek has done an amazing job in porting LightWave3D to 64-bit. The application truly shines and is a showcase for the benefits of the platform. LightWave3D x64 + multicore workstations are the way to go if you want to make 3D graphics happen.