Mod to force t42 (and possibly earlier versions) to 133fsb.

[cirthix]

As I promised to post a while back, here's the mod to force it to 133fsb:

http://img86.imageshack.us/img86/3742/p4190151oe0.jpg (190KB inline image changed to a link)
http://img221.imageshack.us/my.php?imag ... 155yb9.jpg
http://img218.imageshack.us/my.php?imag ... 157jt3.jpg
I don't remember what I had to change (one or two resistors moved), as it's been a while. If someone posts original pics, I can tell you .

Beware that the memory is also overclocked, with pc2700 running at 221mhz and pc2100 running at 171mhz. You can easily flash the spd of the ram to make it think it's slower than it is, making your pc2700 act like pc2100 and then run at 171mhz. You'd likely need pc3200 to run at the higher speed, only one of my 5 test pc2700 sticks was able to do 221 stably.

This is verified 100% stable on my t42 with a 1.4@1.86, but a known-good 1.7 (that could easily do 2.26 in other boards at stock vcore) was not able to do 2.26 stably. I tried a 2.0 533fsb chip which worked fine. This will probably work ok for those of you with 1.4/1.5/1.6ghz chips, but in my case, the mobo was having problems with the 1.7.

to mod the spd:
http://forums.techpowerup.com/showthread.php?t=20349

Moderator edit: Inline images need to be ~50KB AND a notice posted in the subject line.

[gunston]

hey man, you're awesome...

[xyz]

Thanks man thats awesome, i take it these are pics of your modified board ?

I can post some pics up later of a board later and we can see what the changes are hehe.

[xyz]

[xyz]

Here you the pics
Can u elaborate as to what is needed to be moved /changed

[kyrotech]

very interesting.

and yes we need more detailed info about what was changed on those pics =/

please!

[unrortit]

it would certainly benefit hundreds of owners if somehow you can remmember which resistors you altered,
would be a damm shame if all your ingeluarty accounted to
no benefits for fellow forum users,whom would really appreciate more co opertation to expoit my suspected company manipulation schemes.worth living for i say.
i hope it all comes back to you one day for everyones sake.

[cirthix]

I forgot about this thread

This appears to be the only change. I remember only having to change one, and this is the only obvious difference.

They were just 10k ohm pull up/pull down resistors on the clock generator's input pins.

http://img258.imageshack.us/my.php?imag ... 494hf8.jpg

have fun

i still have the modded board+ 1.4 LV@1.86 if anyone's interested . it has a 9600 on it that oc's well. to 450 on the core iirc.

[baisley]

cirthix,

Thanks very much for posting the fix. I have a few questions to ask. Would we only need to install a resistor on the back of the board on those points ? The one circle you have with an arrow pointing down, there is no resistor to move on my T42 board. Do we have to do anything on the opposite side of the board or is this circuit complete by just installing that resistor across the two solder pads that are identified in C204 circuit ?

Now for the real newbie question:

Best way to solder something so small on the board. Pencil tip soldering iron ? Hot air station ? (I have been itching to get a hot air station for quite a while and it is afterall, Christmas

Thanks for your tip on this ! Please let us know if you have any others or could point us to the sources for this and the others.

Happy Holidays

[Mikem95626]

Wow...impressive.
Thanks alot for posting this...I have a old T40 motherboard laying around with a 1.3 cpu in it...the resistors look the same as your pic. I think I will give this ago and see how it goes. If I am not mistaken, the T40 runs the memory at PC2100 speeds anyway, so any PC2700 sticks should handle the 33% bump with ease.

[cirthix]

Ok, you got me to whip out my board and have a closer look at things.

OUR BOARDS ARE LAID OUT DIFFERENTLY! Don't worry though, the mod is still doable, I'm very sure, but you can't do exactly what I did.

I just checked out the datasheet again, and fs1 needs to be pulled from low to high. You can do this a few ways. I did it by moving a resistor from pull-down to pull-up (there are two places to mount the resistor, one is connected from fs1 to vcc and the other is connected from fs1 to vss). Another is to short the FS1 pin to the refrence voltage input.

The surest, layout-independent way to force 133fsb is to do the latter:

http://img180.imageshack.us/my.php?imag ... ba9sp9.jpg




In case anyone's wondering, the datasheet from the clock generator may be found here:

http://www.alldatasheet.com/datasheet-p ... FLF-T.html

[xyz]

So just to confirm if you short vs1 with vref that will force a bump to 133.. and u don't need to mess with any of the resistors etc..

[cirthix]

Yes, FS1 to VDDREF.

[Mikem95626]

Would it be easier to just short fs0 and fs1 together since the pins are next to each other or does that not work?


Once again, thank you for sharing this information with all of us. I have been wanting to o/c my thinkpads for years...they are always just sitting at a desk running on a/c anyway.

[cirthix]

Would it be easier to just short fs0 and fs1 together since the pins are next to each other or does that not work?


Once again, thank you for sharing this information with all of us. I have been wanting to o/c my thinkpads for years...they are always just sitting at a desk running on a/c anyway.

Depending on how yoru board is laid out, this may cause problems. On mine, One was pulled up and one was pulled down, so shorting them would cause both pins to be at vref/2, which would probably cause the system to not post, but would work with undoing it.. If IBM got cheap and tied them directly to vref or gnd, this would cause serious problems (ie, short between power and ground=burned trace somewhere or worse).

In short, no, this won't work.

[ClayMan]

sounds like fun.. anyone tried this recently? thinking about doing this on my t41 for fun..

[Thomcat]

Can this be done on a T41?

[Thomcat]

Wouldn't this be easier?
http://www.notebookforums.com/showthread.php?t=80879

[FTC]

Wouldn't this be easier?
http://www.notebookforums.com/showthread.php?t=80879

No, this does not work on 855 chipset based notebooks such as the T40-T42 (this would work in T43s though...)

[jamiphar]

.......The one circle you have with an arrow pointing down, there is no resistor to move on my T42 board. Do we have to do anything on the opposite side of the board or is this circuit complete by just installing that resistor across the two solder pads that are identified in C204 circuit?.........

I'm noticing the same things on the boards I have. There are a total of six resistors in the area that the photo shows. I have a couple of dead boards that I can take resistors from, does anyone know which ones would work for this?

[aaa]

Hey all. This may or may not help (it's for an R50):
http://forum.thinkpads.com/viewtopic.php?p=409119

[screwdriver]

are there any news?
@cirthix
how safe is the method with connecting fs1 to vref?

[xob7]

are there any news?
@cirthix
how safe is the method with connecting fs1 to vref?

Hi. I'm new here

Here's my mod of i855PM:
1) Make sure your RAM is able to work at 176Mhz ( 133FSB and ratio 3:4 ). I've just flash SPD on my RAM to DDR266.

2) Connect pin54 to pin55 on ICS950810. Tested on my T42.
Try some wire and nail polish
http://galeria.oneit.info/album/Seb/fsb133mod0.jpg
http://galeria.oneit.info/album/Seb/fsb133mod1.jpg

After that you can use Dothan 7x0 with full speed.

RAM clock after mod i855PM to 133MHz

http://galeria.oneit.info/album/Seb/ram1.jpg
( RAM SPD locked to DDR266 )
http://galeria.oneit.info/album/Seb/855at133.jpg
( IBM T42 & R9600, i855PM @ 133MHz, Dothan M750, 2x512DDR350 )

[Binh]

OK. In addition to what xob7 said:

A. Pinmod PLL:
- The simplest method to connect pin 54 to pin 55 is using a soft pencil: just draw a small line between these pin. Hope that you still remember the "pencil method" for overclocking AMD K7 CPU. Note that you don't have to de-solder the 55 pin as many sources on internet said.

- I am successful in using pencil method for T42 (PLL chip is CY28346ZI-2, same as xob7) and T41 (PLL chip is ICS950810). AFAIK, there only these two types of PLL in T4x.

B. Many are interested, how to change SPD of RAM module to force it to work at 266Mhz (resulting after overclock 354Mhz). Here is the intruction:
- Download SPDTool software
- From the File menu, go to Read -> Module 0
- Choose File -> Save to backup original SPD data.
- In the grid on the bottom seach for SDRAM cycle time at maximum supported CAS latency change default value 6ns (167Mhz) to 7.5ns (133Mhz). Assume that you have a DDR333 module installed in the internal RAM slot (under keyboard).
- Go to Edit -> Fix Checksum. DON'T FORGET this step. If you forget it, the computer will not recognises the altered RAM modul and it won't boot up after restarting.
- Go to File -> Write -> Module 0 to save the altered SPD.

Note:
- If you have installed two memory modules, you need to change SPD only for the first module, i.e. the module installed inside the laptop under the keyboard.
- If you still forget to Fix Checksum and the lappy doesn't boot: remove the module and install it to another laptop (not IBM brand) which BIOS does not stop when encounter error in SPD (I use my syster's DELL X300), and reprogramm the SPD.


To xob7: You have the same PLL as my T42p. After pinmodding, the SetFSB for thinkpad X31 stop working though it did work before pinmodding. Do you have a similar problem ?

[xob7]

(...)
To xob7: You have the same PLL as my T42p. After pinmodding, the SetFSB for thinkpad X31 stop working though it did work before pinmodding. Do you have a similar problem ?

Binh I've never tried to use SetFSB so can't tell.

Memory SPD after flashing will look like this: http://galeria.oneit.info/album/Seb/ram_mod.jpg

[sjthinkpader]

...
B. Many are interested, how to change SPD of RAM module to force it to work at 266Mhz (resulting after overclock 354Mhz). Here is the intruction:
- Download SPDTool software
- From the File menu, go to Read -> Module 0
- Choose File -> Save to backup original SPD data.
- In the grid on the bottom seach for SDRAM cycle time at maximum supported CAS latency change default value 6ns (167Mhz) to 7.5ns (133Mhz). Assume that you have a DDR333 module installed in the internal RAM slot (under keyboard).
- Go to Edit -> Fix Checksum. DON'T FORGET this step. If you forget it, the computer will not recognises the altered RAM modul and it won't boot up after restarting.
- Go to File -> Write -> Module 0 to save the altered SPD.
...

This is great. Here is the SPD modification tool I made a few years ago for SDR SODIMM.

SPD programmer

Basically an EPROM writer wired to a SODIMM socket. Editing was done using Excel spreadsheet and DEC2HEX and HEX2DEC commands. Value were obtained from JEDEC specs.

[horacelau]

Two questions:

1. Have anyone tried this with a Banias based T41? Will this make the Banias run 533MHz FSB?

2. What speed does the memory run at with DDR333 and 533MHz? I'm confused about flashing the SPD to DDR266 and why that's necessary. What's the memory speed with SPD at 333 vs 266?

[Binh]

to horacelau:
1. I have success to overclock my T41 1.6GHz Banias (type 2373-9U5 if you are interested).

2. You have to flash SPD to force memory to run at 266Mhz in order to avoid it to run at very high frequency when overclocked. Here is explaination:

A. Assume you have a DDR333 module. In normal mode (not overclocked), the memory run at 166Mhz speed (resulting DDR 333 since DDR doubles speed) and CAS value 2.5, while the FSB is 100Mhz (or 400Mhz in QuadPump speed terminology). Thus we have the FSB:Memory divider 3:5 (you can check it with CPU-Z software, page "Memory").

B. If you overclocked FSB to 133Mhz (or 533 quadpump speed), with the memory divider 3:5 you have RAM speed of 133*5/3=221Mhz (or DDR 442) at the same CAS value 2.5. Only a few DDR modules can do DDR442@CAS2.5.

C. If you reprogrammed SPD to DDR 266, then the computer will set memory divider to 3:4 and thus after overclocking your memory module works at 133*4/3 = 177.6Mhz (or DDR 355).

Some notes:
- The i855 chipset support only two memory dividers: 3:4 and 3:5.
- It support only two values of CAS latency: 2.0 and 2.5. Consequently, the use of DDR400 modules has no advantages since at DDR400 the CAS latency of these modules is 3.0

[horacelau]

Binh,

That all makes sense. Thanks for your response. So your Banias 1.6 was able to run at 2.1?

[horacelau]

Just tried on my T41 with a Banias 1.4GHz -> 1.83GHz using the pencil trick. It does boot up, 533MHz FSB, memory at 177MHz, but the system will freeze after trying to use it. I tried running Windows memtest and that seems ok, so I think it's CPU related. I would think 1.83GHz shouldn't be a problem. Anyone have any ideas?