GeForceFX - NVIDIA Goes Hollywood?
At Comdex Fall 2002, NVIDIA will unveil the heir to the highly successful GeForce4 Ti line of graphics chips, which was in turn presented to the public in complete detail in February of this year. Now NVIDIA seems to be copying its main competitor ATi's modus operandi, presenting its new chip, dubbed GeForceFX and code-named NV30, in stages. At Comdex, NVIDIA will only be revealing details on the chip's technologies and features, which far surpass the DirectX 9 specification. There will not, however, be any concrete information on performance or the card versions that will be produced - not to mention testing samples.
This makes it nigh on impossible to answer the big question everybody is asking and has been pondering for months now. Will the role of technology and performance leader revert from ATi, who captured it several months ago with its Radeon 9700 PRO, back to NVIDIA and its GeForceFX? I'd like to save that question and its answer for the end of this article, and cover the GeForceFX's new features and capabilities first.
| Header Cell - Column 0 | NVIDIA GeForce4 Ti4600 | NVIDIA GeForceFX | ATI Radeon 9700 PRO |
|---|---|---|---|
| Chip Technology | 256 bit | 256 bit | 256 bit |
| Process | 0.15 Micron | 0.13 Micron | 0.15 Micron |
| Transistors | 63 Mio | 125 Mio | ~107 Mio |
| Memory Bus | 128 bit DDR | 128 bit DDR2 | 256 bit DDR |
| Memory Bandwidth | 10.4 GB/s | 16 GB/s | 19.8 GB/s |
| Pixel Fillrate | 1.24 Gigapixel/s | ~4 Gigapixel/s | 2.6 Gigapixel/s |
| Anti Aliased Fillrate | 4.8 Billion AA Samples/s | ~16 Billion AA Samples/s | 15.6 Billion AA Samples/s |
| Max FSAA Mode | 4x | 8x | 6x |
| Triangle Transform Rate | 69 M Triangles/s | 350 M Triangles/s | 325 M Triangles/s |
| AGP Bus | 1x/2x/4x | 1x/2x/4x/8x | 1x/2x/4x/8x |
| Memory | 128 MB | 128/256 MB | 128/256 MB |
| GPU Clock | 300 MHz | ~500 MHz | 325 MHz |
| Memory Clock | 325 MHz (650 DDR) | 500 MHz (1000 DDR2) | 310 MHz (620 DDR) |
| Memory | BGA 2.8ns | BGA 2.0ns | BGA 2.9 ns |
| Vertex Shader | 2 | FP Array | 4 |
| Pixel Pipelines | 4 | 8 | 8 |
| Texture Units Per Pipe | 2 | 1 | 1 |
| Textures per Texture Unit | 4 | 16 | 8 |
| Vertex S. Version | 1.1 | 2.0+ | 2 |
| Pixel S. Version | 1.3 | 2.0+ | 2 |
| DirectX Generation | 8 | 9.0 (+) | 9 |
| FSAA Modi | MultiSampling | MultiSampling | MultiSampling |
| Memory Optmizations | LMA II | LMA II OptimizedColor Compression | Hyper Z III |
| Display Outputs | 2 | 2 | 2 |
| Chip Internal Ramdacs | 2 x 400 MHz | 2 x 400 MHz | 2 x 400 MHz |
| Chip External Ramdacs | - | - | - |
| Bits per Color Channel | 8 | 10 | 10 |
| Special | - | TV Encoder On-ChipExtended ProgammabilityAdaptive Filtering | TV Encoder On-Chip; FullStreamAdaptive Filtering |
| Estimated Price | ~ $200-300 | Not specified | $399 |
