Great work! I'm a fan of this board too. I've run it at 75Mhz for a while without any problems, but I believe that 83Mhz will push the PCI bus too far out of spec for some cards.

You may already be aware that there's a modded bios available with K6+ and >32GB drive support. I can vouch for this working well.

bjt wrote:

Great work! I'm a fan of this board too. I've run it at 75Mhz for a while without any problems, but I believe that 83Mhz will push the PCI bus too far out of spec for some cards.

You may already be aware that there's a modded bios available with K6+ and >32GB drive support. I can vouch for this working well.

Great! Thank you! It's not like I'm going to build something on it anytime soon though, got too much going on 🙁

Relevant:

more inductance => better voltage stability (less ripple)
less inductance => better transient response

Might be interesting to experimenting with 😀

The board uses ICS9169CJ-272 chip. Its datasheet tells all possible FSBs.

lazibayer, the 50-75 MHz settings are there already and 68.5MHz reports 66MHz for some reason. Do the BSEL settings correspond to PCI bus? That would be awesome if this board could overclock while keeping 32MHz PCI clock.

RacoonRider wrote:

lazibayer, the 50-75 MHz settings are there already and 68.5MHz reports 66MHz for some reason. Do the BSEL settings correspond to PCI bus? That would be awesome if this board could overclock while keeping 32MHz PCI clock.

Theoretically yes. On SP97-V there is a FS3 jumper for switching between 1/2 PCI divider and async 32MHz bus freq. It uses ICS9169CM-23 chip with similar input settings:

I couldn't find such jumper on TXP4, TX97 or SP97-XV boards that use ICS9169CJ-272. I am not sure if those boards are designed for async PCI bus.

They are not designed for that at all. Mine Asus TXP4-X is mainly limited by the PCI bus clock limit. 75MHz is fine, but 83MHz cause HDD problems... Maybe some PATA/SATA or CF card conversion could make it? Dunno.

I recapped the board completely back in 2008 for my friend:

...and recently started working on it to make it run at 7.14MHz FSB w/o troubles (the troubles are, that no keyboard or mice is working at such settings...)...

Pentium 90 overclocked to 100MHz (66x1.5 - 3.4Vcore):

AMD K5-75 (50x1.5 - 3.4Vcore)

And few new pics from testband:

...and there you go, 7.14MHz FSB x 1.5 = 10.7MHz repeated again 😀

Works good, except the input devices. But maybe serial mouse could work...?

trodas wrote:

...and there you go, 7.14MHz FSB x 1.5 = 10.7MHz repeated again :) http://s2.postimg.org/w5jp7q1dx/Asus_TXP4_X_P90_at_10_7_MHz.j […]
Show full quote

...and there you go, 7.14MHz FSB x 1.5 = 10.7MHz repeated again 😀

I'd be interested in Norton SysInfo 6 scores if you could run it, especially to compare speeds against an 8088 XT, as noted in this thread: The IBM XT challenge: How slow can you go with i386+ (IA-32) CPUs?

trodas wrote:

Works good, except the input devices. But maybe serial mouse could work...?

This drop in FSB may have thrown all your other buses out-of-wack, but there shouldn't be any harm in trying a serial mouse.

If is there a option, that let me run the test w/o user input (remember, at 7.14MHz FSB = no keyboard or mice!), I gladly do it. Put it on bootable CDROM and run by by autoexec 😀 Should be fun! ATM it is running SuperPi 32M at fastest clocks I managed yet (75x1.5 = 112.5MHz) stable, but that is, of course, quite uninteresting for your purposes 😀
(you will see there: http://hwbot.org/hardware/motherboard/txp4-x/ ...when it finished the run)

Yep, the FSB drop thrown pretty much everything out-of-whack, but if graphic works on PCI bus, then a PCI USB card could work as well, witch mice there... 😀 I have to test the USB ports in the low clock case too. Might be as easy, as to use them... And I did not have ATM a serial mouse, 🤣 🙁 I know, I suxx. I try to acquire one 😀

trodas wrote:

If is there a option, that let me run the test w/o user input (remember, at 7.14MHz FSB = no keyboard or mice!), I gladly do it. Put it on bootable CDROM and run by by autoexec 😀 Should be fun! ATM it is running SuperPi 32M at fastest clocks I managed yet (75x1.5 = 112.5MHz) stable, but that is, of course, quite uninteresting for your purposes 😀
(you will see there: http://hwbot.org/hardware/motherboard/txp4-x/ ...when it finished the run)

You can automate the benchmark in Autoexec.bat.

