Is the current rate of PPC CPU upgrades sustainable?

124

Comments

  • Reply 61 of 96
    nr9nr9 Posts: 182member
    Quote:

    Originally posted by Telomar

    Eh. I'm guessing you aren't actually an engineer or else you'd realise such is engineering. Often you're presented with multiple routes to take and sometimes what looks good sometimes screws up.



    Given IBM pioneered several of the most important, and difficult, technologies and processes currently used in semiconductor manufacturing and managed to bring them successfully online years ahead of competition to claim they don't know what they're doing is naïve.



    Unlike UMC and TSMC IBM only charges by the working chip rather than by the wafer so the poor yields hurt their bottom line more, except in terms of production capacity and nobody has made any great secret that the Fishkill plant has had difficulties ramping up.




    eh are you saying IBM is years ahead of UMC and TSMC? IBM is not years ahead of anyone. If anything, they are behind.
  • Reply 62 of 96
    nr9nr9 Posts: 182member
    Quote:

    Originally posted by Tomb of the Unknown

    And another thing. High school students in your country may well take higher level courses than our high school students do. But you know what? Every American is entitled to a high school education. Every single one. Unlike some countries that only allow kids who pass tough entrance exams to attend or only allow the privileged to be educated. Sure, some kids take high level courses, but the majority may never get beyond trade schools, or the equivalent of a Jr. High level.



    So take your superior attitude and shove it.




    heh you kno what, I have an american education. This is what makes me pissed off about the US education systems, seeing my peers in taiwan light years ahead of me. Who cares if every american is entitled to a high school education when they learn less than others learn in junior high?



    By the way, the majority of people in Taiwan attend high school and college and graduate school. Its a cultural difference in our attitude towards education. People in the US would get a job sooner and their parents wont even pay for their college tuition. Also, the tuition in the US is ridiculously expensive.



    In Taiwan its fair, there little opportunity for favortism in university applications. They take a test over a couple days that covers a comprehensive syllabus of what they learned in high school. If they get a high score, they get to go to a good university. You can see lots of construction workers' children going to National Taiwan University's Medical School. i dont see much of that class mobility in the US.
  • Reply 63 of 96
    nr9nr9 Posts: 182member
    Quote:

    Originally posted by Bigc

    Education does not in any way have anything to do with ingenuity and creativity. There are plenty of people with little education that create amazing products and processes.



    I know plenty of well educated scientists that couldn't solder two wires together or build a shack to live in, let alone change a tire on their car. They work real well sitting in an office working on one specific problem their entire lifetime. In many cases they don't even know where the problem they are working on fits in the scheme of things.



    ...and trolling is an interesting form of science and a waste of an education...




    Specialization is the key to efficiency. Why would u want to be a jack of all trades when u suck at all of them. Why would you want to solder two wires together or build a shack when you can earn more money writing a computer program and use that money to hire someone to solder your wires and build a shack. this is basic introductory econ.
  • Reply 64 of 96
    bartobarto Posts: 2,246member
    Quote:

    Originally posted by Nr9

    this is basic introductory econ.



    Yes it is. You are totally correct, this is introductory economics. Adam Smith advocated maximum division of labour, and if you've taken economics no doubt you've seen an Economies of Scale graph.



    However, as you delve deeper into economic theory you'll find the EoS graph is not hyperbolic (sp?), but parabolic. If EoS gets too drastic, you being to LOSE efficiency, for a whole variety of reasons. Likewise, while DoL appears to decrease marginal costs on the face of it, too much ends up decreasing it for a whole variety of reasons. *Yawn* I'm tired already.



    Barto
  • Reply 65 of 96
    nr9nr9 Posts: 182member
    Quote:

    Originally posted by Barto

    Yes it is. You are totally correct, this is introductory economics. Adam Smith advocated maximum division of labour, and if you've taken economics no doubt you've seen an Economies of Scale graph.



