Pixel filling power
To best understand how the DeltaChrome S8 Nitro relates to its competitors, our handy-dandy fill rate table is helpful, because it shows the theoretical pixel-pushing power of each card. Although these numbers aren't fate, they do help dictate how a graphics card will perform overall.

The extra 25MHz doesn't boost the DeltaChrome S8's performance potential all that much, but the S8 Nitro has fill theoretical peak fill rate numbers that are more than competitive with the GeForce FX 5700 Ultra and Radeon 9600 XT, especially in single-texturing.

At 10.4GB/s, the S8 Nitro leads the Radeon 9600 XT in memory bandwidth, but both cards are behind the class leader, the GeForce FX 5700 Ultra. Remember, though, that DeltaChrome doesn't have a crossbar memory controller like the ATI and NVIDIA chips do, so it may not be able to use its memory bandwidth as efficiently in real-world applications.

Of course, we can put these theoretical peak numbers to the test fairly easily using 3DMark.

Hmm. The S8 Nitro's multitextured fill rate is faster than the competition, as expected, but its single-textured fill rate is well below what we'd expect from an eight-pipe design. This low performance raises red flags for us, because both NVIDIA and SiS have misled the press and the public about the pipeline configurations of their GPUs in the recent past. NVIDIA led us to believe its NV30 was a 8x1 pipeline configuration, but it turned out to be a 4x2 design—that is, it had four pipes with two texture units per pipe. Similarly, SiS said its Xabre was a 4x2 design, but it turned out be a rather unorthodox 2x4 configuration.

S3, however, is adamant DeltaChrome S8 is an eight-pipeline GPU. I inquired about this issue pointedly and repeatedly, and the answer was consistent. S3 says the pipeline config doesn't switch to 4x2 at any point.

The folks at S3 raised some interesting questions about how 3DMark's fill rate test works and whether it's structuring its test scene in a way that maximizes performance from the DeltaChrome GPU. The S3 engineer I talked to said a test using a large triangle with single textures would get eight pixels per clock out of the GPU, because DeltaChrome S8 can operate on eight pixel shader instructions in parallel. All rendering on DeltaChrome, it seems, translates to pixel shader instructions. The test would have to have a very high pixel-to-vertex ratio in order to extract eight pixels per clock. (I'm fairly certain FutureMark has covered that base.) Also, the test would have to use instructions with a single cycle of latency in order to work.

The net effect is that DeltaChrome is much slower in 3DMark's single-texturing test than in the multitextured test, but it also shows that not all rendering pipelines are created equal—a point NVIDIA has been hammering home since just before the NV30's debut.