Intel® GPA Tools Help Zombie Studios Pioneer the Free-to-Play Gaming Space

By Nancy Nicolaisen

 

 

Download Article

Download Intel® GPA Tools Help Zombie Studios Pioneer the Free-to-Play Gaming Space [PDF 1.5MB]

In technology, there are innovators and there are pioneers. Innovators see relationships, connect the dots, and leverage opportunities. Pioneers create opportunity from the raw material of experience and insight. Right now, the first-person shooter gaming niche is experiencing an explosive wave of innovation fueled by improvement in parallel processing technology and leaps forward in graphics processors. It takes something special to be a pioneer of the genre-like, for example, someone whose game play involves actual guns, real helicopters, and live ammunition. Such a company is Zombie Studios, developers of the new free-to-play title Blacklight: Retribution.

Zombie Studios Lead Developer Chance Lyon has been with the company for six years and knows something about battle simulation gaming.

"Both our founders were from high-level military backgrounds. They set out to build a company based on developing virtual reality tools. That eventually that led them into video games. Our company's Serious Games division still creates simulations, virtual reality tools, and a mix of both. We do projects where soldiers train with live ammunition. We detect the trajectory of a shot and translate that into high definition virtual reality environments that respond appropriately. It's not our primary business, but it's definitely still a big part of what we do and who we are at Zombie."

Given that heritage, it's not surprising that when Zombie Studios made the leap into the consumer gaming space, they wanted to move the boundaries of player experience and raise the performance bar for themselves.

Blacklight: Retribution is the second iteration of the Blacklight series, and with this game, Zombie is breaking new ground-at least for the North American market space. Backlight: Retribution is being released as free-to-play content aimed at PC-based gamers. Zombie's entire revenue stream on Blacklight: Retribution proceeds from in-game microtransactions, where players spend small amounts of money to gain assets, capabilities, or amenities that make gameplay more fun and individual. In the free-to-play space, engagement and immersion are absolute requirements for success. "Essentially, we game developers have to compete for customers' time and attention before we have an opportunity to compete for their money," says Lyon.

To create the most irresistible user experience possible, Lyon's small, tightly focused team set aggressive goals for themselves with this release of the game. They used Intel® Graphics Performance Analyzers (Intel® GPA) to get quick prototype feedback and perform data-driven analysis, helping artists and programmers to quickly make informed design decisions. Here's an up-close look at how Lyons and his team used Intel® GPA to accelerate Zombie's development cycle for Blacklight: Retribution.

 

 

 

 

 

Boosting Productivity and Compressing Development Timelines with Intel® GPA Tools

The Zombie team iteratively optimizes the game throughout development, using Intel® GPA metrics reporting tools as in-game overlays. The Intel® GPA System Analyzer Heads-up Display (HUD) gave Lyons' team a dynamic, playtime depiction of the high-level performance characteristics of Blacklight: Retribution. Lyons captured the screen image shown in Figure 1 while analyzing play sequences that he suspected had the potential to become GPU bottlenecks.


Figure 1. The Intel® GPA overlay, shown here running in-game, gave Lyons and his team real-time feedback on the performance characteristics of Blacklight: Retribution during test plays and game development.

A key concern for the team was understanding how well their game played on less powerful platforms used by a significant percentage of their target audience. Monitoring Intel® GPA Frame Analyzer rate metrics gave Zombie's developers an empirical way to determine how well Blacklight: Retribution played across a variety of platforms. Their audience confederation goal for the game involved targeting integrated PC graphics cards. To achieve this and still offer fast, immersive play experiences was a fairly aggressive objective. By using the frame rate data gathered during test plays, Zombie engineers could make informed decisions about how and when to scale back Microsoft* DirectX* 11 features like dynamic tessellation and image reflection. For platforms at the lower end, Lyons and team used Intel® GPA data to decide when to trigger options to reduce levels of shader complexity, pull objects out of a scene earlier, lose scene detail, or even reduce data rates to the client.

"I really like using the performance tools Intel has-the [Intel®] GPA. They're a nice suite of both CPU and GPU profiling tools. To some degree, I just get peace of mind knowing that Intel® GPA tools are integrated and ready to use for testing whenever we're having performance problems. It allows me to keep on top of the artwork and really measure how its characteristics impact the game's performance."

Lyons keeps the HUD display available for constant monitoring of the number of primitive draw calls. He feels that drawing primitives are a good measure of just how complex a scene is and are also typically a limiting factor on low-end machines. Lyon believes that this data-driven optimization is key for delivering a good game play experience on low-end platforms, because once draw calls exceed a certain threshold, it doesn't matter how sparse or how simplified the geometry objects. The sheer quantity of drawing primitives can overwhelm a GPU.

