You can't compair apples and oranges...

Exfoliate

Exfoliate

Geek Trainee
I realize ATI and nVidia's chipsets aren't identical but I was just sort of wondering why nVidia can get more out of say, a 350MHz chip (like the 6800GT) than ATI? If the Radeon X800XT ran @ 350MHz and so did the 6800 Ultra for example, the Ultra would trounce the XT. It must be the architecture I guess as otherwise, aside from a few less transistors on the XT's part, the specs would be pretty similar? So how does it work (more or less)? When it comes to processors I understand that intel has really high clockspeeds yet can be destroyed (in some things) by AMD's and Apple's chips that have far lower clockspeeds, as the fabrication process and architecture is quite different (so nVidia is like AMD...), but the video card thing isn't so obvious to me. Hope you don't mind clearing this up, thanks.
 
There's also Direct 3D vs OpenGL. nVidia is traditionaly stronger in OpenGL than D3D and ATi is vise versa. Now, nVidia has been getting better at D3D, but their chip designs are simply different than ATi's. Different means to the same end.
 
Its the architecture of the chip. There are many factors which could allow one chip to scale with clockspeed better than the other, e.g. heat production (often linked with transistor leakage), transistor technology used coupled with how efficient the circuit along with how complex it is aswell.

The reason why some chips with lower clockspeed perform better clock for clock than their rivals is because they may have higher ILP (instruction level parallelism). Branch predictions units and how accurate they are; how deep the pipeline is as well as on-die cache efficiency (absolutely vital to keep performance good with a deep stage pipeline) and number of registers/data bus width and their respective latencies. Bandwidth available also improved performance in processing units obviously as a fast and efficient I/O is extremely important for eliminating performance bottlenecks.

The reason why Intel were able to scale with their clockspeed more easily than AMD was because of the NetBurst architecture they used. Even so it was meant to scale easily to 5GHz. (Well thats not exactly what happened was it ;). AMD chose the higher IPC route, a decision which probably paid off in the long run.

Many GPU's are actually getting to a state in which they are more complex and often have a higher IPC than desktop CPUs. And with increasing complexity/branch prediction as well as differing sizes of cache/no. registers clockspeed is harder to raise. But GPU for GPU they both have different processing cores and IO system so they're not going to be identical in terms of performance per clock.
 
You must log in or register to reply here.
Back
Top