[Newbie2Mac]

Hi, I know that a 1Ghz G4 Mac has better performance than a 2Ghz Pentium 4 PC, but could anyone explain how that works, that half the processor speed is a better processor?

[SeaFox]

Because the Mac's processor does more than twice the work of the Pentium is how the story goes. Think of the Pentium as a Ferrari and the G4 as a dump truck. Both have to move identical piles of sand across the desert. While the Pentuim can drive a lot faster, the G4 can take a lot more with it on each trip back and forth.
I don't know if this is the answer you wanted. Someone will come in here with an explanation of piping, AltiVec, stages, ect. at some point. But that's the gist of it.

[Grant_G]

Also RISC vs. CISC is a major part of it, but someone with more knowledge of the subject than I will have to explain it all.
G

[iloveitaly]

Well lemme ask u a question. What is the difference from a pentuim, and a pentium II, a pentuim II from a pentuim III, ect,ect. would a 2GHZ pentuim be faster than a 1GHZ pentuim II? Basically Macs use power PC chip architecture which from what i understand is more efficient than the x86 architecture, which is what most PC's use. Also, I think some of what the other guy said about dump trucks adds onto too, dont take it from me I dont know much about Chips and stuff, all I know is that my g4 450 is faster than my 1.3GHZ PC.

[SeaFox]

In reply to:

---

Well lemme ask u a question. What is the difference from a pentuim, and a pentium II, a pentuim II from a pentuim III, ect,ect. would a 2GHZ pentuim be faster than a 1GHZ pentuim II? Basically Macs use power PC chip architecture which from what i understand is more efficient than the x86 architecture, which is what most PC's use. Also, I think some of what the other guy said about dump trucks adds onto too,

---

I have no clue what the differnece is between P2 and P3's. And I was the one talking about dump trucks. I think you meant to direct this toward Grant.

[Azzgunther]

Apparently the size of the pipeline in a processor is very important. My understanding is that you can describe the pipeline size as the "amount of distance" something has to travel.
Thus, a 1ghz system's G4 chip has a pipeline of 8 and a Pentium 4 chips is like 20, over twice as large.
While a vague basis for you to compare with, this is all from something I read but didn't pay much attention to a few months ago. I'd like someone to expand upon and elaborate on what I've said.

[tahoe3]

A P2 at 233mhz is a little over twice as fast as an original P-233. A P3 is similar to a P2, but takes advantage of advanced instructions set. Between a 2ghz Pentium and a 1ghz P2, My money would be on the 1ghz P2 although not by much.

[stockscalper]

The Pentium is a very inefficient chip with long pipelines where data tends to get congested. It needs very very fast processor speeds to function. To do any instruction takes the Pentium 20 steps. Even if an error comes up and the instruction is not completed it still has to complete the 20 steps before it can take on the next instruction. The G4, in contrast, completes instructions in no more than 7 ot 8 steps. In many cases, depending on what is being done and whether Altivec is involved, the instruction cycle can be completed in 2 or 4 steps. So, you can see if the instruction was to go and fetch a glass of water the Pentium would have to take 20 steps whereas the Mac would take 8 or less. So, to be functional the Pentium has to walk faster and complete its 20 steps in the same relative time as the Mac's 8 (or less) steps.
Intel has done a wonderful job of hyping their products and making the public think they really have something special going on. They give catchy names to their products like "Speedstep" technology. Wow, sounds like it really does something doesn't it? All it is is a mechanism to slow down processor speed to save battery power and something every Mac laptop has had that I've owned for the past 9 years. And now their laptops come with Centrino technology. You would think they have some special chip in them, but Centrino is a marketing name which means nothing more than a laptop equipped with the mobile Pentium chip plus a wireless card. Gee, hasn't Apple had wireless laptops for years? Apple should take a cue from Intel and come up with some catchy marketing buzzwords for it's technology. If Intel were marketing cars they would have you believe that a four cylinder VW beetle is faster than the 8 cylinder 500 HP Corvette because the VW ran at higher rpm's (mz in computers).

[d00d]

The problem with pipelines is not so much the fact that steps are subdivided (because for as many stages there are, that many instructions can be in the processor), as branch instructions. Basically, processors with pipeline stages guess what the next command is and begin processing it before the last has finished. If the instruction is a branch, it takes several pipeline stages to be realized and the entire pipeline must be cleared (all stages) before it can be loaded up with instructions again. Also, sometimes instructions get "stuck" on a certain stage if they rely on the results of a prior instruction and the result of that instruction hasn't been realized yet. Processor makers have been working hard on making processors that are good at branch prediction and other ways of avoiding these problems, and as such, the problems caused by this are minimized. Compilers are designed specifically to deal with the latter problem (this is optimization) and have gotten fairly good at it. PowerPC's strength also relies on the number of different processing pipelines (forgive me if I'm using the wrong terminology, it's been a while since I took the class in this). There's one for integers, one for floats, etc. last I knew. This allows for multiple instructions to flow through the processor at the same time (not in a serial fashion, like pipelines) so one instruction in one pipeline can't block another instruction in another.
At any rate, processor performance always relies on the metric used. Dependent on what you're testing, you get different rankings for different processors. As far as I'm concerned, the best thing to do is try out the machines you're interested in and decide which to get based on which one yields the best productivity (that's my personal metric).

[iloveitaly]

I was asking the guy who posted the post not u. I know I clickeds reply to seafox, but I really didn't are what I clicked just wanted to reply.

[Nobody]

PC's are better at some things, Macs are better at others. There is probably no way to predict which one will handle a certain process, application, etc. better than the other. Basically just experimentation. I don't do any of this experimentation however. I just buy the Mac /forums/ubbthreads/images/graemlins/grin.gif don't you?

[AKS9]

http://www.apple.com...myth/index.html
This quicktime video explains the pipeline thing. I don't really know anything about processors and stuff like these guys on the forums, but this video helps quite a bit.

[bracken]

Very informative articles:
Understanding Pipelining and Superscalar Execution
The Pentium 4 and the G4e: an Architectural Comparison
The Pentium 4 and the G4e: Part II: the Execution Core
G4 vs. K7: an architectural comparison
RISC vs. CISC: The Post-RISC Era

[digedit]

Well, for one thing a 1 Ghz G4 is not faster than a 2Ghz Pentium 4 except in very rare instances, but I won't dwell on that, the important thing to understand is that a 1 Ghz G4 IS faster than a 1 Ghz P4 if there was one.
DO NOT use any Apple propaganda to try and understand this. It's obviously lopsided in it's view and the video and speech linked to from above are very carefully crafted to mislead you. First of all the speech is comparing an 867 G4 verses a 1.7 P4 and the accompanying video depicts two processors of the same clock speed giving the fasle impression that the top is the 867 G4 and the bottom a 1.7 P4. Also not mentioned were the specific instructions used for the comparison. Also remember that the pipeline only really plays a major effect when branch prediction is guessed wrong and the pipeline needs to be emptied. Given intels advanced branch prediction used in their chips and optimized compilers this is not as big a problem as you might believe. Also, many of the most demanding cpu tasks are DSP type in nature, meaning that the processor is processing long streams of data and which makes braqnch prediction less of an issue as well. Note that since the video above, that Motorola has increased the lengths of their pipelines in newer cpu designs as well.