P4 default FSB pin mod (400->533) success story

CommandoCATS

Senior member
Jul 8, 2003
517
0
76
*DISCLAIMER* AS ALWAYS YMMV, don't blame me if stuff explodes..etc

A bit of background:
I've been trying to put my Pentium 4 1.6A o/c 2.5Ghz setup into a mATX board because I've gotten ahold of a Power Mac G4 Quicksilver case (Thanks PowerMacMan). [Pics will be available for this mod shortly too]. After a lot of research, I chose a Foxconn 865M01-G-6ELS because it seemed to be have the most (albiet quite poor overall) overclocking features for a mATX board (All the mATX boards from the big names like Asus, Gigabyte, Abit..etc have *no* overclocking features at all). Anyway, I got everything set up, and I ran into my first snag - the MB would only allow a 32mhz FSB higher overclock than your default (which is 100mhz for a 1.6A - ignoring the x4 quad pumping). Well, I thought, 16*132 gives me about 2.1Ghz, thats not *too* bad - so I set the BIOS "superspeed" to 132 mhz, hoping to be greeted by a friendly beep. WRONG, nothing, no beep, no logo... just dead. Now that pretty odd for a CPU that on my ASUS P4B266 can easily hit 2.4Ghz at stock voltage. Well, I figured this motherboard supports 100/133/200 (400/533/800) FSB, so it can't be hitting the chipset's limit, must be that the PCI and AGP frequencies are getting way too high. Wouldn't it be nice if I can mod the CPU somehow so that it thinks the default FSB is 133(533) and sets the correct divisors for AGP/PCI.

So I read some forum posts, and I read something about someone who made his 533 CPU go down to 400 by shorting the BSEL0 and BSEL1 pins together.

The signals table for the BSEL pins are the following:

MHZ BSEL0 BSEL1
400 L L
533 L H
800 H L

Now my understanding of how this works is like this: (while I'm an electrical engineer, but I could still be wrong here) Basically, the motherboard socket puts VCORE (1.5V) on both BSEL0 and BSEL1. That means that if no connection goes from the motherboard to the CPU and back out to ground, that is considered "high" (e.g. the current finds a dead end, so the voltage stays at 1.5V - think of a water hose where you have the end closed: no water (current) flows, so the pressure (voltage) stays high.) The way to make that a "low" is to provide a path to ground (through the CPU). So on the 533mhz CPU, BSEL1 normally is just a dummy pin, a "dead end" for current, while BSEL0 connects to a ground pin internally somewhere in the CPU. By shorting the two, both of them become ground pins, thus the FSB gets detected as 400mhz. Now what I want to do is the opposite, I have 2 ground pins, and I want one of them as a "dead end" pin. Well thats easy enough, just snip off the BSEL1. So I located the corresponding physical pins using Intel's datasheet for Northwood and pried the pin off (it wasn't very hard, used the retracted tip of a zebra mechanical pencil as a lever and just pried). Be careful reading that diagram, pin A1 is "pin 1" on the P4, the one thats marked by the triangle in the corner, and the diagram is referenced from the top of the CPU, so you need to mentally rotate the image. I put the CPU into the Foxconn motherboard, and VOILA! detected as 133x16=2.1Ghz! I then used the BIOS "SuperStep" and bumped the FSB up to 153 (it allows me to go all the way up to 162 now that my "default" FSB is 133). 2.5GHz, perfect!!!

Just for kicks, I volt moded my Northwood up 1.75 default VCore as documented here. I'll maybe post some photos and maybe write a real guide later when I'm less tired. I've been experimenting with this all day (good thing I have this Mac G4 case, it lets me pull the motherboard tray out while the computer is running). I want to see someone with an Intel motherboard try this and see if you can overclock using that :).

Please post your comments/experiences below :)

-A very very tired Zuofu
 

CommandoCATS

Senior member
Jul 8, 2003
517
0
76
Whoa, that could have helped me out like, 12 hours ago:p Damnit! Google failed me! As an update, anyone figure out if this works on 533->800mhz? Though you'd have to have a pretty lucky CPU to get that stable.
 

hemi

Member
Dec 8, 1999
103
0
0
Cool. Glad that worked out for ya. I haven't done it on a P4, but I did take a 1.2GHz Tualatin Celeron and pin modded it to 133FSB (1.6GHz) for use with an MSI board.....always wondered if the same would work with a P4. :)


Tom
 

Mingon

Diamond Member
Apr 2, 2000
3,012
0
0
I have used this on mobile p4's, some boards like gigabyte require it. You can also mod voltage this way
 

Zap

Elite Member
Oct 13, 1999
22,377
7
81
With a combination of pin removal plus shorting out pins, it should be possible to get any documented (and undocumented) voltage and FSB. For instance, there should be an unofficial 667MHz FSB for the P4. Don't know how a motherboard would handle it, but it is available. 533=>800 might be possible with a chip like the Prescott 2.4a, bumping it to 3.6GHz. It is a bit high since out of the two 2.4a chips I've overclocked with, neither would do more than about 3.3-3.4GHz stable, but it is conceivable that on the right board with the right RAM and a lucky chip... Probably no other chip would have any chance at doing that... hmmm, unless a mobile... wait, I have two mobile Celeron 1.6 chips... to risk or not to risk a budget CPU... Heck, I first have to find a motherboard that will run it close to properly at default speeds.

CommandoCATS, my page probably isn't listed on Google. Heck, I run that page off my work computer, LOL. I just post a link whenever someone asks for information like that. Maybe I should update it with 667/800MHz FSB support, plus voltages.
 

THUGSROOK

Elite Member
Feb 3, 2001
11,847
0
0
hey,
alot of 800fsb owners may be interested in this too.
quite a few motherboards lose PAT when OCing. tricking the board into thinking its a 533 cpu may resolve that problem.

:)