Brief Architecture Comparison
|
Fill-Rate, in MT/s |
Memory Support, in MB |
32-bit Colour Support |
AGP Transfer Support |
nVidia TNT |
180 |
16 SDRAM |
Yes |
1x, 2x |
nVidia TNT-2 |
250-300+ |
8-32 SDRAM |
Yes |
1x, 2x, 4x |
3dfx Banshee |
110 |
16 SDRAM |
No |
1x |
3dfx Voodoo3 |
286-366 |
16 SDRAM |
No |
1x, 2x |
3dfx Voodoo2 |
180 |
8-12 SDRAM |
No |
N/A |
3dfx Voodoo2 SLI |
360 |
16-24 SDRAM |
No |
N/A |
Since the original introduction of the TNT chipset many months ago, nVidia has been recognised
as an industry-leading manufacturer of quality, high-performance video chipsets. The TNT was promised to be the
long awaited competitive solution to the poorly developed 3dfx Voodoo Banshee chipset. nVidia hoped to ship the
TNT at nearly outrageous clock speeds for its era, but the final product missed nVidia’s estimates by a large margin.
Even at the lower production clock speeds, the TNT proved to surpass the performance and quality of 3dfx’s Banshee
with ease. However, nVidia still lacked the needed clock rates to compete with 3dfx’s ultimate 3D rendering solution,
the Vodoo2 dual-card SLI configuration.
The V2 SLI setup was (and actually still is) a powerful 3D offering with nearly no equals
in its day. In comparison with the TNT’s maximum fill-rate of 180 approximately megatexels per second, the V2 SLI
offered performance of over 360 MT/s!
nVidia had already spent countless millions developing the TNT, thus could not afford to
again develop a completely new architecture. Instead of pouring more money into more research and development,
nVidia designers worked to prefect the already popular TNT chipset. Through a successful die shrink (.35 to .25
micron) and the inclusion of faster memory, the design lab rolled out the exciting new TNT-2 platform less than
a year after the original TNT release. The TNT-2 offered the performance to compete with any chipset available,
with an impressive 250-300+ MT/s fill-rate, plus enough memory bandwidth to sustain high resolutions with true
32-bit color depth. The TNT-2 offered the needed performance and rendering quality to finally compete with 3dfx
at all levels, from the value to high-end market segments. Even with release of the Voodoo3, the TNT-2 proved more
than capable of surpassing this next-generation 3fx chipset. The TNT-2 scaled through several speed grades, with
the last offering being the TNT-2 Ultra. The Ultra offered the highest performance available, at least until nVidia
could develop its next-generation chipset, the GeForce256.
The nVidia Family
|
Fill-Rate, MT/s |
Memory Support, in MB |
Memory Data Width |
Advanced Shading Rasterizer |
HD Video Support |
GeForce-256 SDR |
480 |
32 SDRAM |
128-bit |
No |
No |
GeForce-256 DDR |
480 |
32-64 DDR SDRAM |
128-bit |
No |
No |
GeForce-2 MX |
700 |
16-64 MB SDR/DDR SDRAM |
128-bit SDR, 64-bit DDR |
Yes |
Yes |
GeForce-2 GTS |
1600 |
32-64 MB DDR SDRAM |
128-bit |
Yes |
Yes |
GeForce-2 Pro |
1760-1920 |
64MB DDR SDRAM |
128-bit |
Yes |
Yes |
GeForce-2 Ultra |
2000+ |
64-128MB DDR SDRAM |
128-bit |
Yes |
Yes |
The GeForce-256 architecture was debuted approximately a year after the original TNT-2 chipset
release. The GF-256 offered a myriad of never before utilised options within the consumer market range. These advanced
options included a 256-bit core rendering path, texture compression, full screen antialiasing, cube environment
mapping, and hardware transform and lighting. Combined with extreme fill rates (up to 480 MT/s) and a new high-performance
memory interface (DDR SD-RAM), the GF-256 proved to be unlike any chipset ever offered before. Even in today's
now exaggerated standards, the original GeForce-256 DDR is considered an ominous presence in regards to rendering
features and actual game/3D performance. However, to continue with its commitment of a new chipset every six months,
nVidia further developed and refined the GeForce platform. The next generation architecture, the GeForce-2, has
proven to be nVidia’s best selling and most compelling marketing offering to date.
The GeForce-2 offers performance nearly unparalleled by any competing chipset. The last iteration
of this chipset, known as the GF-2 Ultra, offers fill-rates in excess of 2,000 MT/s, a level still unsurpassed
by any competitors’ architecture. Other advanced features include second-generation hardware T&L, simultaneous
real-time per-pixel shading, high-definition video support, and even advanced 32-bit w-buffering with the latest
DX 8 drivers. Various versions of the GeForce-2 are now available, ranging from the entry-level MX to the powerful
Ultra chipsets. The MX is a scaled down product with standard SDR SD-RAM and lower fill rates, but still proves
more than capable of dealing with any game currently available. The upper-tier segment includes the GTS, Pro, and
Ultra chipsets. Each of these chipsets are basically the same core, but with varying clock rates and amounts of
memory. The GeForce-2 is regarded as one of the best platforms available, even as nVidia is on the verge of releasing
the costly GeForce-3.
|