Abit BX133-Raid

[trodas]

Oldie board and fun with it

Whem my stepbro get from friend Ondra some time ago this board for testing, I never thought that I will ever use it for anything but testing if this or that Pentium II CPU will work or not. It has, however, Socket 370 and therefore support for Pentium 3 up to 800Mhz (whoooa! ) but also only Intel i440BX chipset, so one is limited to SDRAMs and also only 256MB per bank. From my double-sided 512MB one it "seed" only 128MB... Total size of ram is 768MB. This is better that way more modern Abit ST6-R I have, witch has limit 512MB, thanks to the stupid lame i815EP chipset. There is also promised UDMA ATA 100, but with old 40G whinning Seagate drive in standard IDE connectors is not even need for 80 wires UltraATA cable, because only the RAID channels got ATA100, normal IDE channels are only ATA33
http://www.abit.com.tw/page/en/motherboard...TYPE=Socket+370
In latest bios 12/03/2001-i440BX-W977-6A69KA1CC-72 they promise support for up to 1,1Ghz Pentium 3/Celeron CPU's, but also strongly warn that Tualatin cores are not supported. Hmmmm.

We got for first experiments two pieces of Celeron II processors, 400 and 433Mhz ones. That translate to 66 x 6 and 66 x 6,5 settings. This old suxxkas work at high 2,0Vcore Mobo support Vcore settings from 1,3V to 2,3V and chipset from 3,2 to 3,9V. Default is 3,5V for chipset. But no settings helped the 433Mhz Celeron to reach stable more that FSB 75Mhz (487Mhz) but the 400Mhz Celeron showed up way more overclocking potencial

W/o any problems it go way up to 100 x 6 (600Mhz) and just after a while it started showing errors in Memtest... Unfortunately there is no settings that could let us get rid of these errors, so we lover the clock to 85Mhz and let it be as good backup machine. Unfortunately after just two days of work there begin problems with powering up, mouse started jerking and IDE devices did not want get detected at startup, or the detection took like 2 minutes (or it worked immediately after pressing enter)
This IDE detection seems to be the best stability test - if it took long, then it is necessary to raise the Vcore or lower the FSB... Jerking mouse however continued even on default settings, so it was probably thanks to IRQ sharing or something like that...



Solution was aparent. The caps used on the board are definitively "choosen" types of caps. Probably there was some repair attempts before, on the board, as in the line of nine output Vcore caps are four types (!) of caps:
Jackcon 1000uF 10V (black-gold)
Jackcon 1500uF 6.3V (green-gold)
Canicon 1000uF 6.3V (black-gold)
Jackcon 1000uF 6.3V (green-gold)
Because all these caps, despite the writings "Low ESR" on them, did not fit even into the basic requirments for capacitor usable for computer it become clear that recap is necessary. I did not want much to do it at first, but then I decided that one good reliable backup computer can't hurt and get on it



For recap is necessary:

Abit BX133-Raid mobo
--------------------
3x 1500u 6.3V d8 (GC)
20x 1000uF 6.3V d8 (GC)
7x 100uF 16V d6.3 (RS/GK)
1x 10uF 25V d5 (RS/GK)

I used Samxon caps for their long-time proven undisputed tested quality



As you can see, I get hard with the poor mobo I laped not only the chipset (I would like to see once at lest the FSB 133Mhz, when in bios one can set FSB up to 200Mhz! ) but also the core of Celeron II, because it is nicely cooper one and it is only coated with nickel, so intel could add there his "logo", witch is of course very bad for heat transfer. So I laped it to the cooper. On the picture you can see our wall and the bottom of our outflow for watter from roof



After exchange of caps the board become act much more reliable. Unfortunately, 100Mhz FSB stable was not possible....

117 x 6 - 702Mhz = monitor show signal, so GFX card is working, but no post
110 x 6 - 660Mhz = post screen does fine, but booting into Memtest from CD is not possible
105 x 6 - 630Mhz = is possible to boot, however Memtest showing lot's of errors that are not in rams...
100 x 6 - 600Mhz = is possible to get in windows, but not stable... still errors in Memtest
95 x 6 - 570Mhz = almost stable
88 x 6 - 540Mhz = with maximum 2,3Vcore it is 100% stable (chipset is fine with lowest 3,2V...)



