- Joined
- May 28, 2002
- Location
- Sector 7G
Maximum PC, July 2002
In current architectures like the GeForce 4, small shader programs can be used to manipulate pixel and vertex data. The simple programs are restricted to just a few instructions, and are limited in overall functionality. The P10 will allow game developers to write much more complex program with unlimited instructions. In fact, the P10 will function more like a CPU. Instead of treating onboard memory as a simple frame buffer like current GPU's, the P10 uses the onboard RAM as an L2 cache more main system memory. Caching schemes intelligently shuttle data between the onboard cache and main system memory, enabling an addressable memory size of 16GB. The upshot is that developers who code for the P10 will be able to implement larger textures and increased complexity in 3D scenes.
There may also be 10-bit per channel RGB rendering, output, frame buffer mode for ARGB (2:10:10:10), DVD playback & output, filtering & scaling and RAMDACs with full gamma correction, along with quad vertex shader array and four vertex shader units, same as the Parhelia's rendering capabilites. Just as long as 3DLabs keeps the speed @~360/750. Here is their VPU explained:
(BTW: the reason the new Parhelia is so slow is because they packed too many transistors into the core )
Last edited: