So that quirk only happens in higher resolutions and in windows? No problems at all at 320x200 ? If you have the video just post the link 😀

There is the video with attempted validation:
http://www.youtube.com/watch?v=IsA0UtOXP_o

The fail is the time. To save the validation file, I give the machine 6,5 minutes. Bad idea! When it should produce equal amount of work, the report.txt time, it took about 8h...! Mostly because high CPU load. To see, if the CPU-Z already exited (meaning work is done) I run the taskcmd as well, but just updating the display consume at 10.7MHz almost all CPU time...!
So I run it at evening and at morning, 7:07h the report was done, while we are sleeping...

Next attempt will be w/o limited time, 🤣. Too slow. But you see plenty of the problems with such slow PCI bus. Using just VGA output is fine, tough. So I need to either force WinXP to not have support for this S3 Trio64 card (to make them run in VGA mode, witch is okay) or change the videocard to one, that is not supported by Win internal drivers (again forcing VGA mode).

Best solution is probably to trigger the BSEL = 0 chice on the PLL chip:

If I understand that right, then it is possible to lock on the PCI bus to 32MHz regardless of the FSB clock. That would allow me to underclock to 7.14MHz as well, as to overclock into 83MHz FSB w/o problems (unreliable IDE controller) on the PCI bus...!

Selecting BSEL = 0 could be pretty cool mod and I cannot understand, why Asus did not do it already. When using PLL chip that can do PCI lock - it is almost a crime not to utilize that...! Somebody should be held responsible for that epic fail!

^^ BSEL = 0 looks like "REF/3", i.e. 14.3/3 ~ 4.8MHz, rather than BSEL = 1 which looks like "REF/ 4", i.e. 14.3/ 4 ~ 3.6MHz...

So, judging by that chart, the PCI bus would be ~ 4.8MHz, rather than ~ 3.6MHz.

Yep, but still better 4.8MHz that 3.6MHz and in the 83MHz case it will be all okay at 32MHz.

The question is only "how to activate this" 😢

Working first on the functional keyboard and mice - there are few preliminary photos (the big tin bulges need cleaning, 🤣) of the mod that I done:

And the keyboard and mouse work nicely, but since I have there a K5-75, I did not tried yet the 7.14MHz settings as they did not work with the K5-75 🙁

...

As for the BSEL = 0, then the BSEL = 1 is triggered by connecting 10k resistor to +Vcc (3.5V). Disconecting it, soldering it into one leg, up into the air and connecting the high leg with ground force the BSEL = 0, according to this PDF guide:
http://web.archive.org/web/20000919074447/htt … df/9169c-27.pdf

Could someone confirm my ideas?

trodas wrote:

And the keyboard and mouse work nicely, but since I have there a K5-75, I did not tried yet the 7.14MHz settings as they did not work with the K5-75 🙁

Don't you have other Socket 7 CPU's to test this?

Nope. Only P90 and AMD K5-75. Going to buy something, but don' t know witch CPU have higher chance of working with 7.14MHz FSB 🙁
I would also like to know, if I could manage like 550MHz CPU's or somethning like that...

But to sucesfully be able to force x2 multi I need K6 with the "+" only, that is the first limiter. Apart from that... no special choice. Looked on eBay and find plenty of AMD K6, but most are not "+", so no interest. Some "+" are nice, especially around the 500MHz mark, but... will they post at all? I don't know. I seen AMD K6 2+ in almost same TXP4 mobo:
http://hwbot.org/submission/2221547_patriot21 … 0mhz_499.83_mhz

...

And the Asus TX97-XE, witch bios I use, running with K6 233:
http://hwbot.org/submission/757107_azesmbog_w … min_53sec_352ms

...but that is all withing the specs of the mobo, as the highest support is for K6 233MHz. No mention of the "+" ones and no mention of higher clocks. The voltage settings are pretty limiting too: 1.8, 1.9, 2.5, 2.7, 2.8, 2.9, 3.2, 3.4 and 3.5V. Big jump between 1.9 and 2.5... but with good cooling it should not be a problem. More I wonder, if there is some software that can set ram timings on the Intel 430TX chipset, as I cannot set the TRP to lower that 3 settings. Therefore ending with 2-2-3-4 (7) settings. Bad. I want 2-2-2 ... and perhaps 5 (7) for fastest speed... but definitively TRP 2, not 3. That make huge difference...