    However, as you delve deeper into economic theory you'll find the EoS graph is not hyperbolic (sp?), but parabolic. If division of labour gets too drastic, you being to LOSE efficiency, for a whole variety of reasons. *Yawn* I'm tired already.



    Barto




    what are you trying to imply, people should solder wires, build shacks and change tires on cars?
  • Reply 66 of 96
    powerdocpowerdoc Posts: 8,123member
    Quote:

    Originally posted by Nr9

    what are you trying to imply, people should solder wires, build shacks and change tires on cars?



    That's not he say.



    But let's imagine a company where there is ten workers : 5 are specialized in soldering wires, 1 is an electrician, 1 is a mechanician, 1 is a mason, 1 is an woodcraft, and the last make transports.



    Now imagine this : there is a huge contract and they must deliver the stuff quikly. The 5 workers specialized in soldering are very busy, the others have not a huge amount of work but canno't help, because they are not specialized. The man who make transports do nothing.

    After a couple of day the work is finish, and people have to deliver : as there is only one man who do the transports : they waste time.

    An inspector of work come into the companie, and ask to update all the electricity of the companie to suit the new norms. They will not be able to work until the job is done. there is only one worker who is busy (the electrician) the others wait.



    This is an extreme example : but if specialization is one the key for better efficiency, flexibility is the other. As these two keys are antagonists, an optimum efficiency requires a balance between the two.





    I will note also that hyperspecialisations is bad in research : people are too focused on their subject, and are not enlightened by the experience of others in other fields. In most case, a mix of an eclectic culture, with a drop of hyperspecialisation is a good thing.
  • Reply 67 of 96
    bartobarto Posts: 2,246member
    Of course not. Like I said, the effect on marginal cost is parabolic. You go too far, and workers lose touch with what is required outside their specific task, which is especially important in business today, where employees have to use their brains, and aren't just robots to program anymore (outside of dilbert).



    For instance, controlling the CPU fab as well as the CPU design lets you design both to interact with maximum utility, that is have the fab and design such as to ensure the fastest, cheapest chips. You need people to be working on BOTH for that to happen. As far as EoS goes, when you have highly, highly horizontal companies with huge EoS, they can become so big it is impossible to manage, because after you get to a certain point, where do you go except have further duplication of factories, machines etc?



    EDIT: And at a more fundamental level, what Powerdoc said.



    Barto
  • Reply 68 of 96
    telomartelomar Posts: 1,804member
    Quote:

    Originally posted by Nr9

    eh are you saying IBM is years ahead of UMC and TSMC? IBM is not years ahead of anyone. If anything, they are behind.



    I would strongly suggest you reread the post because quite clearly you haven't comprehended what I said.



    Quote:

    Originally posted by Nr9

    Specialization is the key to efficiency. Why would u want to be a jack of all trades when u suck at all of them. Why would you want to solder two wires together or build a shack when you can earn more money writing a computer program and use that money to hire someone to solder your wires and build a shack. this is basic introductory econ.



    Specialisation is the key to efficiency? Umm...no. Nice theory but try studying how the real world works some time (when you actually get to uni they might even teach you about this too if your university is half decent). There are limits to the benefits of efficiency and contrary to popular belief efficiency is often overrated and not the most important factor. Clearly you've never heard of vertical systems integration. It runs rampant through the processing industries and has for a very long time since economically it produces better results and since often you will tailor processes to suit upstream feeds.



    IBM is not a jack of all trades. They have a process division and a design division. The fact they possess both actually assists them in much of the contract work they do since they have links between the teams and neither team wants to be independent of the other. The manufacturing processes are custom tuned for the specific design and design is done with an eye to manufacturing. Even UMC and TSMC have teams for helping to integrate designs to their manufacturing processes.
  • Reply 69 of 96
    I don't think anyone is 'years ahead' of anyone in the cpu game.



    If IBM have to team up with someone to shoulder the burden of .65 then surely prudent business.



    Like Apple teaming up with IBM is good for both concerned. R&D is shared.



    Apple get access to world class chips.



    Thus a Power 5 derived 980 can't be far behind.



