1. Welcome to the new MacRumors forums. See our announcement and read our FAQ


Discussion in 'Hardware Rumors' started by samir, Sep 18, 2001.

  1. macrumors newbie

    i dont know if i can trust on jobs ,i learn the explination about the mhz mith so i can shut up my friends mouth that use pentium and now i read about g5 with 10 pipeline stage and a high frecuency ,if i am wrong and its design is more complex than the frecuency and pipeline just tellme because i am not sure now .DO YOU GUYS "THINK...THAT IS BETTER WAITH FOR THE NEW G5 64BITS COMPUTERS COUSE I READ THAT THIS APPLE 32bits COMPUTER GENERATION IS DYING...... I WAS DYING FOR A IMAC BUT I RATHER NOT EXPEND MONEY ON A DYING TECNOLOGY (SORRY A LOT OF DYING WORDS

    [Edited by samir on 09-18-2001 at 03:19 PM]
  2. macrumors 65816


    Who says its dying

    The techology is not dying... in fact it is growing. Apple was in some trouble like 5 or 6 years ago but they habe made one hell of a comback. They have new markets springing up every were and they are taking back education. Which is the market that put them on the map in the first place. It's up to you if you want to wait for the G5. I just got a G4 Quicksliver and now we are looking at the possibility of a new CPU in a few months. But those are just rumors for now. As for the Apple Generation... when Apple takes education back... there are going to ba a lot of young people that are going to want one of their own at home. i.e. long time customers.... I am almost 22... I have always been an apple user since my first Mac+, to the IIvx, rev-a iMac, G4 Quicksilver.

  3. macrumors 68020


    There's no conflict

    Using MHz as an indicator of raw speed is still a flawed argument, but the clock rate does still have a significant influence on overall performance. The reason you're concerned is that you have made up a whole new myth. We can call this the "pipeline premise." The pipeline premise suggests that shorter pipelines are always better.

    Of course, it's easy to see that this isn't the case. Superscalar (pipelining) architecture was one of the major advancements that made our modern, zippy CPUs possible. There is no miracle of technology that would ever make it possible to clock an i386 or a 68040 up to gigahertz levels.

    A canonical CPU has four stages: fetch, decode, execute and store. With no pipelining, only one instruction is in the queue at a time. With a pipelining architecture, when the first instruction moves onto the "decode" stage, the next instruction is queued up right away. In the best case, every stage of the instruction queue has an instruction to work on at every clock cycle. The difference in performance is the same as the difference between just hitting a home run and hitting a home run with bases loaded.

    There's a minor (but tolerable) problem with this approach: It's not always possible to tell which instruction is the "next" instruction. If the first instruction is a conditional operation, there are two candidates for "next," and you can't know for sure until the first instruction has gone through the "execute" stage. The way this is handled is that the CPU and the software compiler cooperate to make a fair guess which instruction will be next. When you see the term "branch prediction," this is what is meant. Hopefully they pick right and the pipeline keeps chugging away. If they choose wrong, however, then all the work that has been done in the queue behind the first instruction is useless and must be discarded -- the entire pipeline must be flushed. Every time this happens, you've got to refill the pipeline before you start getting results again, and that takes a lot of time. The performance gain from good guesses, however, more than offsets the penalty for guessing wrong, and you're still doing WAY better than older processors that only worked on one instruction at a time (which, in superscalar terms, is like failing every branch prediction you make).

    Another problem you run into is that your pipeline is only as fast as its slowest stage. In the four-stage processor described above, this is usually the execute stage. If your instruction takes eight clocks to execute, then everything behind it is going to sit idle for eight clocks, even if they all finished in one. The way we combat this problem is to break down the monolithic stages into much simpler substages that take as little time as possible. If your eight-clock execute stage can be broken into eight separate single-clock stages, then the following instructions will never have to wait. Having simpler stages allows you to pulse the clock more frequently because the simpler operations may be performed faster.

    The big drawback here is that the longer your pipeline is, the bigger the performance hit you suffer when you have to flush it. Every time a 20-stage Intel chip misses a branch prediction it takes 20 cycles before the chip starts producing results again, because it has to refill the pipe.

    So here's a very rough and dirty comparison: The current G4 has 7 stages. With a full pipe, instructions are completed on each cycle. If the pipe is flushed, it will take seven cycles before you see more instructions completed. If pipelining were turned off completely, the effective speed of a 700MHz chip would be 100MHz. That's the absolute worst case.

