After getting the bug to overclock my G4 in the mac mini thread, I decided to take the plunge, and see if I could get my DA G4 overclocked.
Full credit must go to Michiro Isobe for documenting the Digital Audio's multiplier and vid settings Here; and to "Thomas" at This site for documenting use of dip switches for overclocking.
My DA G4 started out as a 466, and has since been upgraded with a 733MHz processor (also from a Digital Audio). Let's see how much higher we can go...
I etched a pcb using a single-sided board, and soldered in a dip switch scavenged from an old ISA modem card.
All voltage and multiplier adjustment resistors were removed. The original intent was to use the dip switches for voltage adjustment, but that didn't work out for me. I didn't want to fuss with it, so I painted automative defog repair paint over the voltage resistor locations. To hedge my overclocking chances, I set the voltage to 1.95V over the original 1.90V.
Four 820 Ohm surface mount resistors were placed on the PCB, and four wires went from there to the board. The multiplier uses 1K resistors...this is close enough.
I used a hard drive cable for wiring.
A fifth wire goes from the common side of the switches to the CPU board (see pic above. The side opposite the remaining 4 resistors is common). All wires are stuck to a line of hot glue along the edge of the CPU board, to prevent the small solder points from pulling off the board.
Tip: cheap wirestippers normally can't deal with such a thin gauge wire. Place several pieces of tape over the plier part of the stripper to bulk it up, and the pincer won't close as firmly on the wire, and will strip very thin wire nicely.
I'll see if I can get some better pics, including the PCB.
Preliminary testing revealed a problem with the voltage control, which is when I removed it from the dip switches and painted it in instead.
After that, 800 MHz was an easy jump. 933MHz made it to the desktop, then kernel panicked. I settled for 867MHz.
Several runs of Xbench and an hour of heavy use, and I seem to have a stable system clocked up to the base speed for Leopard.
Having a well-researched plan helps. I didn't, as usual. Improvising often leads to innovative solutions, but can just as easily lead to sloppy work
Note the solder bridge at the top of the dip switch, placed once I learned I only needed one wire since there is a common side for the resistor banks.
There's a bit of copper still under the dipswitch...that was an experiment with a fine-point marker, to see whether it could be used for drawing traces. Apparently it can't.
I do have some concerns about cooling, now that I've upped both the clock speed and voltage, but it's likely the G4 will be getting new fans soon, since it's a noisy beast.
Nice work, nice documentation. I'll have to give that a go myself.