Summary of engine features:
    Sometimes the engine outstrips its time and in generally is designed for modern and (even future) hardware.
    The main goal was to utilize all resources of video boards especially the T&L
    'Cause not all of us have it (even the developers :) the most of engine features (almost the vertex-related)
    are emulated (and very fast emulated) in software.

    Distinguishing feature of XRay engine is high parallelism, which results in low number of situations when CPU
    is waiting for GPU or GPU resources to be available.

In Detail (not sorted by any kind of priority):
* Variable, dynamic geometry detail of scene
    - highly optimized for T&L hardware
    - doesn't modify vertex data - only rearrange indices
        - which results in ZERO overhead
        - disadvantage: can kill GPU vertex cache :(
* Colored dynamic, per vertex lighting
    - Intelligent light sources selection & clipping
        - Makes possible more than 100 dynamic lights per frame at reasonable FPS-es
* Very impressive simulation of atmosphere effects like sky/clouds/suns
    - Of course it depends on art-designers skills but the bumped sky looks great :)
* Fast, full-accelerated by hardware "Hard Shadow" generating algorithm
    - Disadvantage: requires presents of 8 (or more) bits stencil buffer
    - Advantages: the most current hardware has already had it :)
* Full scene anti-aliasing
    - Supports Accumulation Buffer & super sampling technologies as 3dfx's T-Buffer (R)
    - Engine designed in mind to use Multisampling than it will be available on hardware
        - Enables to use "motion blur", "depth of field blur" and "soft shadows"
* Table W-Fog & Z-Fog support
    - automatic rollback to emulation via vertex fogging if hardware doesn't support it.
* Intelligent use of triple buffering scheme
    - results in FPS-es increase on video boards which have enough memory to hold all buffers
* Motion compensation scheme
    - Hides many visual artifacts produced by FPS-es instability
* Light maps and dynamic lightmaps
    - Up to 3(three) texture stages that blends using almost infinite combinations of pixel-shaders
    - Disadvantage: some pixel shaders requires DirectX8 and appropriate hardware support
* Flare system
* Particle system
    - Can be highly accelerated by upcoming hardware, which will be available at fall'00
* Bump mapping
    - Even "DotProduct3" emboss bump mapping supported
* Environment mapping
    * Cubic environment mapping
        - Realization fully depends on hardware capabilities
    * Spherical environment mapping
        - Real alternative to CubicEnvMap on all current hardware
        - Doesn't highly depends on geometry complexity
* Reflections
    - Pre-rendered textured reflection simulation can be applied on any object or part of it
        without any performance penalty
* Curved surfaces
    - 'Cause even modern hardware can't do the tessellation in real-time - it is done on load-time
    - Tessellation is based not on the real splined surface definition but on the control surface NURBS-tessellation
* N-Patch support
    - Real-time support is only guaranteed for DirectX8 with appropriate hardware installed
* Single-mesh skeleton animation
    - High-speed blended animation system capable of up to infinity number of bones interpolation & modulation
    - The first 256 bones blending will be accelerated by GPU (if it capable of doing such thing :)
    - Combined with N-Patch technology can give photo-realistic models visualization in real-time
* Portal-style, non-linear subdivision based visibility detection system
    - Optimized for T&L hardware in the way of grouping primitives to large-size batches
    - Can handle up to 20,000 polys in one frame on software transformation(PII 450) and up to 80,000 in T&L(GeForce2)
    - Greatly reduces overdraw - which is the main slowdown on even high-level video boards
    - Enables to use more texturing effects compared with todays engines (Even Q3 & Unreal)
* Intelligent hardware state caching technology
    - Results in speed increasing on bed-written rendering code (not all of the programmers code good, isn't so? :)
* Dynamically generated textures
    - High-speed fractal & particles based generating
    - 20% of Unreal(TM) engine based on this technology - it is approved.
* OBB-style collision-detection system
    - Capable of ideal determining collisions between high-poly models in high-detailed world
    - Disadvantage: Ray-picking is not highly optimized now
* Independence of engine from world and objects it holds
    - All objects lies in independent DLLs and the Engine itself doesn't know of them until level load time
    - Results in increased speed of development (and especially team development)
* Powerful scripting
    - Full realization of C-language
    - Only 6-8 times slower than real (x86) code   (isn't great? :)
    - Enables fast AI development and almost infinite number of texturing effects
* The most important part of the mathematics is optimized for PIII SSE & K6/K7 3DNow instructions


* High quality HRTF-3D sound, accelerated by hardware if possible
* Low-penalty & "low-delay" reaction technology (about 1-3ms to start HRTF 3D Sound Source to play)
* No "glitches" when starting a sound even in full HRTF model
* Incredible support for sound environments
    - Requires EAX-capable sound card
    - A3D now used only to speed up some calculations (handled by DirectSound subsystem)
* Sound clipping and sound reflection
    - If Diamond Corp. continues support for its A3D standard - will be realized on hardware
* High speed streaming sound decoding
    - Means playing MP3/MP2/WMA/ADPCM tracks with almost no penalty to performance
* Context relative multiple-mixed music streams
