Wii's graphically superior to anything last generation!

[bdbdbd]

HappySqurriel said:

bdbdbd said:

It's pretty shitty to compare specs between different architectures, it's also not about the performance, but also how it is used.
Sports car with 100kW of power, has better performance than a truck with 400kW. But try to pull load of 20 tons with sports car with 400kW, when it's easy for a truck with 100kW.

Back in the 80's and early 90's "Bits" was the meaningless stat that everyone was using to determing processing power. Intel pushed "Hz" because, although Intel processors were overall faster, AMD and Cyrix processors were generally faster per "Hz" and ran at a slower speed. Now it seems like everyone is pushing "Cores" as the meaningless measure of processing power ...

The truth is that "processing power" is far more complicated than that and can not be easily measured across architectures. Hypothetically speaking a processor that was designed to be used for game development and the size (and cost) of the chip were not issues could have an extra wide bus (1024 or 2048 bits wide), have 48 64-bit wide registers for holding 3 Matricies, and include Matrix Multiplication/Addition/Subtraction/etc. functions.

Matrix-Matrix and Vector-Matrix multiplication is done far more often than anything else in a modern 3D game. In an old architecture (like the N64) to do matrix multiplication would require 32 Memory movements, 64 Floating point multiplications, and 48 floating point additions. With this hypothetical processor this could be reduced to 1 to 2 Memory Moves, 1 Floating point multiplication (64 at the exact same time) and 2 Floating point additions. In comparing these two very different architectures the "old architecture" would need to run far faster, and have far more cores in order to be similar in performance to the theoritical processor I described.

People have to have something from what to say "mine is better". It's pretty much about how specialized processors are. Comparing specs between GPU:s and CPU:s, you'll notice, that GPU:s often have higher specs than CPU:s, at the same time while they work at lower clock cycles. As far as i've understood differencies between high and low instructions, you can get the maximum performance from CISC by using code that is "complex" enough. I couldn't find the right words to describe what i mean, but i think you understood. :)