    Now consider a G5 with a ten stage pipeline. At 1GHz (1000MHz), the worst case (no pipelining) would result in... 100MHz. To recap, a 1GHz 10-stage chip has similar worst-case performance to a 700MHz 7-stage chip. It's worth noting that the 1GHz will still have better average performance than the 700.

    See, it's all a balancing act. MHz alone cannot determine a processor's performance. Nor can the length of the pipeline. In fact, the combination of the two is still dependent on the efficiency of the branch prediction, which is a much too complicated topic for either Motorola or Intel's marketing literature.

    But note that all of the proposed G5 speeds are comfortably above 1GHz. Definitely a fair trade. That 1.6 is gonna rock.
  4. macrumors 68030


    why did you type all of that? you started out sounding like you had a point that would be controversial, but just ended up correcting yourself and saying what we new all along, shorter pipelines are more efficient than longer.

  5. macrumors 68020


    If that's what you got out of it, then you missed the whole point. The G5 has more pipe stages than the G4, and that granularity allows for higher clock frequencies. What you seem to have missed is that the clock frequencies the G5 is able to support are disproportionately high when compared to the increase in pipe length. When you do the math, even given the longer pipe the G5 is more efficient than the G4.

    That's what I was saying about balance. Intel went overboard. They kept adding stages so they could keep bumping MHz, because to Intel the clock speed is just a marketing point, and by overusing the pipeline trick they sacrificed efficiency to keep that number moving upwards. That is the heart of the "Megahertz Myth." The real "bad guy" behind it is neither high clock speeds nor longer pipelines, but Intel's ham-fisted marketing-driven engineering practices over the past five or so years.

    Exposing this myth is, of course, in the interests of Apple's marketing efforts. But unfortunately, a lot of people have come away with the idea that clock speed means nothing and that a longer pipeline is always worse than a shorter one. That's really only a mirror image of the original fallacy. As we saw from the original post in this thread, this marketing tack is coming back to bite Apple because now people who don't really understand how this stuff works think that the G5 is somehow worse than the G4.

    Motorola has done an excellent job with the G5, adding stages very conservatively, and milking the maximum performance from every additional stage. Frankly, it's a fairly impressive feat of optimization.
  6. macrumors newbie

    P4 pipe's

    According with arstechnica article (www.arstechnica.com) titled "The Pentium 4 and the G4e: an Architectural Comparison", we have a complete analisis of the P4 pipeline structure, in this analisis you can read this:

    Stage 5 - Drive: This is the first of two of Drive stages in the P4's pipeline, each of which is dedicated to driving signals from one part of the processor to the next. The P4 runs so fast that sometimes a signal can't make it all the way to where it needs to be in a single clock pulse, so the P4 dedicates some pipeline stages to letting these signals propagate across the chip. I've actually never seen a "Drive" stage in a pipeline before, and neither has anyone else whose P4 write-ups I've read. I think this may be a first. It's definitely there because the P4's designers intend for it to reach such stratospheric clock speeds that stages like this are absolutely necessary.

    No words needed.
  7. macrumors 6502

    Gelfin has a good point.

    Apple has gone from 4 (I heard the G3 is a 4-stage) to 7 to 10, but always with good reason. Apple seems to increase their pipeline stages by 3, and tripple their real-world performance per Mhz, hmm, seems like a good (and ironic 3-3) tradeoff. You see to counter the event of a pipeline flush a faster frequency is required (yes I know you know) but the trick is to keep a proportional line at the most. What I mean is never loose performance overall. Some tasks can predict the next step quite acuratly, and benifit a LOT from high Mhz. I've heard P4s start up in 13 seconds for example. Apple wants to keep their pipeline low and increase Mhz, but ONLY if they can show that the performance per is better. Perhaps the G5 can predict the next step better than the G4, I don't really know. But having a 64-bit main-CPU will definatly kick-butt. 256-bit AltiTivek would scream. Quad cross-bars can truly almost qudruple performance.

