Leighton
Member
Registered: 21st Feb 01
Location: Liverpool
User status: Offline
|
quote:
Originally posted by Richie
Intel get high clock speeds, but what matters more is the clock cycles per second, which AMD win in that respect.
im not sure what you mean here "clock" and "cycle" have a 1:1 ratio. Take 3 GHz for example. That's 3000 MHz, they call that clock speed. What's a MHz? A Hertz is 1 clock cycle, so a MHz is a million cycles, etc. i think what you mean is intel designed a Long pipeline whitch help chip designers to scale up in MHz. But it also be harmful when you have a branch misprediction. If you dig into Assembly language and CPU documents for both Intel and AMD, Intel's CPUs have a penalty of up to 14 cycles every time it misses a branch in code. Here's the kicker: CPUs always miss the first time around. So when you're in a very big loop of code, that penalty really adds up and you start to see that 3 GHz figure dwindle down. AMD chips handle branch mispredictions a lot better.
but you are on the right lines. i have just build a Xeon server but that was aplication specific my next boxes will be AMD Opteron or 64 chips with cost , power consumption and performance
|
