Tuesday, December 16, 2008

RT-DAT: Noble technological failure

Many years ago, at the turn of the millennium, I was highly excited about a little game from Shiny Entertainment known as Messiah. The player took on the role of an adorable angel known as Bob, sent down from the heavens to clean up a grungy cyberpunk Earth. Bob was able to jump into the bodies of every human he could find, allowing him control of their bodies until they died. Messiah had been heavily hyped a couple of years earlier for a technology known as RT-DAT, which stands for Real-Time Deformation and Tesselation. Essentially, the character models’ polygon counts would adjust in real time to keep the frame rate consistent. Dozens of models on the screen? Just lower the poly count and the game will be fully playable with no slowdown. What’s more, users with high end computers would potentially be able to view character models with bajillions of polygons, comparable to the models you see in animated films. Shiny’s former president David Perry was adamant that the technology was going to be huge.

Unfortunately, that didn’t happen. To my knowledge, the only games to make use of Shiny’s tech were Messiah, Sacrifice, and VR Baseball 2000. So what went wrong here?

Well, first off, the tech just didn’t work as well as it should have—in Messiah, anyways, which was supposed to be the flagship title for RT-DAT (though technically VR Baseball came first). I haven’t played Sacrifice, which was released later in the same year, so I can’t say if things improved much there. But looking at Messiah’s character models up close revealed a blobby mess. At times, it was actually a pulsating blobby mess. Because the poly counts were being adjusted in real time, the characters’ bodies sometimes appeared to bulge and contort in distractingly weird ways. When they weren't doing this, the polygons would often settle into a shape that just barely resembled a human being. Take a look at this classy young lady’s body.



Now check out this scientist's arm, body, and left side of his head. Ech.



Now, remember the effect that water had on Gizmo from Gremlins’s body, minus the furballs popping out? At times, that’s kinda how the characters looked in motion. And as for the potential of seeing super-high-poly models? Try the game on a modern machine. Seriously, try it, it’s a fun game, Good Old Games has it downloadable for $5.99. But don’t expect the polygon models to look any better than they did on a decent system when the game was released.

Speaking of which, the technology was introduced in mid 1997, but Messiah, originally planned as a short project, ended up taking longer. A LOT longer. The game wasn’t released until March of 2000. In those days, a 3-year development cycle was almost unheard of. When Messiah was announced, the age of 3D-accelerated polygonal gaming was in its infancy. This was a time when everyone was oohing and ahhing at Lara Croft’s amazing triangle boobs, so natururally, RT-DAT seemed all the more amazing, since there was the potential to see super-high-poly boobs, provided there weren’t too many sets of boobs on the screen at once. In 2000, engines like Quake 3’s were upping the ante considerably for poly counts, and running well on a wide range of machines to boot. Guess what else was released around the same time as Messiah? Yep, the original Serious Sam tech demo, where dozens of enemies would rush the player at the same time. Good-looking models and high frame rates.. so why was RT-DAT necessary again? Really, I think gamers would rather take a frame rate hit—which seemingly ends up being fairly small in well-coded titles—than witness characters warping and bending in front of them. Furthermore, the plethora of other options that PC players have always had to adjust their graphical settings make RT-DAT seem even more like an oddity that didn’t really need to exist.

Maybe the technology would have been more widely used if it had been better implemented, or if the timing had been better, or if Messiah had made better use of it by actually including areas with more than a scant number of onscreen characters. Scaleable 3D models are still used today. For example, low-poly assets were used in a game that I worked on, SimCity Societies. When the camera got to be a certain distance from a building, it would switch to its low-poly version. I’m sure a lot of other games use this method as well. But not since Messiah I haven’t seen anything like the real-time deformation and tessellation that was supposed to be the next big thing in game technology. The technology and the games in which it was used have been relegated to obscurity. And maybe that was a good thing, to see RT-DAT go the way of voxels. Ah, voxels, that’s another topic I’m going to have to write about…

No comments: