i855PM 533FSB laptop mod

Long story short, the i855PM chipset will run a FSB of 533mhz. I did some research on my laptops PLL (CY28346) and discovered that it supports 533 (133x4) and supports the required dividers on the PCI and AGP buses. Shorting pins 55 and 54 will set both S0 and S1 HIGH. Unfortunately the ram divider doesn't change with this mod so the memory buss speed is 179Mhz after. The problem is compaq has saw fit to limit rams speeds to 266Mhz in my EVO N620c. So the chipset still sees it as DDR266 ram, so sets the DDR266 timings. My ram was unstable at DDR358 using DDR266 timings so I used a utility called SPDtool to flash the rams SPD info to run DDR333 timings at the apparent DDR266.

Stock speeds: 1.7Ghz @ 1.35V, DDR 333 @ DDR266.
Resulting speeds: 2.28Ghz @ 1.3V, DDR333 @ DDR 358.

 

So far so good. 24 hour prime and S&M stable. Max temp reached during testing was 53c (127F). So far its running faster (by nearly 600Mhz), cooler (by around 10c) and running on less voltage (1.48v compared to 1.3v) then the stock banias core. The north-bridge does run slightly hotter but the change in heat output is negligible.

Known issues so far:
Incorrect frequency and L2 cache size in the bios info page and some CPU info software. (probably due to running an unsupported dothan core and/or fsb speed.)

Resuming from deep sleep states resets the FSB to stock 400 (100x4) and sometimes hard locks the system. Using RMclocks "run HLT commend" option solves this problem.

There is approximately 1 in 10 chance of a FSB of 66Mhz loading on boot. On further investigation S0 and S1 voltage drops on boot. sometimes enough to set LOW on both. Shorting S0 and S1 to VDD (3.3V) could fix this problem but is untested.

 

This is quite interesting, thanks for detailing your work. I had very little clue that the i855PM chipset supported anything other than plain-jane 100mhz FSB speeds and the prospect of breathing a little more life into this old workhorse is exciting.

I've got a Toshiba Portege M205 tablet with a Dothan 745 1.8ghz that has been running at 112*18=2016mhz via ClockGen (using ICS 952618 as the PLL). Changing the FSB in this way has the side-effect of bumping the AGP and PCI bus speeds, meaning that at greater than 112mhz FSB and 37.59 PCI the sound cuts out. Surprisingly, my Hyundai and Micron PC2700/DDR333 sticks have been fine at ~187mhz.

Though that ClockGen PLL setting works, my motherboard really has the ICS 950812, the other commonly used i855PM PLL (besides the Cypress one you have). There is a datasheet available here (or attached) and more information on that part and its variations here.

The ICS950812 appears nearly identical to the CY28346, with pins 54 and 55 representing FS0 and FS1. Theoretically, shorting these two together should produce a 133mhz FSB, but I haven't been successful at this yet. Some people suggest shorting them with a pencil mark, others suggest connecting FS1 to VDDREF, and then others suggest adding a resistor. I'll keep trying, perhaps lifting/cutting the foot of FS1/pin 55 and connecting it to FS0/pin 54 or perhaps VDDREF...

There are a few guides/threads out there on this subject, such as this one about adding a 1k resistor to a Dell D600 (same CY28346), and this one on shorting and resistor additions in various Thinkpads.

 

I tried a pencil mark and resister and got no response. The only way I could get a FSB of 133 was directly shorting FS0 and FS1. I first attempted it by simply shorting the pins with a small screwdriver. In my case (and probably the norm) once the FSB was set on boot I could take it off and stability test and the FSB will not change back. But be careful, I have heard that some mobos have "LOW" FS pins shorted to ground. If that's the case you will be shorting "HIGH" voltage to ground. Check FS0 and FS1 for dead shorts to ground first.

 

what type of wire did you use for the short? I know if the diameter is too thin, it will get 'heat creep' and move enough so that on say, the next boot, it will not give proper contact, thus giving the lower readings.

 

The S0 and S1 pins on the PLL are simply soldered together. There the second and third pins from the top left in the pic of the PLL.

 

nice.

next time, use duct-tape

 

Hello, i have a 2003 laptop with a compal cl50 motherboard with ICS950810cg, its very similar to that in the photo. I have bought a now cheap dothan CPU to use on it but the current 400mhz fsb limits the cpu speed to 1,6 instead of the original 2,13 if it had 533mhz.
If i connect the pins 54 55 named FS0 & FS1 with a copper wire will it work? im kinda afraid it might burn or something....

Is there no way of simply making it run at 533mhz through software?


tks