1) Download Sysinfo from the above link if you don't have it already.
2) Use the following entry in Autoexec.bat, at the end of the file, type:

1CD "PATH to the directory where your sysinfo file is"
2SYSINFO /DEMO
3
4e.g.
5
6CD C:\SYSINFO6
7SYSINFO /DEMO

... The demo should run automatically.

trodas wrote:

Yep, the FSB drop thrown pretty much everything out-of-whack, but if graphic works on PCI bus, then a PCI USB card could work as well, witch mice there... 😀 I have to test the USB ports in the low clock case too. Might be as easy, as to use them... And I did not have ATM a serial mouse, 🤣 🙁 I know, I suxx. I try to acquire one 😀

I'm not sure if USB mice and keyboards are reliable in DOS, even if the buses are running to spec - never tried it myself.

USB mice can only work in DOS if the BIOS supports PS/2 emulation, and it doesn't do so for keyboards only. I've got a PCChips M530 that doesn't have it, my MVP3 SS7 mobo has it for keyboard only, but all my stuff that is newer than that has it.

trodas wrote:

...and recently started working on it to make it run at 7.14MHz FSB w/o troubles (the troubles are, that no keyboard or mice is working at such settings...)...

Works good, except the input devices. But maybe serial mouse could work...?

The problem on these boards is that the Holtek keyboard and PS/2 mouse controller is expecting a clock signal of 6 or 12MHz. When you set the FSB to 7Mhz the 24MHz clock now becomes 14.318 divided by 4 = 3,57MHz. Very low clock for the controller. The signal gets further divided by 2, because on normal FSB settings it uses 24MHz out of the clock generator. 24MHz is too much for the controller, but it is a perfect number to be divided. So dividing it by 2 or 4 gives a 12MHz or 6 MHz clock that the Holtek controller on the Asus boards uses.

In theory one can mod the motherboard to support all FSB settings without problems. Pins 2 and 3 are the clock inputs for the Holtek controller, so lifting or cutting the pins from the motherboard and then feeding it clock signal directly from a 12MHz clock oscillator should fix completely the input problems on these boards.

I had a skype session with Elianda who was very nice and gracious to help me test the undocumented 7MHz FSB setting on his ASUS TX-97XE and we had the exact same problems. The Holtek Controller simply cannot cope with the low clocks that are fed to it and the input devices stop to work.

OMG...! That is damn good idea, mate! I checked and yes - Holtek HT6542b controller have the pins 2 and 3 marked as OSCI/OSCO signals ( http://www.alldatasheet.com/datasheet-pdf/pdf … EK/HT6542B.html ) and they say it support from 6MHz to 12MHz clocks.

Now I'm fairly confident that I can create and add there a small circuit board with simple crystal driven clock generator, but if you can suggest schematics and components to build one, I would be *VERY* grateful 😉 It can be easily powered from 5V or 3.3V lines on the mainboard (just near the Holtek KB conrtoller I see empty places where filtring caps could be and I bet I measure a +5V there...), so a slight mod will fix all my problems... Now that would be absolutely stunning, great and amaizing - all in one!

I could order parts from Digikey for that... and get going. I hope they sell some nice 12MHz crystals 😀 Could you help me out with that project? In turn, I test anything anyone ever ask on that mainboard with any settings I manage to work (eg. not 83MHz FSB, that is over the top for my HDD... maybe with some PATA to SATA or CFcard reduction that can handle 41.5MHz PCI... but untill I have something like that, then FSB 75MHz is the best I could do).

What do you think about the serial mouse on COM connector?

I wish i could help you and help myself actually. My electronics expertise is i'm afraid to say non existant. This is simply in theory, if one can build a small clock circuit to feed the input pins on the controller directly problems with input on undocumented FSB settings.

