Mainstream/Value parts
Audi S3 S8
S3 has plans for two more products to service the mainstream segment: the S8 and S8 Nitro. The S8 family is based on the exact same core as the F1 family, right down to the eight pixel pipes. S3 just takes an F1 core and clocks it lower, right now they’re targeting a core clock speed of anywhere from 275MHz-300MHz. This is a lower clock speed than RADEON 9600/GeForce FX 5600, but thanks to its eight-pixel pipeline, S8 doesn’t need high clock speeds for optimal performance. In fact, a 300MHz core clock frequency would put the S8/S8 Nitro well ahead of the competition in fill-rate.
In fact, with eight pixel pipelines, the S8 family matches up more evenly with the core architecture of the RADEON 9500 PRO.
![S3 DeltaChrome Update [ An S8 reference board @ 800 x 550 ] > View Full-Size in another window.](images/02-s.jpg) An S8 reference board
|
|
![S3 DeltaChrome Update [ The DeltaChrome mobile reference board @ 800 x 511 ] > View Full-Size in another window.](images/03-s.jpg) The DeltaChrome mobile reference board
|
|
Like the F1 family, the S8 will take advantage of a 128-bit DDR memory interface with a clock speed target of 275-300MHz. This would provide the S8/S8 Nitro with up to 9.6GB/sec of peak memory bandwidth, the same figure as ATI’s RADEON 9600 PRO. When you combine this with its 8x1 architecture, the S8 could be one compelling product in the mainstream segment.
Where we really see the S8 doing well is not in the upper end of the mainstream segment (where Nitro resides) but towards the low-end. It could be argued that the RADEON 9600 and GeForce FX 5600 are underpowered parts, in some cases falling short of GeForce4 Ti 4200. This is because these cards don’t have the fill rate to keep up in older games. Thanks to its eight pixel pipes, the S8 won’t have this problem. If S3 pairs the S8 with high-speed memory, they could do really well in the lower half of the mainstream segment.
The S4
For the value space, S3 has prepped the S4 and S4 Nitro. As their name implies, the S4 series employs a four pixel pipeline architecture, with one TMU per pixel pipeline, and supports the same feature set as the other DeltaChrome products. This puts S3 at a distinct feature advantage.
Like the other DeltaChrome products, the final core clock frequency for S4 has not been determined. S3’s range varies from 300-400MHz, so we’re looking at a fill rate that ranges from 1.2-1.6 Gigatexels/sec. S3 has settled on a final clock speed for the memory however: 300MHz.
It will be interesting to see how the S4 performs in comparison to its closest competitor, GeForce FX 5200/5200 Ultra. S3 is quick to point out that their dual hardware vertex shaders should give them an advantage over NVIDIA in systems with slower processors, a condition that’s prevalent in the value space. S4’s memory subsystem will fall short of GeForce FX 5200 Ultra however, which could give NVIDIA an advantage at high resolutions.