[bma_mat99]

If for example you have a 2.8 ghz pc, and a 2.5 ghz powermac, how come the powermac is faster.. arn't ghz a measure of speed.... please set my mine streight about this whole hertz thing, how does it work, and what are the real things that show the speed of a computer, and what are the L2 cache, etc.... thx.....

Dont laugh at my stupidity, im 16, and new to macs.... and dont have that much computer knowledge... i know the basics.

[flypenfly]

www.arstechnica.com read the articles on CPU's

Basically, mhz is a measure of clock cycles of a particular cpu. Its not like Miles Per Hour or anything like that.

[CaptainHaddock]

Hertz is technically a measure of "cycles per second". Gigahertz (billions of hertz) can measure the cycle speed of your CPU, but that's only relevant when comparing to other, very similar CPUs.

Think of a CPU as a room full of accountants and their calculators. Each "cycle", each accountant can push one button on his calculator. So yes, if the accountants in Intel's room can push buttons faster, all things being equal, they will get more work done. But there are other considerations:

1. How many accountants are in each room? If the G5 room has twice as many accountants, they can work more slowly and still get more done.
CPU equivalent: number of operations per cycle

2. How smart are accountants? The G5's accountants can do more work with fewer button pushes because they're smarter than Intel's accountants.
CPU equivalent: the G5 has a more efficient RISC instruction set, and it has more registers

3. How well-coordinated are the accountants? Intel's accountants often do the same work twice or do unnecessary work. G5's accountants don't do that as often.
CPU equivalent: piplines and branch prediction, i.e. knowing ahead of time what operations to prepare for

4. How good are the accountants' calculators? G5's calculators can handle bigger numbers.
CPU equivalent: the G5 uses 64-bit instructions and addressing, the P4 only 32-bit

5. How well managed are the accountants? The bus and memory cache on a Powermac can generally use the capacity of the CPU better than an Intel machine. Also, a larger L2 cache holds data that the CPU can access more quickly than regular RAM.

6. How well do they handle special tasks? The G4 and G5 processors have special Altivec co-processors for handling difficult vector processing.

So there are many reasons why a 2.5 GHz Mac is faster than a 3.0 GHz Intel machine. In fact, Pentium 4s are often slower than equivalent Pentium IIIs because of some poor design decisions.

[bradoesch]

If you know anything about cars, then Hertz is very similar to your tach. It's a measure of speed but not results. Two motors both running at 7000 rpm can have totally different outputs. Just like your example of a 2.5 GHz G5 and 2.8 GHz PC processor.

This is exactly why you can't determine the speed of a system by it's MHz or GHz.

[CharlesS]

Originally posted by bradoesch:
If you know anything about cars, then Hertz is very similar to your tach. It's a measure of speed but not results. Two motors both running at 7000 rpm can have totally different outputs. Just like your example of a 2.5 GHz G5 and 2.8 GHz PC processor.

This is exactly why you can't determine the speed of a system by it's MHz or GHz.

Exactly:

RPM = revolutions per minute.

Hz = cycles per second.

revolutions = cycles, so you can convert RPM to Hz by dividing by 60 seconds per minute. Then, you can convert to MHz by dividing by one million, or to GHz by dividing by one billion. So the question is, do you measure the speed of your car by RPM (convertable to MHz) or by MPH (actual work done / distance travelled) ?

[Xeo]

This turned out to be a good thread. I like the analogies you've brought up. They seem to fit pretty well.

Wrong forum, of course. Where to put it, though? I'll be back.

[Catfish_Man]

Originally posted by bma_mat99:
If for example you have a 2.8 ghz pc, and a 2.5 ghz powermac, how come the powermac is faster.. arn't ghz a measure of speed.... please set my mine streight about this whole hertz thing, how does it work, and what are the real things that show the speed of a computer, and what are the L2 cache, etc.... thx.....

Dont laugh at my stupidity, im 16, and new to macs.... and dont have that much computer knowledge... i know the basics.

Unfortunately, it turns out the measuring the real speed of a computer in any general way is amazingly difficult. The SPEC benchmark attempts to do this, but even it only claims any real accuracy for scientific type computations, and maybe a few others, and is still broken in a few important cases (mostly due to modern compilers using sneaky tricks). The real way of seeing which computer is faster is to go find the program(s) you want to run, and see which computer they go faster on

As for the cache, and other important architectural details I highly recommend the arstechnica.com articles (I really started getting interested in this stuff when I read them at about your age), but I'll give it a shot:

basic computer:

processor <---> storage

The difficulty with this is that any storage large enough and permanent enough for modern uses is too slow, so we add in higher speed temporary storage (ram) and it looks like this:


processor <---> memory <-----> storage

This has the same problem though, accessing ram is waaay faster than a hard drive or cd, but still slow compared to the processor. So we add another layer of smaller, faster storage:

processor<--->cache<--->memory<--->storage

Then as chips get smaller we move some of the cache on chip to speed it up even more

[processor <-> on chip cache]<----->cache<---->memory<---->storage

These days, the most cache crazy processor I've heard of (Intel's upcoming Montecito chip) looks like this:


[two processor cores <-> level 1 cache <-> level 2 cache <-> HUGE level 3 cache]<--->memory<----->storage

[kupan787]

Originally posted by bma_mat99:
If for example you have a 2.8 ghz pc, and a 2.5 ghz powermac, how come the powermac is faster.. arn't ghz a measure of speed.... please set my mine streight about this whole hertz thing

Even if you just look at the PC side of things, you could buy:

2.4GHz Celeron
2.4GHz Celeron D
2.4GHz P4 Northwood core
2.4GHz P4 Prescot core

Now, these all have the same GHz rating, but all will perform differently (just looking at the prices for the various parts, and you can see Intel feels the same way). I can't tell you exactly why the different chips are better/worse (I don't know x86 all that much), but I do know there are varying levels of cache, and instruction sets available to each play into the "speed" somewhat.

[Eriamjh]

Originally posted by bma_mat99:
...i know the basics.

I'm not so sure you do...

[banninated68]

Originally posted by bradoesch:
If you know anything about cars, then Hertz is very similar to your tach. It's a measure of speed but not results. Two motors both running at 7000 rpm can have totally different outputs. Just like your example of a 2.5 GHz G5 and 2.8 GHz PC processor.

This is exactly why you can't determine the speed of a system by it's MHz or GHz.