Card Spec's
Specifications
- 166MHz 128-bit 2D/3D core
- 16MB of 166MHz SDRAM
- 350MHz
RAMDAC
- 333 MTexels/sec
- 166 Mpixels/sec
- 7 MTriangles/sec
-
2.66GB/s peak bandwidth
- Resolution support of up to 2048x1536
-
Glide/Direct3D/OpenGL support
- Integrated TV/S-video output
- AGP 1x, 2x
support
- Single pass trilinear mip-mapping
- 16-bit 3D rendering
support
The ultra-high speed RAMDAC provides resolution support upto
2048x1536 at an ergonomical 75Hz refresh rate. Speaking of which, it supports
refresh rates from 60Hz to an amazing 160Hz! The Voodoo3 does not provide
support for 32-bit colour rendering or AGP texturing. But performance-wise, it
really takes off thanks to a dual 32-bit rendering pipeline capable of
unleashing 2 texels per clock cycle.
Bundled stuff with the retail box
includes: Voodoo3 driver/tools CD, Need for Speed 3, Descent 3, Unreal and a
Quick Install manual. These games are enough to keep you hooked for hours and
drooling at how quick your new Voodoo3 3000 is. A detailed manual isn't included
but sufficient on-CD documentation is available in HTML format. Upon first
inspection the thing that catches your eye most is the large silver heatsink
that rests upon the Voodoo3 chip and the Voodoo3 certainly needs it! This heat
sink is removable as and when needed to accomodate a specialized cooling
fan/heatsink combination. But more on that when PCinsight reviews cooling units.
There are 8 2MB German-made Seimens SDRAM chips seated on the board as well. The
board I received has 6ns SDRAM chips. By calculating the inverse of 6ns (1 /
0.000000006) then dividing that number by 1,000,000, you get the resultant speed
of the SDRAM in MHz. Neat trick huh? As the Voodoo3 2000 and 3500 had RAM access
times of 7ns and 5.5ns, respectively. The SDRAM speed of the 2000 is 143MHz and
the speed of the 3500 is 183MHz. The graphical rendering feature set of the 3000
is identical with the 2000 and 3500. The overall speed difference of the 3000
and the faster 3500 is simply a small percentage difference. 10 percent points
to be exact. The only thing that the 3500 offers above the 3000 is video editing
capabilities and an FM/TV tuner, which of course comes at a price premium. The
3000 does offer TV output through an S-Video connector built into the 3000
itself next to the analog out. Included with the kit, is a cable with an S-Video
out connector on one side (to plug in the card) and a composite video connector
on the other side. Using this, you can take a regular RCA cable and attach your
TV or VCR on this end. The 2000 doesn't offer any of this. But all in all, you
get what you pay for, right? Like the rest of the Voodoo3 line, the 3000 has a
fixed 1:1 ratio between the core clock and SDRAM speed. So by increasing either
of these speeds, you are in effect increasing the other.
Performance
Results
Before we get
on with the actual benchmarking here is a description of the testing platform
that we used. Now before you break out into uncontrollable laughter when looking
at the configuration, understand that Pentium III 600s and Athlon CPUs are not
easily available in India. "Well, how about a Pentium III/450?" -- Uhhh, well, I
just get can't afford that right now! ;) So in the mood and understanding that
we are in, let us loudly label the following setup as the PMSC (or Poor Man's
Supercomputer).
Testing Platform
Celeron 333 - Slot 1 (clocked at 333MHz and overclocked to
416MHz)
Soyo SY-6KL motherboard (66x5 @ 333Mhz, 83x5 @ 416 MHz)
96MB
SDRAM
Samsung 0432A 4.3 GB Hard disk drive (UDMA/33 enabled)
Viewsonic
E653 15" Monitor
3dfx Voodoo3 3000 drivers (ver. 1.03.04)
Windows
98
For OpenGL testing, Quake3 Demotest v1.09 was used in the normal and
fastest settings. Since the Voodoo3 line does not have support for 32-bit
rendering, we did not run the high quality setting. Quake 2 was also used to
test the OpenGL performance of the 3000. For Direct3D performance, we used
Expendable, Forsaken and 3DMark99 MAX Pro with all tests and enabled. For your
viewing pleasure we have also added benchmark results under Unreal. Each of
these tests will be run 3-times with the end result being the mean value of the
3 runs.
|
RSS Newsletter
