"
Hi Mike,
I finally pulled together all the information I needed to modify the core
voltage of my cpu. IBM was exceedingly helpful in providing me the
relevant information (which isn't up on their website). I am quite
grateful for their assistance and the level of information they provided.
On my processor module (and this may be different from even most other IBM
ZIFs), the VID[4-0] resistor settings start out at 00000 for 2.05V. A 0ohm
resistor populates each of positions R3 through R7. R3 corresponds to
VID[0] on up to R7 which is VID[4]. The relevant resistors are boxed in
purple in the top view of my IBM ZIF/PCM (Processor Card Module) attached
to this email.
I bought one of the first blue PowerMac G3s running at 400MHz. I ordered
it the day they were announced and received it a few days later. When
afterwards, the information on the speed settings became available on your
site ( http://www.xlr8yourmac.com/G3-ZONE/yosemite/OC.html ), I
overclocked my machine to 450MHz.
This was still in February as I recall. At that time, I had no problems
with stability. As spring rolled along, the ambient temperature in the
house increased and I started seeing the occaisional random crash. They
increased in frequency as the temperature increased so I had to back off
to 400MHz again. Since then, and since seeing the information on how
someone had bumped up the core voltage of his PowerForce G3 ZIF (
http://www.xlr8yourmac.com/guest_rev1/500mhz.html ), I decided to look
into doing something similar for my ZIF so that I could overclock my
machine again.
As I had mentioned to you earlier, I couldn't find the specific resistors
responsible for the settings. Fortunately, IBM provided me with the
information (basic circuit diagrams for my processor module circuit
board). That was very gracious of them. I really am grateful to the people
at IBM who helped me get this info.
Yesterday, after desoldering resistors 3,5,6, and 7 (for an VID[4-0]
encoding of 11101), I had the core voltage at 2.2V.
I was able to increase
the clock speed to 473MHz (95.5MHz bus x 5). My machine simply would not
boot at 500MHz. I then tried sticking my PCM and heatsink in the freezer
for a few minutes. With no fans blowing on it (since that would just have
heated things up faster) I was able to run at 500MHz for about 12 seconds
before the machine crashed. (Note that you shouldn't do this a lot since
if nothing else, the condensation can potentially damage things.)
Last night, I did some more soldering. I moved the resistor in R4 over to
R5 for a core voltage of 2.4V. Since then I've been running at 500MHz
stably (so long as the cache is running at 5:2 or slower). I have a single
2" cpu fan blowing on my heat sink. The machine immediately dropped into
OpenFirmware when I tried booting at 519.7MHz (94.5x5.5). I didn't bother
with 550MHz. I'll see over time how stable this setup is. I've not been
running it at this speed/voltage for very long yet.
The reported processor temperature through Powerlogix's G3 Cache Profiler
decreases as the core voltage increases so don't take the figure in the
attached screen shot without at least a 22C/40F correcting factor.
Regardless, it's still way below IBM's recommended maximum operating
temperature of 65C for this version of the 750. The voltage is pretty
close to IBM's recommended maximum core voltage of 2.5V. I'm not willing
to push it any further.
With adequate cooling (by this I mean active cooling :), I might be able
to push it up to 600MHz or beyond (sez the optimistic guy). BTW, anybody
using peltier junction coolers should be putting them on a completely
separate power supply since a single peltier can draw more current than
the average computer power supply's rated maximum at 12V just by itself
(with no motherboard, drives, etc. attached). [Note - a combination Peltier/Heatsink/Fan combo I bought at a recent PC Show cited 2 AMPs of 12V current usage - that's 24 Watts of 12V power. Also note that as reported in the main site news in the last month or so, two beige G3/Peltier users reported RAM failures after 6 months or more of use, both showed damaged capacitors on the DIMMs. Condensation is another issue as the Peltier surface area is much larger than the contact patch of the G3 CPU (many times larger). Condensation could short out your motherboard or CPU.-Mike]
(The remainder of this is directed at the readers and not specifically
you, Mike, since I think you are well aware of the dangers.)
I would advise people against doing this kind of modification. It takes
fairly tricky soldering/desoldering of miniscule surface mount resistors.
Your only practice should definitely not be on a working (expensive) piece
of electronics like your processor module.
If you must do this, you should have the right tools including (but not
limited to) a temperature adjustable soldering iron with a sharp,
elongated tip (these surface mount resistors are _tiny_). Care should be
taken to keep from heating up the circuit board too much or you'll start
burning out traces and rendering your processor module useless. Stray bits
of solder could be deadly.
Unless you _really_ think you can handle all this or are prepared to buy a
whole new computer at worst for a relatively (to the potential cost) minor
speedup, just don't do it. You stand a chance at destoying most of the
components in your computer and loosing data by doing this. Consider
carefully the risks and the real benefits.
Note that your processor module, even if it comes from IBM, may be very
different from mine. If you must try this, do not rely on the same 5
resistors doing the same thing as mine. I've not yet found a ZIF module
with the same circuit layout on any of the machines where I work. Be sure
of which resistors are responsible for encoding the VID settings on
_your_specific_ ZIF module before seriously considering this.
I take no responsibility for any of these modifications if you damage
something. I recommend against doing this. It is risky.
Hehehehe! 500MHz Whee! =)
Ryan Tokarek
http://www.cloudmaster.com/pulsar/
"
