Proof that GHz dont matter

[Gnufsh]

Saying clock speed doesn't matter is alot like saying it's the only thing that matters, incorrect. The performance of an architecture depends upon it's design. It can be like a P4 and do less per clock, allowing it to achieve a higher clockspeed. Or, it could do more and not be able to clock as high. Those are both valid ways too squeeze more performance out of a chip. A chip's speed is only one of many things that contribute to a chip's performance, but it is one of those things, and shouldn't be totally discounted. Personally, I'd rather have a 1.6GHz Athlon or Pentium-M than a 2GHz P4, but I'd rather have a 3.6GHz P4 than a 1GHz G4. Even assuming the G4 did an average of three times as much per clock than the P4, the P4 would still be faster.

edit: For those you you who say GHz are meaningless:
giga is the prefix for 10^9 and a Hz is equivalent to 1/s, so GHz are 10^9/s or 1000000000/s. See, they have a meaning...

 

[Albigger]

... So, if

chip 'A' does infinite amount of work per clock cycle but its operating frequency approaches zero

and

chip 'B' operates at an infinite frequency, but the work it does per clock cycle approaches zero


the REAL question is:

is chip A faster/slower/the same speed as chip B????

 

[Gnufsh]

That really depends on how the speeds approach what they approach. You need an equation to take the limit.

 

[@md0Cer]

I know we already know this, but just to drive the point home:link

The worlds fastest supercomputer is running only at 700MHz.

Another way to prove that point would be with the upcoming dualcore CPU's. They will run at a lower clockspeed than their single cored counterparts.

Also, a CPU's architecture has allot to do with this. A good example would be a Pentium M with a 10 stage pipeline vs. a Pentium 4 Prescott with a 31 stage pipeline.