Question regarding different kinds of processors

[Sunner]

I've asked this question before on several occasions, but never gotten a satisfactory answer.

The primary question is the difference between general purpose MPU's such as P3's etc, and graphics chips, what makes a P6 class chip capable of so much higher speeds on the same process, as compared to, say a GeForce.

Take the Klamath and the TNT for example, both on a .35 process.
The TNT topped out at 90 MHz, and with good cooling, one could run them upwards of 120 MHZ, while the Klamath reached 300 MHz, with some OC'ing upto 350+ MHz.

Now granted, a GFX chip rarely comes with the same kind of cooling as does a CPU, but no matter the cooling, a GPU will never even begin to aproach the speeds of a CPU.

Whats the deciding factor in how far a GFX chip can go, process and cooling aside?

 

[nam ng]

It's the same fundamentals as RDRAM VS DDR SDRAM, deep pipelines, high latency, high clock... against wide pipelines, low latency, low clock, with all else equal.

If both implementations are well done, neither have the performance throughput advantage over the other but for usage specific implementation advantage.

An example usage specific implementation advantage of a single port 100Mb/s switch usage, against 10 ports @10Mb/s switch (10/100) usage, all operating at max bandwidth condition.

Implementation specific advantage - Traditional x86 is equivalent to 10/100 HUB (cpu=100, other devices=10), while AMD + nFORCE equating to 10/100 SWITCH (cpu=100, other devices=10).