The Second Level Cache Divider
On the web is a lot hoopla going on about Athlon's L2-cache divider. The Athlons that are currently available are all using a divider of 2, which means that Athlon's L2-cache is running at half the core-clock. Upcoming models that run at 750+ MHz will use dividers of 2.5 and 3 until AMD has integrated Athlon's L2-cache on the die. There's several websites that tell you to change the SMD-resistors on Athlon's PCB to change the L2-cache divider. Tom's Hardware knows that this is not necessary. The L2-cache divider can actually be changed via a software setting. It's usually the BIOS of the Athlon-motherboards that sets the divider and the BIOS is designed that it automatically changes the divider for higher clock-frequencies of Athlon. Thus you normally don't need to mess around with those settings when you plug an Athlon overclocked to 800 MHz into your board, the divider will automatically be changed by the BIOS.
Unfortunately we don't know the software settings for the L2-cache divider yet, but we are working on it. Until then, I'd still like to advice against messing around on the PCB unless you really feel that you have to.
Epilogue
We are currently working on a PCB-layout that we will publish as soon as possible and some other workaround that might not even require a complete OC-card. This will also be published as soon as we could make sure that it works.
Please excuse that I wrote this article in a way that clearly addresses people who know how to work with electronic circuits. I am aware of the fact that most of you won't be able to do this operation. However, I only want to get the word out on how to make a proper OC-card to every company that feels able to produce it. This way there will soon be a lot of overclocking-cards available for each of you and you don't need to do any soldering operation by yourself anymore.
Addition [Updated]
The 56 Ohm FID Pull-Ups
AMD's original design has one flaw, which made me re-check my circuitry-diagram several times before I released it on Saturday. The pull-ups for the FID-selectors are with only 56 Ohm against a voltage of 5V very low, resulting in a hefty power dissipation of 440 mW for each of the four resistors when the FID-dipswitch is in 'ON'-position. This results in an unnecessary heat-production of the device and most likely in a short life span of the resistors. The pull-ups have to work against pull-down resistors of 1000 Ohm on the Athlon-PCB, so that 220 Ohm should suffice just as well. At this resistance each SMD-resistor would only have to 'burn' 113 mW, which should not result in any excess heat production.
Creating a Source for 3.3 V Out of the 5 V supply
Brian Andersen sent me a good idea to 'circumvent' the voltage-regulator to produce the 3.3 V out of the 5 V power supply. He suggests to use three 1N4001 diodes in series. The voltage drop across each diode is about 0.6 V, so that you would reach 3.2 V, which would be close enough to the needed 3.3 V. Thanks Brian!
