[QUOTE="JMR09"]The PC has the capabilities of having much, MUCH more advanced graphics than any console game. The problem is in regards to scaling for those who don't have high-powered PCs, which usually results in developers not taking complete advantage over the advanced resources available to them. They do want to make money and be successful after all.
CwlHeddwyn
This is why pound for pound the console gamers usually get a better performance out of their hardware than PC gamers. There are just too many CPUs, GPUs and RAM chipsets out there for PC developers to fine tune games for- this is why PC games rely in RAW power rather than efficiency. Fortunately as its already been shown most modern PCs have a lot more processing power than the consoles. Back in 2006- 2007 however this was not the case- the consoles were Very competitive back then.Geforce 8800GTX was released a few days ahead of PS3.
Geforce 8800(G80) is designed to be more efficient compared to G70/RSX.

To minimise RSX/G70's shader/math unit stalls, G80 has decoupled texture and shader/math unit design i.e. a feature for NVIDIA Giga-threads and AMD Radeon HD's Ultra-Threads. AMD Xenos includes decoupled texture and shader unit design, but it's only has 64 threads.NVIDIA Giga-threads and AMD Radeon HD Ultra-Threads has thousands of threads.
Another G70/RSX's design flaw

A stall in"FP32 Shader Unit 1" will stall the rest of shader units.
With AMD's Radeon HDs, shader/math units pipeline design was flatten which maximises shader/math unit usage.

---------------------

As GPU raster rendering solution, transferring data between CELL's SPEs and RSX is considered to be less efficient compared to a single "fat" GPU solution i.e. the ping-pong transfers between RSX and SPEs adds latencies.
Unified shader architecture is considered to be more efficient compared to older non-unified shader architecture i.e. maximises shader/math unit usage.
Multiple thousands of data registers in DX10 Geforce/Radeon HD GPUs adds to math compute efficiencies i.e. attempts keep more data in the GPU cores in the fastest known data storage tech e.g. registers. Radeon HD 4870 has 2.5 Megabyte registers storage.CELL only has a few hundreds of data registers.
Log in to comment