How does incorporating various graphics features come into the design of a game? When starting the design of a game, there are a set of graphics technologies already available - in your opinion, what are the criterias that influence what graphics technologies get included and what gets ignored?
This partially goes back to your previous question about minimum specification and target release platforms and timeframe, but the chosen graphics features depend heavily on the concept of your game as well. For example a technology producing dark, gloomy environments might not be the right one for a kiddies’ game or a flight sim.
Finally, the most significant input on what technologies are chosen comes from the team’s programmers - Carmack being a case in point, but this happens in every project, with every developer.
In your opinion, what makes a game reviewer proclaim that a game is "graphically impressive"? Is this a universally accepted phrase, with a common set of criterias (creative and colorful texture art, high resolution textures, high polygons, as examples) for judging graphics? Or is the phrase "graphically impressive" more subconscious and subjective and differs from individual to individual?
It’s definitely subjective.
In my opinion, the single most important thing is that the looks must be original and fit to the game. That’s why I’ve been extremely happy about Remedy’s radiosity-rendered light maps – it made Max Payne look dark, gritty, realistic, and also unique. That’s also why many developers who license an engine modify it heavily to suit their purposes.
And after you have the technology about right, you of course need the right tools, great artists, art direction, plus good coordination between the content team and the programmers.
Higher expectations from the consumers include better graphics, better storyline, better AI, better sound, better weapon mix... the list goes on. How do you, as a game developer, juggle and piece together a good mix of so many game components and content?
It’s really very simple. You get an excellent team together, nudge people to the right direction, and suddenly you have a good mix of everything. If people are enthusiastic about what they do, the end result simply can’t be bad. I’m very happy to work with so many truly talented people.
We, as a team, also try to learn as much as possible. Sometimes that means intense debates over improving gameplay, deep analysis of other games, learning about marketing or developing our ways of working. The industry is still developing at such a pace that constant learning is a must to survive.
DirectX 9 is more of a evolutionary step up from DirectX 8 with many lauding its focus on higher precisions. John Carmack has said on record that floating point framebuffers should be the way forward, which is another evidence of his wish for higher precision. What would you personally like to see in the future of graphics technology?
Like most game developers, we’ve usually been the user for the APIs and only had minute comments to the specification or development process of e.g. DirectX.
Floating point is definitely the way forward in many cases, and currently the FP usage in DX9 is a bit limited. I’ve heard rumours from researcher-friends that some new algorithms that they’re trying are producing artefacts even with the 24-bit precision, so 32-bit accuracy will then perhaps open up some new opportunities.
Also the “free programmability” is not yet quite free; both floating point and programmability come with performance penalties, heavy in some cases. Cost of every rendered pixel is increasing dramatically. Still, I think the general direction of the development is right, though. We just need to be patient…
With DirectX 9 being as new as it is, it would take some time before games start taking advantages of its newer (than DirectX 8) features. Pixel Shader (PS) and Vertex Shader (VS) 2.0 provides for some really good effects, as can be witnessed in Futuremark's new 3DMark03 "Nature" test scene. As we progress along, what is the likelihood of "game patches" or updates that add in PS and VS 3.0 support?
Well, we would first need to see the hardware naturally. But in any case patching existing content to run with a higher Shader version will not give you much of anything unless that content is originally planned to run e.g. in multiple passes on an older Shader version.
With multiple render targets (MRTs), rather than send the geometry multiple times, it is possible to send the geometry once and store per-pixel the geometry information relevant for rendering that pixel (you store world position, normal, diffuse color, etc). During follow-up lighting passes you only need to send a quad to apply the light effect. The advantage of this technique is that vertex processing is only done once, and light passes are really simple (just quads). For Remedy's current and future game development work, have you considered using MRTs for your rendering technique?
Decisions on this kind of approaches also depend a lot on what you’re trying to achieve with the game, and to my knowledge we haven’t announced any new games or technology... And we aren’t yet developing one either. Our focus is kept tightly on Max Payne 2 now, as a small developer we can’t afford to do many things at once but rather do one thing and do it to the best of our abilities.
Stencil shadows is one of the latest buzzwords, especially so hyped by the upcoming id Software game DOOM3 and is a big determinant in the "immersive environment" factor. Do you have any comments (the pros and the cons) about stencil shadowing? What other shadowing and/or lighting techniques do you think may be better alternatives?
Doom3 looks impressive, but the lighting model is somewhat limited, and that’s not definitely the only way to go forward. With free programmability increasing, the alternatives are almost limitless. Whether the other alternatives will be better, depend of course on what you want to achieve.
We’ve only been throwing around some neat ideas during coffee breaks over here, but focus is on Max2, so I can’t tell you anything concrete on this right now. But I wouldn’t be at all surprised if we’d go a different way than Doom3.
Final question - do you have any comments about the state and quality of video card reviews by various websites? Do you think they have done a good job in providing information to the game-buying public about which is the better video card to buy for playing games? If you were asked to review a video card, would you have a different approach/method compared to the existing ways practised by various hardware websites?
It’s a difficult job, reviewing video cards.
I think too many sites just show tons of results with tons of different benchmarks without actually saying very much. Testing extensively is good, but a graphics card review should be doable in 2 pages, not in 20. And it should say something more than “X is faster than Y as is visible from the graph”.
I think the most important guiding factors are that a reviewer should be aware of what they don’t know (and investigate matters), and he should abide by “the ethics of journalism”.
I'd like to thank Markus for his time and yes, I'll keep 15 questions to 15 questions and not 30 or more next time! :) I'm sure many gamers are waiting in anticipation for Max Payne 2 - let's hope we don't have to wait as long as we did for the original Max Payne!
- Please feel free to comment on this interview here.
