Specifications
Both the R420 core and the NV40U core are based on the 0.13-micron manufacturing technology. The R420 has 160 million transistors as opposed to the NV40U's 222 million transistors.
Now that's a lot of transistors! Unfortunately, because of the GPU's design process, these transistors aren't really as efficient as we are led to believe.
NVIDIA previously, for their NV38 core, decided to implement a 4x2 configuration for their pixel pipelines as opposed to ATI's 8x1 configuration. Because of that, they were slower when shaders were used due to wasted clock cycles.
This time around, however, they have corrected their configuration and implemented a 16x1 pipeline architecture which is basically the same path that ATI treads. Unfortunately, they locked down four of the pipelines for the X800 Pro.
Another thing that NVIDIA decided to implement is Shader Model 3.0. This supposedly allows the shaders to work faster and removes the bottleneck from the hardware as far as shader commands are concerned. ATI on the other hand decided to stick with the old Shader Model 2.0.
NVIDIA's choice of Shader Model implementation natively supports DirectX 9.0C whereas ATI doesn't. Will this prove to be a disadvantage? We'll find out soon enough.
When it comes to performance, ATI's offering looks very intimidating indeed on paper. With a higher fill rate and core speed, it looks faster than NVIDIA's. The only things that the stock 6800GT has that beats ATI in this comparison is memory bandwith and memory clock speed.
<<< The Contestants! (Continued) : Previous Page | Next Page : The Test Bed, Testing Methodology >>>