PPC750L (Copper G3) Stepping 2 voltage tweak

Through a process of elimination, I was able to determine why so many of the copper stepping 2 OWC (Apple) G3 ZIFs have stability problems. I relieved a 466 MHz ZIF of said type from a beige owner who had no luck getting it to run stable for any extended period of time. The theory was that his beige didn't like the ZIF and that it would (should) work fine in my B. I popped it in and booted up at 500...but then it crashed...hard. So I then tried 450. Same result...hard crash. I tried swapping RAM and checked my heatsink and clip to make sure that I was getting sufficient cooling. These changes (moving the RAM) made no difference in the stability of the processor.

So, I got out the old soldering iron, removed all the 0 kOhm resistors and used a rear window defroster repair kit to take the voltage from the default 2.05V to 2.10V (slight voltage increases often help stability because they (in most cases) give the CPU more headroom). Booted at 2.10V core...CRASH! So then I proceeded to try 2.20 volts, 2.30 volts, and 2.40 volts. As the voltage increased, the stability decreased (I was unable to get past the Mac OS load screen at 2.30 and 2.40 volts).

It was about this time that I remembered the voltage settings on a 450 MHz stepping 2 ZIF that I bought from OWC last July...that particular ZIF came with the voltage set to 2.00 volts (which I noted as strange at the time considering that my OEM stepping 1 400 MHz ZIF came at 2.05). Considering that the stability of my machine had decreased as the voltage increased, I decided to drop the voltage to 2.00 VCore. Booted up at 500 MHz...STABLE! Then I tried 550/275...again, STABLE! I loaded and saved state in VPC 3.0 3 times in a row to test for stability and at 2.00 volts, my machine did this at 550/275 with NO glitches. I am using a 80 WATT TEC (running at 27 watts) to keep the temps below ambient, but I don't think this is necessary.

So, there you have it. For flaky stepping 2 PPC 750L, dropping the VCore may solve stability problems.

To change the CPU voltage it is necessary to set the proper code on the VID registers (see RED BOX area in photo below).

In the past, this has been accomplished by moving the 0 kOhm resistors around using fine tipped soldering tools and 1337 soldering skills. The prolonged exposure to heat during this process can damage the ZIF and may cause the solder pads may lift off the PCB making it impossible to set the desired voltage. By using electrically conductive paint (rear window defogger repair paint in this case), these challenging and often harmful aspects of voltage tweaking can be avoided (for the most part).

Step 1: I removed all the 0 kOhm resistors (solder is needed to do this). There was no need to keep track of them or worry about damaging them in the process as they will not be needed later. After they have been removed, I made the remaining solder on the pads as smooth as possible (put a drop of flux or paste on each pad and tap it with a hot soldering iron).

Step 2: I painted over the entire VID block area (the paint dries rather quickly and the area is very small so it is virtually impossible to paint one bridge at a time). Once the paint has dried, I scraped the excess off using an Xacto knife. The final result has 5 bars of remaining paint (one bar connecting each pair of solder pads). I was careful not to let the paint connect horizontally between the pairs of pads (see blue arrow).

Step 3: I set the VID code by breaking the necessary bridges (in the case of my ZIF, the bridge on VID# 4). The measured resistance of the bridges I connected is less than 1 Ohm, so excessive resistance (causing a bridge to be seen as open) should not be a problem. I was concerned that the defogger would have high electrical resistance, but this is not the case.
-David Giessel (aka Tycho)
http://homepage.mac.com/dgiessel "

Related Articles:

• 2nd Report on Voltage Mods (Illustrated)
• Original Article on Voltage Mods