    Between faster Mhz, wider registers, better instuctions sets, and multi-core processing, the G5 will EASILY make up for 3 pipelines. Now all Apple has to do is convice us the Mhz isn't *That* Bad.
  8. macrumors regular

    First...for all of you to know....If you ever see the Gestalt number...that is actually your "guessing" mechanism. It looks at the whole process instead of parts of it so that it can guess where to send it. I have never seen the term Gestalt used in Pentium or AMD chips, though (probably never decided on a name for it).
    Now, dealing with the former post about a clearing of a pipeline when a guess is wrong:
    G4 866 2/3 MHZ chips in a 7 stage misguess: 123.8 MHZ effective clock speed for that instruction
    P4 1,866 2/3 MHZ chips in 22 stage misguess: 84.84 MHZ effective clock speed for that instruction
    (does anyone know the stage # for the Athlon and P3 and G3 and Celleron or anyother processer?)

    Just remember folks....the processor is not everything and Apple is falling behind in bus speed, Ram speed, and other things, too. Lets just hope that they can come up with a major breakthrough soon.
  9. macrumors 6502


    I've heard that Apple doesn't plan to increase to ATA/100, DDR-Ram, or 266Mhz Buses until they feel it's vital. Apple is lazy abut these things, they only added 133Mhz bus to compliment the GeForce3 w/AGP4. Apple might release a new G4 with a 266Mhz bus/ram and ATA/100 -or- SCSI160 standard drives if the G5 isn't ready soon enough. Just a guess though.


    PS: I just got 10.1 and it ROCKS!!!
  10. macrumors 68000


    Re: Rumors

    I'm honestly not as impressed with 10.1 as I thought I'd be. Oh well.

    Anyway, regarding the topic at hand, I think you'll see DDR-RAM in the next generation of Macs...I think that it really *is* a necessity. You'll never see SCSI inside a Mac again, though. Those days are just gone.
  11. macrumors 6502


    I noticed a MUCH bigger speed increase in my G4 vs my G3, I think a lot of code is altivek-enabled now. But X.1 aside, SCSI is still big for apple, they'll probably kick it out ASAP when FireWire2 is around, but I've hear that 10.1 has ATTO RAID for SCSI, I don't know what that means, but it's supposed to be better.
  12. macrumors newbie

    Re: Who says its dying

    I hope you're right, but if Apple doesn't address its weakness in gaming, there won’t be an "Apple Generation", regardless of how well they couple to the educational market. Kids are adaptable and don’t think twice about switching platforms. If you can’t play cool games then having a Mac isn’t cool.

    I’ve been a Mac user for 15 years. My kids (ages14, 12) were raised on Macs, both at school and at home. They are constantly bombarding me with requests to buy a PC.

  13. macrumors 6502


    I think PCs are like smoking, if you give a 8 year old some drugs and he vomits for a week, he'll have a very small chance of becoming an adult smoker. If they are clean until they're 19 say, they could become a smoker much more easily.

    If a kid sees how often PCs crash, and how most of the PC-only games suck, they'll think twice. I also figured out a common misconception: "Macs are bad for gaming." macs are great for gaming but: 1) Less games 2) Most are ports, port run off PC code and thus can be slower. OK not PC code, but save and read PC documents and were designed for PCs. I'm really hoping the G5 comes out soon, between 800Mhz+ G4s and LCDs in iMacs, Apple could really market it as a powerhouse computer. Remember a few years ago, people would kill for a G3 as fast as the iMac, now it's like a joke in the pro-market.
  14. macrumors 6502

    Then again I'm the Apple poster child.

    I think it's fun to go into CompUSA and bombard customers with facts and stats that are not in favor of Wintels, shal we say.
  15. macrumors regular

    I'm applying at Office Depot. Think of the damage I could do to the people there. Their best selling machine is a $300 e-mail machine with a windows-like interface. And best of all....I talked with one of the guys working there and he went to a confrence on WinXP and said it REALLY BLOWS. You can only upgrade your system 3 times (installing drivers & such) before you call Costomer Service and get a new key-code....you even have to give your SS# to get the code in the first place. Big Bro looking out for you or what?!
  16. macrumors 65816


    Re: Re: Who says its dying

    They probably just want games. There is one thing that Windows has and that is more game titles. Get Vertual PC so they can play the same games their friends have.. and they get to see why the Mac is better.. it can do it all. I have been slowly converting all my friends to Mac. The latest on was a hard core Win head... then he got a job that enjoled web desine... his compainies lab had Mac's and PC's for him to use, but the end of the summer he was using only the Mac's Now he want one of his own. He's bee asking me what he should get.
  17. macrumors 65816



    Be a Mac missionary. Conver convert convert!!!!