So Samxons caps worked well once again, even the limits of Celeron 400, created with 0,25um technology, is simply unsurpassable. Joe sayied that over 600Mhz are these black PII very hard to get. And I belive this. With better cooling and higher voltage into the CPU it will be sure possible to reach 600Mhz, but what the hell now.
Black Celeron/Pentium II are simply made with 0,25um technology. Green ones are made with 0,18um technology and these can overclock over 1Ghz, so when I get some into ma hands... And what about Tualating 1400Mhz? (100 x 14) I got one in my Abit ST6-R and it goes with 1,600Vcore to 120 x 14 = 1680Mhz! The Abit ST6-R did not make stable higher FSB, but this mobo can, as the FSB settings go way up to 200Mhz + it can instease voltage to che chipset, however I don't know any SDRAM's that can make that



Powering of the AGP graphic is reasonably well made. Jut shame that the Vcore regulator is not at least two phase one...



This board has an ISA slot too, but I think I could even desolder it and no-one will probably miss it...



For cooling I used lapped Titan heatsink with quiet (but also w/o RPM sensor...) Arctic Cooling fan. Heatsink was meant for Socket A. It is very nicely quiet, thatnk to the fan, but the bottom was terribly odd. On glass piece I was lapping it with grit 180 for over 30 minutes (!) before it's bottom was even flat! On Socket A, where the core is small it probably did not matter much, but when we tried to use it before on Celeron II, witch die is pretty big, it did not cooled a thing and the overclock has to go down significantly...
Originally it has also temperature control (there is a hole in the side of the heatsink), but it get broken and as consequence Ondra's CPU died on him, when cooled with this Titan... (could be also because of hos odd the surface was...)

So, lap, lap and once again - lap heatsink to be flat



Futhermore I discovered a great page of guy goes in nick Rayer, who was played much with this board and moded the bios for it - Rayer moded bios Abit BX133-R So now it does support even Tualatin Pentium 3 CPU's (!) and there is many interesting infos (in czech) about it:
http://www.volny.cz/rayer/hardware/abitb133.htm
For example the info about shared IRQ for PCI slots - PCI slots from top 3 and 5 are shared So better did not use these PCI slots 3 and 5! If that help me to fix my mouse jerking problems, it will be great

Mouse jerking was finally solved by reflashing of bios again with these parameters:
flash bios.bin /cc/cd/cp
These settings clear completely the flash, hence clearing well the settings and as consequence, the board suffer of no more troubles at all! Ufff, we got a pretty hard time to figure that one Using the Abit recommended flashing way ended in nightmare - never do it!

 

[trodas]

And some more old hardcore fun

...so, since nothing can be found on the net, can it be that we are the first who tried Pentium II IHS remove?

Here is classic Pentium II CPU, it is a Celeron 266 (66 x 4). It is a dead one, so we can't broke it, lol. IHS seems to be soldered to the rest of the CPU. What now?

I got a idea to simply unsolder it. Two 75W transformer soldering irons heated up the IHS enought, so I manged to get a knife under one of the corner - and that was it!

IHS from Pentium II looks this way from bottom:

As you can see, the core is higher that the rest of the top of the CPU only by the thickness of the thermal tape Intel used. As I heard, the Pentium II was manufactured by first soldering the IHS to the CPU and then inserting the core from bottom and glue it in place. After first attempt to clean out the tim I realized that the whole place bellow the IHS is from cooper!

...and tin is now everywhere...

After realy intensive cleaning, when the Intel remainings of thermal tape (?) did not react well on technical spirit, nor on technical petrol - yet it reacted at least a little on acetone:

Detail of P II core:

Conclusion.
This is almost useless. Yes, the IHS can be removed from Pentium II CPUs. Even the core can be, with hard work, cleaned. However the black border is higher that the CPU die and even the die is only by like thicknes of two sheets of paper higher that the rest of the CPU. Theoreticaly is possible to lap the black border away and then a little bit the core too, but it seems to be necessary the cooler will have bumped the bottom part, where it will touch the die. Theoreticaly it can help the overclocking a lot, because I have no reason to believe the thermal conductivity of Intel solution was nowhere near good and the poor contact between just the IHS and the core making me being sure about possible overclock improvment.
However now it is necessary to obtain a pure cooper heatsink and mill away most of it's backgroud at about 0.5 mm thickness to prevent shortcut between bottom of the heatsink and most of the pins of the CPU

 

[trodas]

A great sucess todays!