In theory again if we had the schematics of these boards it could be that we could fix the PCI clocks at permanent 33MHz regardless of FSB settings. If you check the undocumented settings you'll see that when selecting the REF/2 setting(14.318 / 2) the PCI bus will be clocked at 3,57Mhz (14.318 / 4). This will probably negate the use of PCI cards which for me is important. I use a PCI VGA card to save me an ISA slot.

j^aws -

SYSINFO /DEMO ... The demo should run automatically.

Luckily, I checked that on normal, overclocked P90. It does not. First it display a red requester, asking for NORTON.INI ... enter fixed that and then it show hardware table and after a while the graph with measured speeds (give me 365.1 or something like that). So please provide the NORTON.INI, because I cannot use enter when downclocked. It does not work...

carlostex -

I wish i could help you and help myself actually. My electronics expertise is i'm afraid to say non existant.

That is sad, however you get a good idea. Maybe we can find someone who can suggest, what to do. First time I did something like that back in 1991 on Amiga 3000. Overclocking the FPU from 16 to 32MHz involved breaking one line on PCB, where the clock was supplied to the 68881 FPU and then soldering a wire from the crystal to the FPU pin. That was it.
Last time I checked was around 2001 and the machine was still operational, despite not even a cooling was added to the FPU. But - there was a guide for it, people done that before. As this TXP4-X goes, that is uncharted teritory. For example I have no idea, if I can just cut the lines on the bottom side of the PCB to stop the clock going into the controller. I would prefer that to cuting the legs, because cut on the line(s) on PCB is possible to repair, when things won't get as planed...

Teory about the 33MHz for PCI clock is nice, but there is a problem - we did not have the schematics. And unless Asus give them out (you can ask 😀 ), then we never will have them. So I would concentrate on the PS/2 problem with downclocking. And mine PCI S3 Trio64 v1 card is working hapilly on the slow PCI bus, so... why downclock would negate anything there? No reason 😀

trodas

I hadn't read the datasheet completely so i have no idea what kind of circuit or components the controller needs.

On page 7 you can see the diagram for a PS/2 mainboard implementation for the Holtek controller. On the bottom you can see 4 (possible?) connections for the crystal oscillator to the OSCI and OSCO pins or 4 (steps?) on how to implement it. I don't know how to interpret it properly but it seems you can pretty much use a 12MHz crystal oscillator almost directly. If those 4 points are steps (most probably) it seems still to be a rather simple circuit where you need to ground the signal at some points. Either way only the OSCI pin seems to be taking a signal. The Holtek controller then uses that signal internally and resends it to where it needs like keyboard clock and mouse clock.

Keep in mind this is a very crude and layman interpretation of electronics. It seems however, to be a very simple mod to do even for an amateur.

Well, according to the schematics it looks like that the Holtek HT6542b PS/2 controller has a built in clock driver, so all what is really need to do is to solder it on the legs 2 and 3 (while separating them from the mainboard first by cut & bend up) and I should be good to go.

So I ordred a 12MHz crystal and when it come, I give it a try and hopefully it works and overcome the problems at 7.14MHz FSB with input devices gone.

trodas wrote:

Well, according to the schematics it looks like that the Holtek HT6542b PS/2 controller has a built in clock driver, so all what is really need to do is to solder it on the legs 2 and 3 (while separating them from the mainboard first by cut & bend up) and I should be good to go.

So I ordred a 12MHz crystal and when it come, I give it a try and hopefully it works and overcome the problems at 7.14MHz FSB with input devices gone.

I'll be anxiously waiting for your results. I'm also pleasantly surprised to know that a PCI S3 Trio will work fine with the motherboard in the 7MHz FSB setting.

Yea, waiting for the crystal. Graphic: In VGA mode - noproblemo. I have video of the validation attempt, 32min long, if you are interesed...

In graphic mode it looks like this:

(S3 Trio64 v1)

I did not remember any graphic quirks on Win98se back in 2008, but that might be S3 Trio64 v2 ... or it might run in VGA mode only. All depends on what the Win98se do with the card w/o drivers... Or maybe too old caps on the Trio that cannot handle the superlow clock?