    No-brainer. A company like IBM has sustained chip development for years. Intel haven't and won't run them out of town.



    IBM are here to stay. As long as IBM produce goods. Apple stays.



    Incentive for IBM to do world class r n' d and Apple to sell more systems.



    The future is looking good for the next five years at least.



    If you'd posted this thread when Moto' was doing Apple Towers then my answer would have been different.



    Lemon Bon Bon
  • Reply 70 of 96
    nr9nr9 Posts: 182member
    many of you seem to imply that the concept of foundries and fabless chip designers is over specialization.



    this is not the case! the foundries make good processes give design help to adapt a specific circuit to their specific process, the chip designers manage the architecture of the chip. these are two different tasks.



    empiricaly this model has been very successful and more successful than the vertically integrated model of chip making.
  • Reply 71 of 96
    So, Nr9...any word on when G5 iMacs are coming?



    Will the 980 hit by late Summer? Will Apple call it the G6? Is it another generation ahead to justify that title?



    Just wondering if you had any other titbits other than the Powerbook CPU info...



    Lemon Bon Bon
  • Reply 72 of 96
    nr9nr9 Posts: 182member
    Quote:

    Originally posted by Lemon Bon Bon

    So, Nr9...any word on when G5 iMacs are coming?



    Will the 980 hit by late Summer? Will Apple call it the G6? Is it another generation ahead to justify that title?



    Just wondering if you had any other titbits other than the Powerbook CPU info...



    Lemon Bon Bon




    sorry i only have reliable information for the powerbook g5
  • Reply 73 of 96
    Quote:

    For instance, controlling the CPU fab as well as the CPU design lets you design both to interact with maximum utility, that is have the fab and design such as to ensure the fastest, cheapest chips. You need people to be working on BOTH for that to happen. As far as EoS goes, when you have highly, highly horizontal companies with huge EoS, they can become so big it is impossible to manage, because after you get to a certain point, where do you go except have further duplication of factories, machines etc?



    Aren't you essentially advocating a Fordist model of production when most modern production systems have been based on outsourcing production to specialists. cf see Apple.



    The ability to control fab and design has done very little for Moto, where the design ability fars outstrips fab capabilities. Also I'm not sure how you're arguing this for IBM, when in fact there are many specialist vendors working directly with IBM in East Fishkill, and previous plants, on the fab process.
  • Reply 74 of 96
    Hi there.. I live in Sweden where everyone, i mean EVERYNONE gets a university degree if they are qualified. Tuitions? No.. Universities are completely free, and so are child care from years 0 to 6, and then there are even state funded school-pay (up to $200 a month) until you graduate university. How's that?



    Anyways.. All we ever see of Taiwan is cheap litte plastic toys that come in McDonald's Happy Meals.
  • Reply 75 of 96
    powerdocpowerdoc Posts: 8,123member
    Quote:

    Originally posted by Henriok





    Anyways.. All we ever see of Taiwan is cheap litte plastic toys that come in McDonald's Happy Meals.




    OK some comments of NR9 where not cool, but it's not a reason to start a flamewar : it's not AO here
  • Reply 76 of 96
    bartobarto Posts: 2,246member
    Quote:

    Originally posted by cowerd

    Aren't you essentially advocating a Fordist model of production when most modern production systems have been based on outsourcing production to specialists. cf see Apple.



    The ability to control fab and design has done very little for Moto, where the design ability fars outstrips fab capabilities. Also I'm not sure how you're arguing this for IBM, when in fact there are many specialist vendors working directly with IBM in East Fishkill, and previous plants, on the fab process.




    I'm saying you can go too far. The PowerPC 970 was designed to be produced at Fishkill. If it was designed to be produced anywhere, it would have become a slower, less capable CPU.



    Nr9 is arguing that fabs and CPUs should be totally seperate to ensure maximum DoL (and presumably EoS through increased use of captial in a single area). That's fantasy for high-tech production. When making CPUs, there needs to be very strong collaboration between the CPU designers and the fab to ensure the fastest, cheapest chips. Owning both the fab and the CPU design can only benefit IBM due to the increased collaberation AND it's in their interest to have one optimized for the other.