 

 

 

 

 

 

Investigating Game Optimization Subtleties Using Intel® GPA Tools

Because Zombie's teams are both small and tightly focused, one of the most difficult things about rapid game development cycles is keeping the work of the artists coordinated with that of the software engineers. Lyons and his teams found that gathering Intel® GPA analysis data fosters better communication of technical ideas and requirements. This communication, in turn, helps artists and developers work together to quickly craft solutions to complicated problems. Intel® GPA provides the diverse teams with real-time, empirical evidence that establishes the exact cost of artwork strategies and choices. This information results in effective, targeted optimizations that can be easily and directly validated by a second pass of the Intel® GPA tools.

Many of these kinds of improvements simply couldn't be effected by "guessing" where art-related slowdowns occur, because not all problem areas are so clear-cut that they identify themselves. "Very rarely is there one thing you can optimize and get a large performance gain. It's usually trying to optimize ten things by 5 percent to get an overall gain," says Lyons. Because Blacklight: Retribution is GPU bound, the bulk of Lyons' optimization time is spent interacting with artists. The teams have collaborated to produce big successes by identifying problem assets using the Intel® GPA Frame Analyzer.

In one such instance, Lyons used Intel® GPA data to look at rendering of the ground in several game levels. The ground was comprised of an aggregate of fairly complicated terrain pieces:

"Using [Intel®] GPA Frame Analyzer tools, we were able to clearly define the source of the slowdowns. We discovered that the entire level's terrain was being rendered in a single draw call. This redraw was happening no matter where the player was. Once you included the [DirectX*] 11 tessellation on it, it took 300,000 triangles to draw one piece of terrain. Given solid data about the nature of the problem, we could justify doing the significant amount work it required to cut up the terrain and devise a strategy to draw it iteratively, in a smaller portions."

Intel® GPA data helped the Zombie team both to discover a subtle problem and to create an innovative and effective remediation for it (see Figure 2).


Figure 2. The Zombie team used the Intel® GPA Frame Analyzer to prove that a single relatively unimportant scene element-which they dubbed the "expensive tree"-was using an unnecessarily large amount of the overall scene budget. This kind of data helps developers and artists work together to craft effective solutions to art-related performance problems.

 

 

 

 

 

 

Investigating Outright Mysteries Using Intel® GPA Tools

Every veteran game developer has two or three things he or she checks out as a matter of course when frame rates go off a cliff. However, not too many would think of looking into menu drawing. When fellow development team members came to Lyons and reported that the game's main menu was rendering about 10,000 polys but running at only about 15 frames per second on the team's Intel® Sandy Bridge system, all knew that performance degradation of that sort on a high-end machine predicted real trouble for typical gamers. There was no obvious explanation for this extremely poor frame rate, because other than the menu, the scene consisted of a single character and a fairly simple background.

Lyons used Intel® GPA Frame Analyzer to drill into GPU data and isolate exactly what was taking up all of the rendering time. Much to the engineers' surprise, he discovered that the seemingly innocuous-looking background was actually a complex composition that relied on several layers of full-screen blur, all going on at the same time (see Figure 3).


Figure 3. The Zombie team had a real mystery on their hands, trying to ascertain why a simple menu draw took 47 ms to render on an Intel® Sandy Bridge system. This was the game's main menu, so it would have been problematic on slower machines typical of what the average player would use.

Lyons was able to make specific, data-based design suggestions as he worked with user interface designers to refactor the entire background for the main menu (see Figure 4). "Frame rates went from basically 15 to a solid 60 immediately," says Lyons.


Figure 4. After in-depth analysis with the Intel® GPA Frame Analyzer, the Zombie team discovered that, though the menu appeared to be a simple graphical element, in reality it required an extraordinary amount of resources to draw. Engineers and artists worked together to refactor menu drawing and achieved a 4-fold increase in frame rate.

 

 

 

 

 

 

Conclusion

Zombie Studios development and artistic teams found innovative ways to use Intel® GPA tools to make their development process more efficient and to keep everyone's efforts coordinated and focused on the ultimate goal: better game play for the end user. The Zombie team's use of Intel® GPA's GPU profiling and performance statistics shows how real-world developers can harness the power of Intel® GPA to speed identification of performance issues and foster better communication about the ramifications of graphic design decisions. Ultimately, using the Intel® GPA tools made it possible to optimize early and iteratively. As a result, games run faster, delivery cycles are compressed, and there are no late-stage surprises-a big win for developers, artists, and gamers.

 

 

 

 

 

 

For More Information

 

 

 

 

About the Author

Nancy Nicolaisen is the author of numerous books on software engineering techniques. She specializes in the design and development of solutions for small mobile and embedded systems. Her involvement with the game industry dates back to 1981, when she worked at gaming pioneer Imagic, developer of Demon Attack and other classics.
 

有关编译器优化的更完整信息,请参阅优化通知