![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7) |
Xaber 600
Architectural Block Diagram |
High level of integration |
|
![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
Of course, no product launch
would be complete without the introduction of a few new
marketing terms. This one is no exception; some
carry over from the Xabre 400 launch and a few are new.
Vertexilizer
Engine
The first new term reflects
the Xabre 600's ability to distribute the vertex shading
load between the GPU and CPU in order to achieve what
SiS considers to be an optimal balance between dedicated
graphics hardware and overall performance. The
benefit of this design is two-fold. First, as CPUs
become faster, the graphics processor scales
accordingly. Secondly, the vertex shading can be
upgraded independently of the hardware, so
theoretically, the shading engine could be upgraded to
support the Vertex Shader 2.0 specification contained
within DirectX 9.
![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
SiS' own demonstration of
the Vertexilizer Engine scaling with processor
performance.
Pixelizer Engine
Pixel shaders can't be
emulated, so SiS does include a Pixel Shading engine on
the Xabre chip itself, compatible with the DirectX 8.1
1.3 specification. Coupled with the Vertexilizer
engine, SiS is able to claim full DirectX 8.1 support.
Frictionless
Memory Control
As graphics hardware becomes
more powerful, addressing the need for memory bandwidth
becomes ever-important for hardware manufacturers.
ATI implemented first included HyperZ on it's RADEON,
and NVIDIA soon followed suit with the Lightspeed Memory
Architecture. Both implementations are designed to
optimize available bandwidth using compression, fast
Z-buffer clears and occlusion culling. SiS also
delivers a similar strategy, and is aided by a 128-bit
memory bus on which resides 64MB of DDR memory operating
at 300MHz. Without the benefit of the FMC, the
Xabre 600 enjoys a peak memory throughput of 9.6GB per
second.
Xmart Technology
SiS has wrapped its latest
software developments into a package dubbed Xmart.
The first component of Xmart, XmartDrive, provides for
frequency throttling of the Xabre 600. Though it
may be an interesting feature for mobile users, desktop
customers don't need to pay much attention.
Secondly, SiS has developed XmartAGP, similar in design
to ATI's SmartGART, which automatically sets AGP speeds
to match the capability of your motherboard (again, not
something that requires much attention).
XmartVision is also included to adjust brightness in 3D
environments that may be a little too dark.
Most significant is the
Xminator II driver package, which addresses many of the
issues originally encountered with the Xabre 400.
SiS has added a texture slider in both the OpenGL and
DirectX property pages, as well as an overclocking
utility. It should be noted that using the Xabre
600's default texture setting, 3D quality suffers
heavily. Maybe we've been playing too many games,
maybe we've tested too many video cards, but the
discrepancy between the GF4/RADEON 9000 is significant
enough to be a distraction. Thus, our benchmarks
reflect performance with both DirectX and OpenGL sliders
set to "Quality." Just for the sake of comparison,
though, here is how the Xabre 600 performs at each
setting:
![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
Clearly, the texture slider
has a significant impact on performance (and quality as
well), so if quality doesn't bear the same weight as
performance, consider this graph representative of
performance in 3D Mark.
|