I trying to get my hands on any K6 locally, but no luck now. One fella in Czech rep. says he have the CPU, so I offered nice price _IF ACTED FAST_ and no response ever since... I quess that plenty of people are intersed, if the AMD K6 2/3 CPU's will work there w/o bios tweaking...

Patched BIOS, with K6-III+ support, can be found here:
http://web.inter.nl.net/hcc/J.Steunebrink/k6plus.htm

manual page 16 2.0 volt setting:
http://www.systemhaus-brandenburg.de/download … p4-handbuch.pdf

maybe worth reading:
http://www.tomshardware.com/reviews/oldie-tuning,216-2.html

trodas wrote:

Nope. Only P90 and AMD K5-75. Going to buy something, but don' t know witch CPU have higher chance of working with 7.14MHz FSB 🙁
I would also like to know, if I could manage like 550MHz CPU's or somethning like that...

Pentium 90 should work fine. When Elianda and i made our test we used a Pentium 100 if i'm not mistaken. The P90 should work with 7.14MHz FSB. Just try it and see if the crystal mod fixes the keyboard problems.

trodas wrote:

... As for the BSEL = 0, then the BSEL = 1 is triggered by connecting 10k resistor to +Vcc (3.5V). Disconecting it, soldering it […]
Show full quote

...
As for the BSEL = 0, then the BSEL = 1 is triggered by connecting 10k resistor to +Vcc (3.5V). Disconecting it, soldering it into one leg, up into the air and connecting the high leg with ground force the BSEL = 0, according to this PDF guide:
http://web.archive.org/web/20000919074447/htt … df/9169c-27.pdf

Could someone confirm my ideas?

"Shared Pin Operation - Input/Output Pins 5, 12, 13 and 32
on the ICS9169C-27 serve as dual signal functions to the
device."

and

To program (load) the internal configuration register for
these pins, a resistor is connected to either the VDD (logic 1)
power supply or the GND (logic 0) voltage potential. A 10
Kilohm(10K) resistor is used to provide both the solid
CMOS programming voltage needed during the power-up
programming period and to provide an insignificant load on
the output clock during the subsequent operating period.

I can see that Pin32 is BSEL - what I'm not sure is where exactly you place the resistor? Between Pin32 and VDD, or Pin32 and GND? I'm sure someone could clarify...

I can see that Pin32 is BSEL - what I'm not sure is where exactly you place the resistor? Between Pin32 and VDD, or Pin32 and GND? I'm sure someone could clarify...

Well, there is a 10k resistor right near the Pin 32 and on the other side of it, is Vcc 3.16V for the chip. So it is fairly good bet, that this is the one. My plan was to solder a wire from Pin 26 (GND) to this resistor, while I solder it by just one side to the PCB (the connection to the Pin 32), while other side will "hang in the air" and I solder the wire to that end.
That should give the "zero" byte on start, forcing BSEL = 0 and giving locked PCI bus clock.

Why I say "plan was"? Because when the 83MHz FSB setting works stable, then there is suddently no need for this. My friend told me before, that FSB 83MHz was NOT stable, but maybe the keyboard mod played it's role as well, but now it IS stable, as you can check from these kinda long runs:

So, unless this mod can fix the L2 cache, then I have no desire to do it. I already soldered the wire and... it is just about to desoldring the resistor, soldering it back "on top position" and soldering the wire to the "air" end = fairly simple, but the PCI clock is not used for the cache. Instead, IMHO a FSB clock is used for the cache...

Witch is why when I enable the "Speculative Lead Off" the mobo at 83MHz settings refuse to boot... 🙁 It does post, tough. But that it is...

Trodas, do you mind testing the Pentium 90 with the 7.14MHz FSB and see if the crystal mod fixes the keyboard troubles with this setting?

Already done. The "crystal mod" fixed all the input devices troubles at 7.14MHz FSB. Machine run just great. The only problem I encountering is the IDE speed / booting support, but that is another topic: SSD boot on Asus TXP4-X impossible?