Todays I added 12 (!) 4,7uF SMD caps to Abit BX133-R board Socket 370 and the maximum FSB of Celeron II 400 (66 x 6 by original) went from 88Mhz to 94Mhz! Great sucess! Yea, only 564Mhz, but hey - I get nice improve!

Next move - perhaps rewinding of the Vcore coil with thicker wire?
Or getting pure cooper heatsink?
Or exposing the die?

...or all there above?

 

[trodas]

Anyone know for SURE (I depend on the CPU, so I did not want it to be killed!) that the Abit BX133-R can handle VIA C3 Nehemiah CPU? Vcore is 1.400V...

In Abit BX133-R specifications are:
CPU: Supports Intel Pentium III /Celeron Socket based processor

There IS a support for VIA C3 tough, in the bios download section is there mentioed support for "VIA Cyrix III CPU".

Witch VIA Cyrix III CPU's this include? Does it include VIA C3 Nehemian 1200Mhz I have? With Vcore 1.400V? Anyone know?

http://www.abit.com.tw/page/en/down..._ON_SCREEN=BX133-RAID&pSOCKET_TYPE=Socket 370

 

[Wile E]

I don't know about your cpu problem, trodas, but I just wanted to say that you're recapping and modding articles are absolutely great, Keep up the good work.

 

[trodas]

Thanks a lot, Wile E Let's hope I can get hold of many interesting hardware, so I can show many interesting projects as well...


About Abit BX133-Raid and VIA C3 Nehemiah 1200Mhz CPU - now I know, yes, it does work. There is a little error on display:


(can I claim world record, 4MHz only? Just let me made CPU-Z screenshot Current record is 15Mhz...)

but the CPU is working well:



No ram that can handle the 133 x 9 as the VIA C3 Nehemiah 1200Mhz should have, but what the helll... It works!

 

[Wile E]

Sounds like you need to try to find some better ram.

 

[Carcenomy]

I'm impressed just by the work being done on the now antique 440BX chipset! I'm suprised Abit built any S370 boards with 440BX chipsets to be honest...

Trodas - stick some beefy single-sided DIMMs in it - the 440BXs thrive on them in stock tune so with your uprated caps and tweaks, the thing should hum along pretty good

 

[Carcenomy]

Actually... Trodas, where the hell did the parallel and serial ports go?

 

[trodas]

LOL - yeah, the ram kind of suxx Neverless I could utilize it for the contest to get the lowest 3DMark score... Then we get back to the overlock, lol

I removing parallel and serial ports on almost all my machines I recap. For better cooling - hot air from CPU escape right away off the case, so, no restrictions And someone actually use these ports, I mean, ever...?

 

[Wile E]

LOL - yeah, the ram kind of suxx Neverless I could utilize it for the contest to get the lowest 3DMark score... Then we get back to the overlock, lol

I removing parallel and serial ports on almost all my machines I recap. For better cooling - hot air from CPU escape right away off the case, so, no restrictions And someone actually use these ports, I mean, ever...?

I know I never used them, not even once. lol That's what we have USB and Firewire for, nowadays.

 

[trodas]

Exactly that is why they has to go
I remove everything I feel as unnecessary

 

[trodas]

And before I do some more hardcore modding - there are still one probem and one question.

Problem.
The bios (altrough it offer generously up to to 2.3Vcore for CeleronII CPU's) limit harshly maximum Vcore selectable when Tualatin CPU is detected. From default 1.45 to just 1.55 - witch is useless at best. Setup is saving the value at ofset 41h and it also does change value on 3Ah, where it is used just to show the correct Vcore in bios.
Hardcore way is using RayeR utility to directly write to CMOS wanted value and it become active on next reboot.
( http://www.volny.cz/rayer/programm/abitvc.exe )
However a clean way - eg. kick out the code that does limit the Vcore options based on detected CPU is wanted.
Anyone know how/can do it?

Question.
Altough 2.3V might sound enough for most, for hardcore overclocking of Celeron II this is simply not enough. Is this mobo voltage regulator capable of delivering higher voltage, say at least 2.5 ot 2.6V...?