How are Intels locked?

entropyentropy Yah-Der-Hey (Wisconsin)
edited April 2004 in Hardware
I've been wondering this...as i don't really understand why a lower multiplier is better, i guess i'm just curious lol. But, since intel sorts their cpus according to what they can or can't handle, and let's use the C line for example, they all have the same fsb. so that isn't what they change. what they change is the multiplier. now, there's no obvious way to do this externally, and they all start out with the same multiplier. they test it by running it, seeing where it stands, and dropping the multiplier. probably all done by computers, but still done. so the only way they can change the multiplier is internally...now what i'm wondering is, if intel can do it, why can't we? from what i'm seeing all they do is give the cpu some sort of command or shock (very carefully controlled perhaps) that tells it or forces it to drop the multi. ok, if they shock it i can see where perhaps they sever some connection to do it, but if it's code, couldn't we figure out some way to repeat the process, except reverse the change? now, anyone correct me if i'm wrong cuz i just thought of this using what i gathered off forums and such. anyone willing to explain why a lower multi is better, feel free as well. thanks!

Comments

  • mmonninmmonnin Centreville, VA
    edited April 2004
    They are multiplier locked. The latest AMD Barton procs are as well.

    Lower muiplier is better because that allows you to run a higher FSB and keep the same CPU speed.

    100*20 or 200*10. The RAM is running twice the speed in the second instance. The whole system speeds up yet the CPU is still the same.

    Some Engineering samples are sent to reviewers with an unlocked multi so it is possible.
  • entropyentropy Yah-Der-Hey (Wisconsin)
    edited April 2004
    i know it's multiplier locked ... but how? it has to be internally and it has to be changeable at one point or another, and if it was changeable at one point, some reverse engineering has to be able to make it changeable again!
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited April 2004
    It's hardlocked at the microscopic level.

    //EDIT:

    Before someone tries to call my bluff, I'll explain why this is not only plausible, but probable.

    The circuits for both the Pentium 4 and the Athlon series are fabricated on the 130 nanometer process. Any metallic trace that must interface with the core of the CPU is a 130 nanometer, or finer, wire. That said, both the Athlon and the Pentium 4 have traces running through the CPU to control multiplier. In order for said wires to fit in the core, they must be fabricated on the aforementioned 130 nanometer process which is smaller than fractions of the width of a hair.
  • edited April 2004
    Actually, in the case of the later Pentium 2s, it was just a case of making one pin unreadable to the motherboard (by painting it with nail polish). I would imagine that messing with the pins on any of these newer processors would unlock them as well. The key is finding which pin(s) to mess with, and how.

    You have to keep in mind when the cores are produced, they (the producer) don't know what speed the core would be running at. For instance, in the case of the 1800+es and the 2400+ CPUs, AMD only wants to make one core so they could cut manufacturing costs. At the same time, they want to be able to use these same cores sold as different speeds, thus charging different prices and filling different consumer levels, all with one product. So, they take a core from the pile, stick it on the organic (or ceramic) package they want to use, and the package itself would "tell" the core what speed to run at.

    Now sure, the conductors actually interfacing (touching) the core will be only 130nm (assuming the 130nm process isn't JUST used within the core, and not the packaging interface), but once the conductors are traveling within the packaging, they are most certainly larger, so they can do necessary things like make a large enough contact with the SM caps and resistors, and in some cases the pins to interface the package with the motherboard. All one would have to do to make contact with these conductors (using equipment which is available at the consumer level) would be to connect to them to either the pins, or the SM components.

    Oh, and let's not forget that it'd be much, much cheaper to use larger conductors on the package, especially when they don’t really need to. We all know they want to keep the manufacturing process as cheap as possible.


    So I'm not saying you're wrong, but don’t count on being right.
  • edited April 2004
    I remember reading on the hardforums a couple of years ago when they had some really sharp folks posting in the Intel section that they actually tested all of them unlocked, then somehow they blow a fusible link in the die itself for locking the proc. Don't ask me specifics because this was way over my poor head, but I did get the gist of the conversations about this.
  • SputnikSputnik Worcester, MA
    edited April 2004
    what muddocktor is talking about sounds like PLD or PALs if anyone knows what they are. that doesn't seem unreasonable at all from cerntains perspectives, but it'd be kinda difficult at the 130nm level, then again, i've only used chips using around a 650nm process
  • qparadoxqparadox Vancouver, BC
    edited April 2004
    Its possible to have a blown link, the easiest way, however, would be to simply lock it using a laser. Believe it or not DRAMs are almost all manufactured faulty. After testing of the DRAM chips, lasers are used to link only working cells to the address and data bus (or remove bad ones depending on the process). Don't believe me? Check out this: (PDF 1MB)
    http://www.gsilumonics.com/process_online_ordering/pdfs/108.pdf

    It'd would be trivial and cheap for intel to directly make new links on the chip after manufacturing to effectively make the multiplier truly locked. AMD tries to minimize costs they they have changes made on the external packaging which are hackable.
Sign In or Register to comment.