THE REVOLUTION WE
TOOK FOR GRANTED: HOW
PHYSX CHANGED GAMING
B
ack in the late 1990s, the first wave of 3D
accelerators made proper 3D real-time
rendering viable at last. And the Pentiums
and Athlons of the time had become powerful
enough to handle basic physical effects too,
such as some in-game objects falling under
gravity or bouncing off walls. It was an age
when the idea of a game being “photorealistic”
some day, was seriously looking less like
science-fiction, and more like an inevitability.
Still, it was early days. Very early days.
Quake was the first proper 3D rendered
game, and even without a 3D accelerator, it
could bounce a grenade off solid surfaces,
and have an LPB’s rocket exhaust light up the
environment as it hurtled toward your face.
But those effects were extremely limited.
When you killed a monster, and it collapsed
onto the ground, that wasn’t the game
dynamically calculating how the monster would
fall. It was a pre-packaged animation.
In the quest for “better graphics”, developers
eventually came up with the idea of using
procedural animation, generated according to
the rules of a physics engine. One of the first
games to showcase this... uh, in a certain sense
of the word... was Jurassic Park: Trespasser
back in 1998. It had so-called ragdoll physics,
where enemies were made of a collection of
“hard body” objects jointed together. When an
enemy was killed, the physics engine would
calculate how the poor dino fell into a gracelessThe story of the fate of PhysX is hard to frame. Is it about a plucky little engineering company going
up against the might of the big GPU brands, and forcing them to get with the times? Or is it about
the all-consuming beast that is NVIDIA, suppressing innovation to maintain its market dominance?
The truth, as always, lies somewhere in the middle.W OBJECTS IN SPACE
starting to stress even the most powerful CPUs.
It’s not like the maths is hard, necessarily, it’s just
that there’s a lot of it going on.
Eventually Havok came up with the idea of
writing a new engine that could harness the
power of a GPU to do all this calculation with
hardware acceleration.
Unfortunately for Havok, their timing was off.
The idea was sound: use the GPU to do physics
too. But they missed their window. Or perhaps
they were focused on maintaining a viable
business and didn’t want to burn too much cash
on R&D.
In any case, there was a new company on the
scene, with a new kind of engine, and they were
about to change physics simulation so profoundly
that barely ten years later we’ve forgotten alllump on the ground, rather than just play a pre-
animated death sequence.
Sure, Trespasser was kind of terrible as an
actual game, but rag-doll physics became THE
hot graphics tech for the dawning of a
new millennium.
Then, in 2003, Valve announced Half-Life 2
and the Source engine. The E3 demo showed
barrels bouncing and floating “realistically”, a
floppy mattress, and various other examples of
really sophisticated procedural animation. The
game lived up to the demo, and so the future
of gaming set: physics was a thing you had to
have if you were making a top-tier 3D title.CRY HAVOK
Part of what made the Source Engine possible
was Valve’s use of a heavily modified version of
the Havok physics engine. This was - and still is- one of the most popular SDKs for physics, and
the Irish company that created it was snapped
up by Intel in 2007, and then by Microsoft in
- You’ve no-doubt seen the Havok logo on
more than a few splash screens.
But if there’s one thing we can say about
game developers, it’s that they push hardware
to the max. Within a few years, Havok had
figured out how to make its engine simulate all
kinds of physical effects, especially fine-detail
stuff like rippling cloth and realistic smoke.
But procedurally animating a character’s
calculate how the poor dino fell into a graceless cloth shirt in real time in meaningful detail was that barely ten years later we ve forgotten all
about them. Waitt, how does that work?
LET’S GET PHPHYSICAL
Physics engigines competed for licensing dollars bycloth shirt in real time in meaningful detail wasOne of the offshoots of GPU
development: supercomputing!From the olden days.The idea was sound: use the GPU to do
physics too. But they missed their window.