UMBC’s got GAIM » Blog Archive » Game developers on academic research

Game developers on academic research

March 17th, 2008 by Marc Olano

In the past month, I’ve had the pleasure of attending two conferences bringing together academics and the games industry. The first was I3D, the Symposium on Interactive 3D Graphics and Games, sponsored by ACM SIGGRAPH. The second was GSCSE, or Game Development in Computer Science Education. GDCSE was formerly the Microsoft Academic Days in Gaming Conference, but this year’s was co-sponsored by EA and held in cooperation with ACM SIGCSE. One of the most interesting things to come out of both was the perspectives of games industry veterans on the roles of university research in the games industry. Here are a few of my notes, from most researchy (I3D) to most industry focused (GDCSE).

Planning game-targetted research

Jason Mitchell, Valve (I3D)

Believes academic research is best focused further out (5-ish years). Game companies cannot take the long view of technology, somebody has to.
Consider memory as well as computational requirements.
Scalability: many academic papers work great (for one object, seen alone)
To really understand the limitations, intern at a game studio
Helps if you can develop research in an existing game engine
Some areas where research is needed

Games use shaders in zillions of permutations, how can you better develop and manage them?
How can data amplification limit memory use or memory bandwidth in exchange for computation?

Andrew Willmott, Maxis (I3D)

Maxis does sandbox games, so more user-generated content introduces new problems
Run-time generation of all kinds of data and acceleration structures
texture atlases, physics hulls, geometric levels of detail, impostors.

Naty Hoffman, Sony (I3D)

Games don’t always want physical correctness, but improving physically-based methods is an important starting point because they can be more robust.

Research tied to a game

Hao Chen, Bungie (I3D)

MSR Asia developed some research tech specifically for Halo 3. For something this closely tied to a shipping game, it needed to be a very well defined problem, and they needed a good fallback plan. The game couldn’t be late because the research didn’t work out as planned.
In addition, they had several 6-8 week on-site researcher visits, to ensure that the researchers understood the needs and limits of the game.

Dennis Wixon, Microsoft Game Studio (GDCSE)

Usability testing in Forza and Halo 3, make sure game isn’t too hard or too easy

What the industry is looking for

John Cash, Blizzard (GDCSE)

Need: algorithms, data structures, c++, algebra, verbal & written communication skills
Want: multiprocessing, trigonometry, linear algebra, calculus
Like: have a specialty and passion (but realize, when you start, you’re not going to be “the graphics guy” or “the physics guy”, you’ll be “the new guy”)

Colleen McCreary, EA (GDCSE)

CS courses: Data Structures/Algorithms, Discrete Math, Linear Algebra, Operating Systems, Graphics, Artificial Intelligence, Low-level programming, Physics, Parallel Processing, Networking, Testing
Useful: Experience with large code bases, Open ended problem solving, debugging, design and engineering, full-cycle projects, complex application development, changing last minute parameters

Posted in CS, graphics, research, technology |

TrackBack URI

Game developers on academic research

Leave a Reply

UMBC

Other blogs