Slashdot Mirror
AMD To Hide MHz Rating From Consumers
pezpunk writes: "Tom's Hardware is reporting here that AMD's next-generation Athlons will be identified by model number rather than Mhz rating. This means that an Athlon will be designated an "Athlon 1600" even though it's only a 1.4Ghz part. The true clock speed of the chip will NOT be shown either on the chip itself or even in the BIOS. Apparently, they're desperate to compete with higher-clocked Pentiums in the minds of consumers -- proof that even the underdog can pull dirty marketing tricks =("
On the one hand, as has been pointed out a dozen times, MHz is a pointless number. It's like talking about engines in terms of liters. Higher numbers are not always better than lower numbers.
More importantly, CPU speed has stopped being an issue for most people. I know, I know, there are always some people who love to claim to be the exception to the rule, people who insist they need to solve systems of fifty million linear equations or that they do aircraft design at home, but for most people, even professional programmers, speed has gone beyond what we know what to do with. When the 333MHz Pentium II rolled around, I started coding in the highest level language I could find, be it Lisp or Smalltalk, because what I then saw as excessive performance afforded me the luxury. Now we have processors that are five times faster, and I don't think about speed at the hardware level.
Slowness is usually something that's outside of the realm of hundreds of millions of operations per second. For example, Internet Explorer takes too long to start up on my machine. Lots of people apparently think that a faster processor would fix that. And other people complain that a game is stuttery, and think they need more CPU performance, when half of the time it comes down to a buggy video driver.
Your posting, while well thought-out, is technically nonsense.
I write this as a person with a bachelors in Computer Engineering who is currently completing masters in EE.
First of all, the waveform in question is a square wave, not a sine wave. So I don't see how pi comes into...well, anything. You go though some basic trig to prove that f=c/L (frequency=speed of light divided by wavelength). So what?
Some architecture triggers on the rising edge, some on the falling. It does not really matter.
The rumor you heard about Intel architecture "counting" both rising and falling edges is silly; what counts is the number of pulses, not the number of rising and falling edges.
Now, there may be a basis to that rumor in that some architectures where the CPU runs at a multiple of the bus speed and triggers on both the rising and falling edges. The older Athlons, for example, run at a 200Mhz clock speed. But the external CPU bus runs at a 100Mhz clock speed.
Does this mean that AMD is cheating? That they are "claiming" 200Mhz when it is only 100Mhz?
No. What it means is that the Athlon triggers on the rising edge, then half-a-period-later it triggers again on the falling edge. Assuming that the Athlon triggers on a rising edge, this could be accomplished by inverting the clock, and ORing the signals together (although it is not that simple, you get the idea). So for each external 100Mhz clock pulse, the CPU fires two internal clock pulses. And the speed is doubled. So your 1Ghz Athlon runs externally at 100Mhz with a 5x multiplier. Inside it runs at 200Mhz with a 5x multiplier. 200x5=1000. See?
And since you trigger every half-period, you cut the time of the period in half. f=1/T, where f is frequency and T is period. So when you cut the period in half, you double the frequency.
That is why the new 266Mhz FSB Athlon chips need to have the external clock speed set at 133.
So why not just run the PC board at 200Mhz and forget all this silly clock-doubling hardware? It is not that easy. Desinging a glass-epoxy PC board to work at 100Mhz is hard; 200Mhz even harder. As you go higher in speed, harmonics in the microwave regions begin to creep in and most digital designers are not ready for that sort of variable. Plus, it raises the cost of everything in the PC. Remember that your PCI ports still run at 33Mhz on most machines....
Okay, so why is the Mac faster at the operations you used in your tests? It's a different architecture! You are comparing apples (pardon the pun) to grapefruit here. It's like saying that if a 10-cylinder diesel truck is at 4000RPM and a 2-cylinder moped is at 4000RPM, they should be going the same speed.
How many CPU cycles does each operation take on the G4? How many CPU cycles do those same operations take on the Intel? What about differences due to setup and OS lag? Is the compiler optimized for the CPU? If so, is it using out-of-order execution? That is the sort of thing you need to know for a test like this. The same operation may take 10 cycles on the Intel and 1 on the G4. So, for that operation, the G4 would be ten times faster. If an operation takes one cycle on both machines, the Intel would be twice as fast as it has twice as many cycles per second. Cycles Per Second, or CPS, is also known as Hz. And the Intel chip is running at twice the clock rate as the G4. Trust me. What it is not running at is twice the speed, since operations on the two machines take different numbers of clock cycles to complete.