Rear View Mirror

Rear View Mirror

By David Jefferies

January 27th 2010 at 2:55PM

Black Rock's technical directors gets looks over his shoulder

Three and a half years is a long time in video game development. During that time our studio completed the transition to the new generation of consoles, released three titles, joined Disney and turned into Black Rock Studio. Usually when I write this column, I choose a topic that’s currently the focus of much discussion in the studio.

Back in the summer of 2006 we had recently finished MotoGP’06 for the Xbox360. It was our fourth MotoGP game and the first for the new generation of consoles that had launched a few months earlier. We invested most of our efforts in a complete graphics overhaul. Like the other development teams working on titles near the console launch, we had to make bets about what graphic technology would define the new generation without the luxury of seeing what had worked for those that had gone before.

We eventually settled on upgrading our lighting engine, which I documented in this column with articles on Lightscattering, High Dynamic Range, Bloom and Shadows. We also encountered another problem that would become endemic on this hardware – that of data size explosion. Halfway through the project our ‘nightly’ builds were actually taking all night and most of the next day, and our 100Mbit network that had served us well up until then started to creak under the pressure of all the data.

This prompted an article in this column about Distributed Data Builds, describing how we refactored our build system into a job manager that distributed the build jobs to as many machines as it had available.

There was also an article about Scalability, discussing how the use of higher-order surfaces and texture scaling meant we could ‘dial’ the complexity of the data set up or down. This was particularly important with early hardware that kept evolving with no-one knowing what the final capabilities would be.

ROCK ON
In October 2006, Disney bought our studio, Climax Racing, and we renamed ourselves Black Rock Studio. Disney had big plans for investing in the studio and they wanted our ambition to grow with it. While our team carried on working on MotoGP’07, elsewhere in the studio Pure was being developed.

Team sizes started growing and we changed our team structure – so instead of being partitioned into art, code and design we formed teams consisting of all three disciplines that would be responsible for entire features. At the time these articles were all about how to deal with big teams.

Then in the summer of 2007, after we had launched MotoGP’07, we embarked on our most ambitious project to date, Split/Second. The challenge was to create a new toolchain for generating the assets for the action sequences in the game. The continuing trend of asset creation being the bottleneck in games production can be partially addressed by more powerful, streamlined tools but the trend will surely continue.

Split/Second has extremely ambitious visuals which prompted another render engine overhaul. I’ve been writing about our experiences over the last six months with articles on everything from Deferred Shading to Colour Banding, to Stereoscopy. Gameplay might be king, but many of the challenges of this industry still revolve around the asset generation and rendering required for high fidelity graphical images.

Looking to the future, we’re trying to mitigate the main technical problem that keeps coming back at us – data size. We had one crack at this a couple of years ago when we wrote our distributed build system but, since, our export times have continued to climb and our source assets just get bigger. Some of this can be solved by hardware – the Split/Second server farm has 256 processor cores and a terabyte of on-board RAM – but this doesn’t help the iteration time of our artists and designers.

To improve these times we’re currently working on making all of our pipelines live update friendly. Hit save in the tool and, in the background, the asset will be saved, exported, deployed to the console and appear in-game without having to restart the level. It’s not rocket science, but it’s something we’ve never had before. It is a big job for us, so we’re starting with some of our simpler pipelines such as Colour Cubes and then expanding from there until it works for every asset type.

TIME TO SPLIT
While our Core Technology team is making that happen the Split/Second team is working on shipping the title. With just a few months to go, we are in a critical period. I need to concentrate all of my efforts on helping make it our best game ever, so unfortunately this is going to the last article I write.

It has been a pleasure writing them, thanks for reading them, and good luck developing video games.