    Barto
  • Reply 77 of 96
    amorphamorph Posts: 7,112member
    Can anyone name one foundry that has not had teething problems on 130nm?



    The simple fact is that 130nm is where current technology and tools start getting stretched, and 90nm is everything 130nm is, only worse. So, yes, IBM's had trouble at Fishkill. So has Intel, and Texas Instruments, and STM, and TSMC.



    IBM might have promised more than they ended up delivering, which is a mistake (and unlike the post-antitrust-suit IBM to do), but nobody who's trying to push the industry forward (which means anyone who isn't still on 200mm) is having trouble.
  • Reply 78 of 96
    Quote:

    Originally posted by Amorph

    Can anyone name one foundry that has not had teething problems on 130nm?



    The simple fact is that 130nm is where current technology and tools start getting stretched, and 90nm is everything 130nm is, only worse. So, yes, IBM's had trouble at Fishkill. So has Intel, and Texas Instruments, and STM, and TSMC.



    IBM might have promised more than they ended up delivering, which is a mistake (and unlike the post-antitrust-suit IBM to do), but nobody who's trying to push the industry forward (which means anyone who isn't still on 200mm) is having trouble.




    IBM has been delivering just fine. Original expectations was a 1.8 GHz G5 to start, but they over-delivered 2GHz. And now only one month after Apple was able to catch up to the overwhelming G5 2GHz demand, Apple and IBM are able to release dual 1.8 GHz G5 PowerMacs at a more attractive price point and lower the price of the 1.6 GHz G5 into more of a sweet spot. The only reports of problems are unconfirmed rumors about nVidia, which as it turns out IBM and nVidia is delivering the 5700 GPU out of the fishkill plant.



    So the reality is that IBM's Fishkill plant is delivering. In fact, it has become common knowledge that Apple engineers have been given 90NM 970 processors and 980 processors to design around. So, the actual deliverables are indcating that everything is going very well.
  • Reply 79 of 96
    wizard69wizard69 Posts: 13,377member
    Yes the future does look good and I do hope that Apple implements those parts as soon as possible. The interesting question is where are those parts going. I still have this sneaking suspicion that the "980", as rumored, is destined for the low cost line and is really a Power 4 derivative.



    The only issue I have is that the problem with the 1.6 and 1.8 GHz machines is that they were grossly over priced to begin with. I do not believe the reconfiguration has anything to do with the production output at IBM. People simply don't want the 1.6 GHz machine because it was over priced and half the machine of the others. Going dual on the 1.8 GHz machine just gives apple a way to get rid of excess processors and change the value equation a bit.



    In any event I hope that we aren't underwhelmed with the new intros at the begining of the year. I could just imagine the screams if another round of G4's are staged. Of course an upgraded G4 done right would not be bad either.



    Thanks

    Dave





    Quote:

    Originally posted by stingerman

    IBM has been delivering just fine. Original expectations was a 1.8 GHz G5 to start, but they over-delivered 2GHz. And now only one month after Apple was able to catch up to the overwhelming G5 2GHz demand, Apple and IBM are able to release dual 1.8 GHz G5 PowerMacs at a more attractive price point and lower the price of the 1.6 GHz G5 into more of a sweet spot. The only reports of problems are unconfirmed rumors about nVidia, which as it turns out IBM and nVidia is delivering the 5700 GPU out of the fishkill plant.



    So the reality is that IBM's Fishkill plant is delivering. In fact, it has become common knowledge that Apple engineers have been given 90NM 970 processors and 980 processors to design around. So, the actual deliverables are indcating that everything is going very well.




  • Reply 80 of 96
    cowerdcowerd Posts: 579member
    Quote:

    IBM has been delivering just fine.



    Well they apparently bailed out on SiLK at 130nm because of big problems.
Sign In or Register to comment.