    I hate going to stors that sell computers. The "tech guys" that work there know little to nothing about anything. I probably know more about windows than they do... you can emagin how much they know about Mac's. They proably have never heard of UNIX before. You have to educate the public.
  18. macrumors 604



    Use At you Own Risk *******
    Hack Dell's site and then replace it woith apple's site@!!!!
  19. macrumors 6502

    Re: Or....

    I would If I could, except instead of a 10.1 main-link, I'd put up the G4 page (the Jpg that shows the 733 outperforming the 1.7Ghz P4, for 1/2 the cost).

    I'm looking forward to the G5, not actually for the G5 as much as G4 iMacs. Imagine if Apple did a medioquer Mhz update, but switch to G4s. Ahh 600/700/800Mhz G4 iMacs

    Think Different
    Think Q2 2002
  20. macrumors member

    Just a quick post from a Windows guy. I use XP. I like it. I know 12 other people who use it. All of them love it. Of the people I know who use it (including mac people) NONE of them dislike it, except one, who had hardware problems that turned out to be the harddrive, not the OS.

    But anyway, I was just wondering why there is such a big need to "convert" people to mac? What's the big deal? Use it, enjoy it, get on with your life.

    Oh, and for god sake, please stop using jpegs from apple.com as proof of your superiority. It's not exactly reliable.
  21. macrumors 68000


    Buddy, the comment, "It's not exactly reliable" is not an argument. If you want to make a point, you have to do a lot better than that.

    More to the point: Mac users like to convert because we are better people. We do this out of the goodness of our hearts. Think of us as the religious missionaries of Silicon Valley. Moses parted the seas, and we shall divide and conquer the Wintellians.
  22. macrumors member

    "Not exactly reliable"

    Why is that not valid? Are you saying that using figures from apple's marketting department are in no way edited to put their product in a good light compared to the competition? A HA HA HA HA HA HA HA HA!

    I'm not critisizing Apple here. What do you think marketting is for. Downplaying your weaknesses (yes, they exist) and overselling your strengths (and yes, they do have them too).

    Apple, Intel, Microsoft, McDonalds, AMD, Ford, I don't care. They're all the same. When you go to Ford and ask them what the best cars are, guess what they're going to say? No, really, guess...

    What you're saying is that it's a valid technique to hack Toyota's website and post a picture from Ford saying that "We're the best cars in the world". That's evidence to you?

    To get a bit more back to what you probably meant, I still claim that saying "It's not reliable" is completely valid. It's not. Do you want me to post specific refutations to specific claims? I don't have that information. I wish I did. I'd love to know some facts. Not so I could go on a McBashing Spree. Just because I hate the fact that Mac people seem so damned selective in their statistics.

    [note] I really didn't come on here to troll or stir up trouble. Apologies if that's how it seems.
  23. macrumors member

    Follow up

    And by the way, what if us Wintellians don't WANT to be converted? Ministry by the sword, if need be? : )
  24. macrumors member

    Scuse me, Wintellian Here

    Hey fellas....

    As a FORMER Wintellian (hehe, funny name that) and a mac lover for 6 months now (hey aint that a record) i would like to say that being educated was one of the best things to happen to me. I in fact did the educating myself and because of the financial status of my school have managed to get another 7 iBook 2001's sold. Fantastic for me as i have the original, all scratched and loved, and more fantastic for apple. Now, i am waiting for that check from them for those sales. yeah right!

    anyway. just to say, there was not one person out of those 7 who considered getting a mac at anytime. One was my girlfriend who rolled her eyes at me getting a laptop, let alone a mac. She said 'I will have nothing to do with it'. Yeah right. Day i got it a tentatively showed her and she instantly liked it. Wow. What a reception for something that she so thoroughly disliked.

    Morale of the story....Educate all people. They make the choice don't they? I think the wintel community at large does not know of Mac OS X. Educate them, tell them and maybe they will change. In my case, most likely they will change.

    I think all that needs to happen for Apple to succeed is to simply tell everyone about their products. There are bound to be many people out there who would change if they knew, including me. In fact, to show my support i managed to not use Windows or an intel box in 1 month. I have had to stop that recently, as my parents complained that the network failed. Funny thing was DAVE said the network was fine, Windows said there were not any other computers on the network.

    Well, thats my first post. Thanks for reading and i hope to be around more often on my iBook.

Share This Page