I (Joe) bought a Duron 800, did the pencil work, and tried to overclock it on an Abit KT7. The machine would only run at 700 or 750 Mhz.
Put the chip in the Soyo, ran fine.
I then used conductive paint rather than the pencil on the Duron, tried it on the KT7. Now it works.
Meanwhile, back in New York, Ed gets told by some people that a thicker application of pencil lead fixes the problem; gets told by others that it doesn’t.
MSI takes a look at our first MSI board, and has no problem with the multiplier using unlocked engineering chips. They also report that AMD told them that some CPU’s can not be modified to allow
multiplier adjustment, even if you connect the traces on the CPU.
What Is Going On Here?
Is there an explanation that ties all these loose strings together? We think so.
Let’s go over what we know:
What Happens In An Unlocked Chip?
The L1 bridges on a Duron/TBird are the link between the CPU and the multiplier-changing circuitry of the motherboard. If the link is completely cut or is otherwise inoperative, the multiplier-changing circuitry cannot work. If the link is partially cut or partially inoperative, the multiplier-changing circuitry is limited to certain multipliers.
What We Think Is Happening
We see three possible reasons (and combinations of those) why a Duron/TBird will not properly shift multiplier (which is two more than we thought an hour ago :)).
- The CPU does not generate enough of a signal to span the connected bridges.
We think AMD’s explanation holds a grain of technical accuracy around a ton of FUD. We certainly believe AMD is aware of this phenomenon; we greatly doubt they deliberately did anything to cause it.
We suppose they might have made a couple of flawed CPUs that might not generate enough current to span a perfect connection. We can believe a few more might generate slightly less current than the norm. We certainly could believe AMD doesn’t exactly rigorous check and reject chips that might have this fault. However, we don’t believe this to be an irremediable flaw of some chips.
What’s much more believable as major causes are the next two items:
- The homemade connections are not good enough to carry whatever signal the CPU is generating across a homemade bridge.
This can simply mean a lousy pencil job.
It could mean the pencil used didn’t have enough graphite to adequately connect the bridges.
It could mean the signal is weak enough that graphite just can’t connect the bridges adequately for that motherboard (see below). After all, graphite isn’t as good as a conductive pen or metal in carrying a circuit. For most people and CPUs, it’s proven to be good enough, but it may well not be enough in some cases, especially if . . .
- Different motherboards have different sensitivities to the current flowing through the bridge
Based on our experience, our Soyo obviously can accept a weaker connection than our Abit. Once we improved the connection on the Duron by using conductive paint, the problem vanished on the Abit. The MSI people using CPUs with a metal connection between the bridges obviously had a better connection that we got from a pencil.
Which one is my problem?
It doesn’t matter, because the solution for all of them is the same: improve the connection.
We will continue to experiment with this, and will update when and if we have further developments.