Question 6

Beyond3D : DX8 specifies Version 1.0 and 1.1 Pixel and Vertex Shaders. Do you believe that the small extensions that 1.1 offers are essential?

Croteam : Errr ... sorry, but I'm not familiar with DX8 features (yet!). I'm still knee-deep-in OpenGL. :-)

MadOnion : In vertex shaders, the difference is essential. For example the skinning we use in 3DMark2001 would not be possible without the vertex shader address register. But I don't think we will see any DX8 HW without VS 1.1 support (since the version naming change was done at the last moment; v1.0 is a subset of the original DX8 spec). I don't have an opinion yet on whether PS 1.1 is essential or not since we've played less with pixel shaders than vertex shaders, and everything we've done so far works with 1.0.

NVIDIA : 1.1 pixel shaders are necessary to take advantage of the full hardware shading capabilities of the GeForce3. If you only have 1.0, you'll miss out on some of the best special effects. :-)

Question 7

Beyond3D : There is a big gap between hardware technology being used and being released to the consumer. How do you plan to avoid this issue in future game/app engines? Some kind of extensions for the 3D engine so that new future hardware supported effects can be added quickly?

Croteam : Actually, what I really want is for people to start buying this marketing hype! I'm sick of all these 'technology demos' and 'white papers'. If I should pick one thing that I like in DX8, it would be an attempt to standardize all the stuff that different IHVs have done 'for themselves'. I mean, developers are having really hard time to implement most of today's hardware features, just because IHVs are not cooperating with each other. Every year we got a fair deal of 'new' features without API extensions that actually support it.

MadOnion : For game developers, the problem usually is not adding the support in the technical sense. The problems come from two angles:

1) Content Creation. For example if you want to add good looking bump mapping, you may need to go through all textures in all of your 30 levels in the game.
2) Tech effects on game play. For example if you add shadows to a 1st person game, the AI should react when he sees the player's shadow. This kind of change may then require some changes to the levels to keep the game play fun.

Both of these take lots of development time, and in a large project monthly game development costs being maybe something like $100k, it might not make sense financially. Things that require content or game play changes usually need to be specified in the beginning of the project.

But games can do quite a lot with how they scale on different hardware; scale textures, scale level of detail, drop some special effects and so on.

NVIDIA : Scalability in the game engine. This includes both multiple geometric levels of detail (LOD) and multiple shading models, so that the engine can fall back on slower or less capable hardware. Realistically, though, most of the enthusiasts will have pretty recent graphics hardware.

Question 8

Beyond3D : We now see a huge range of performance and features in consumer video boards. Many computers still ship with TNT2 M64s while others will soon be upgrading to NV20 and the likes. How can you ever target this huge span of performance and features? Are dynamic adaptable and flexible 3D game/app-engines the way of the future?

Croteam : Yes. You'll always have 'TNT M64' and 'NV20' at the same time. It is not polite to ignore all those 'M64' owners just because they cannot afford 'NV20'. At the same time, it is not wise that you don't provide all the 'bells and whistles' for the 'NV20'.

MadOnion : I think it's the game content that needs to adapt. Of course the 3D engines are a part of it. But honestly, I don't think big manufacturers should be shipping with inadequate graphics cards, or at least clearly tell the buyer that it is not suitable for high-end gaming use without an upgrade. The same applies to the skimpy memory amounts in some of the new PC packages. It's like selling a Ferrari with a VW beetle engine.

NVIDIA : See above!

Question 9

Beyond3D : Do you think that T&L up to now has been "used" or just been "supported"? Do you believe that 3D game/app engines need to be re-engineered to really use T&L or can you hack old engines?

Croteam : It depends on the engine. First step, to 'support' T&L, is rather easy. But a move to 'exploit' T&L can be very difficult. Not to mention the bottleneck of all bottlenecks - AGP! :-)

MadOnion : It depends on how you've planned your game engine. If you had enough abstraction level, it's easier to do large changes and really "support" something. With us it meant almost a full engine rewrite, but because our own tech API, with minimal impact on the applications being developed at the same time.

Secondly, this is a content question. The content needs to scale wider than before to really take full advantage out of HW T&L, and developing that may be much harder than changing the 3D engine.

NVIDIA : I think that software developers who wrote their own T&L pipelines during the DX7 timeframe did themselves a disservice. They prevented themselves from being able to take advantage of hardware evolution. The GeForce era of hardware accelerated T&L simply offloaded the CPU, making higher performance possible. GeForce3's programmable hardware vertex programming makes it possible to do things that weren't possible on PC's before. There just wasn't enough floating point horsepower. I think that in order to really reap this benefit, games need to be re-engineered from the ground up.

Question 10

Beyond3D : Companies like NVIDIA, Imagination Technologies and 3dfx (RIP) are talking about 3D in portable devices like Mobile Phones, PDAs, etc... do you believe in this new 3D-market-segment ? Can you imagine Quake3 deathmatches on Mobile Phones?

Croteam : Yes, especially while you're driving! :-) Car sales will go straight thru the roof!

So, the answer is no - I don't believe in this 'new 3D market segment' (and don't make me laugh :). Maybe I could imagine that, but then ... the phone wouldn't be 'mobile'! :-)

MadOnion : Personally I don't believe in 3D or action gaming on mobile / pda devices because of problems with the display technologies, power consumption, bandwidth, differences with operating systems & APIs. But first and foremost the biggest problem will be the controls. Already the Sony console controls are much worse than mouse/keyboard combo for fast paced games; what about a pen control or 4 buttons from which only one can be pushed down at once (as in the otherwise excellent Compaq Ipaq). And every device has different set of buttons or other controls.

But for example multi-player strategy games may work really well. You can control your empire from anywhere. What could be better?

NVIDIA : I think that the best opportunity for Quake3 deathmatches on mobile phones is being able to trash talk your opponents AS you frag them! Seriously, there is no free lunch (and, no free power, either), and GPU quality graphics isn't coming to handhelds or mobile phones right away. However, it is clear that more and more devices are going to get better graphics. I think that our goal is to get great graphics on 100% of PC's, both mobile and desktop, before we start to focus on other platforms. Did you know that more PC's are delivered worldwide in 2 weeks than there are PDA's, period?