KT Multiplier Issues 527

Multiple Issues Mystify

During the time the KT133 and now the KT133A chipsets have been around, we have either run into or at least heard about a number of problems people have been having setting the multiplier.

These problems have been occurring across the board. No board is completely immune to this. I’ve seen some comments in some forums claiming that Abits and Asuses are immune to this, like hell they are.

On the other hand, this problem isn’t endemic, either. For every board with this problem, several under the same circumstances work perfectly fine.

Over the past couple months, we’ve thrown out a number of theories as to why this might be happening, and suggested corrective action. We have found that any of these approaches work sometimes, and not others.

At this point, we’ve concluded the following:

  • There is no one cause to all this. A particular problem may be due to one of a number of causes.
  • It is very likely there is some inherent weakness in the Via chipset. We don’t know exactly what.
  • We are uncertain whether or not these problems can be addressed (or worsened) by the mobo manufacturer. Some seem better, some worse than others, but we suspect our information may be skewed.
  • We suspect there is likely to be some problem with some AMD CPUs.
  • We think it very likely that many of these problems may be caused by the interaction of multiple factors (i.e., a particular CPU may provide a little less voltage to transmit the multiplier signal and a particular mobo might need a little more voltage to read the signal).
  • This demonstrates where the review process fails. Not all equipment is created equal. We or Anandtech or anybody else can test a board, and we get no hint of a problem. You buy the board, and it just won’t work. Or vice versa. We’re not lying dogs, you’re not lying dogs.

    Keep in mind what the equipment manufacturers do as opposed to what we do. The equipment manufacturer’s goal is to make sure that all equipment that goes out the door meets certain minimum standards. They can’t test every single board for three days for every contingency; they do some rough and ready tests to weed out obviously bad items, and that’s that.

    We overclockers, on the other hand, do not run our equipment the way the manufacturers want. That AMD 1200Mhz CPU is supposed to be run at 12X100, not 9X133. No manufacturer is testing to see if their product can do what we want it to do, nor weeding out those products that don’t meet our standards.

    We hold a higher standard, so it is perfectly normal for equipment to vary in performance a lot more than it would if we ran strictly at spec.

    So don’t ever, ever think that just because person A can do something, you automatically can, too, or vice versa. There are no certainties, just probabilities.

    What Can I Do About It?

    We’ve come up with a checklist of solutions that at least solves some people’s problems, ranging from the simplest to most extreme actions.

  • If you connected the L1 bridges, improve the connection Take another look at your CPU and inspect your pencil work with a magnifying glass. Many people find they didn’t do as good a job as they originally thought. Pencil traces can also easily wear away (see below).

    What kind of pencil did you use? The “harder” the pencil (#1 is softest, #3 is hardest), the poorer the conductive connection. If you used a hard pencil, or don’t know what pencil you used, wipe away the earlier pencil markings, and redo them with a soft pencil. Keep in mind that soft pencil traces can easily wear off, so if the problem recurs again after a while, you’ll probably have to redo it.

    If this doesn’t work, use a better connector than a lead pencil, use a conductive ink or conductive paint to connect the L1 bridges.

  • Set your equipment first to 100Mhz Don’t be impatient. If you want to lower the multiplier so as to run a higher FSB, don’t go running at 133Mhz or better first. Set your machine at 100Mhz, and lower the multiplier at that setting. Cold boot the machine a couple times until you are sure the lower multiplier has “taken,” then try for 133Mhz or better. Do NOT do things like flashing BIOSes until you are already running at a lower multiplier. If you do, you may lose lower multiplier options.

    Find out where your CMOS reset jumper is, and use it at the first sign of trouble. Your number one priority is to keep those lower multiplier options available.

    You may find that your BIOS gives you no multiplier options less than the default. If that’s the case, and you can get your hands on a lower-multiplier Duron/Tbird, put that in the machine, set the multiplier as low you can, and reboot/flash BIOS while it is in the machine (at 100Mhz, of course).

  • Play with the voltages a bit There are more than a few chips out there, especially the higher speed ones, that really don’t like higher voltages. You may find that a processor won’t work at 1.85V, or even at default, but will work at lower voltages. In some extreme cases, it can be as low as 1.6V.
  • Hardwire the L7 bridges to 1.85V Some people have reported their multiplier problem goes away when they do this. This is simple enough to do, and easily reversible, just connect all the L7 bridges together with a pencil. (I realize this is contradictory to the last suggestion, what can I say? Sometimes one way works, sometimes the other.)
  • Hardwire the L3, L4 and L6 bridges to the desired multiplier See HERE. This should be the last step you should take, because it is potentially the most dangerous and could destroy your CPU. Not only do you have to connect bridges, you have to break them. This is delicate mechanical work. If you don’t have the skill and patience and experience to do this, either don’t do it, or have someone who does do it. No false pride here; you won’t be a macho guy if you botch it, you’ll be a fool with a dead processor.

    People have reported success using either an Exacto knife or a minidremel. Before you do this, try to find some accounts from others who have actually done this. When you actually do it, gently cut or wear away the bridge. These copper bridges are hard, but very thin, don’t leave a gash into the processor, just enough to break it, not obliterate all traces of it. If you do it, it’s solely your responsibility.

    Email Ed


  • Be the first to comment

    Leave a Reply