debs3759 wrote on 2022-09-16, 16:17:

Sphere478 wrote on 2022-09-16, 15:58:

2 hasn’t been known to mod to 3 or 3+ […]
Show full quote

dj_pirtu wrote on 2022-09-16, 14:30:

Got myself another P5A v1.06 today (with K6-2/500 and 128MB RAM), need to mod it and run some tests with K6-III+. Hope this one is OK.

2 hasn’t been known to mod to 3 or 3+

Only 2+

You can make a 2 with a 2+3+ though by moving the second resistor

Anyway, my personal experience with 1.06 hasn’t been that great. Recommend a 1.04

Pulling wt/wb to vcc3 using a 1k reisitor seems to have worked with others. On 1.05 1.06

I think he was talking about modding the motherboard to run the a K6/3+

I assumed. But for others reading who don’t know any better I tried to cover all the bases 🤣. 😀

Tried that P5A yesterday, not so good overclocker... With non-plus K6-2/500 I got only 105MHz FSB stable, 112MHz gave error in 3DMark00 and 120MHz didn't always give picture and when booted, in DOS nothing worked stable.
Display card was Voodoo3 AGP. Tried with loose memory timings too. Kept CPU near that 500MHz default clock.

Put it away and I will keep using TX97 (K6-3+/500 so 6x83FSB) with Geforce2MX PCI and Voodoo2 PCI, got some sound cards there too: Aureal Vortex2, SB16 and Gravis Ultrasound Classic. Rock stable system and not so much slower than ALI or VIA with 100FSB.

And no, I'm not using motherboard's cache at all, it only slows things down and anyway it's unstable @83FSB.

Try replacing the caps

I had enough recapping for awhile with my KT7A mobos, recapped three and none of them didn't benefit nothing from a recap. First one got only 136FSB, second one took 145FSB and third one works 150FSB.

Plain silicon lottery with those but when I get my motivation back again I'll try recapping P5A.

OK, new try with that P5A which I wrote earlier. Did that '+'-mod with 1.2k resistor, put a K6-2+/550 in there and did some testing.

This mobo takes now 120FSB, multiplier x5 and CPU voltage 2,2V so 600MHz and it seems to be perfectly stable. Don't know what was the problem before, maybe that non-plus K6-2 didn't like FSB at all.

Tried 124FSB too but that isn't stable, even if I raise IO-voltage 3,5V -> 3,6V. But 120FSB is really enough to make this fly with K6-3+.

EDIT: I was too optimistic, 120FSB wasn't fully stable, whatever I did. 115FSB needs 3,5V IO-voltage and Internal Page Detect: disabled in bios. K6-2+ takes 633MHz (115x5.5) but needs 2,4V 😁
EDIT2: And seems like stability is getting slowly worse... So it's the damn caps.

dj_pirtu wrote on 2022-11-21, 12:11:

EDIT: I was too optimistic, 120FSB wasn't fully stable, whatever I did. 115FSB needs 3,5V IO-voltage and Internal Page Detect: disabled in bios. K6-2+ takes 633MHz (115x5.5) but needs 2,4V 😁
EDIT2: And seems like stability is getting slowly worse... So it's the damn caps.

dj_pirtu, does the replacement of the caps help you?

I have similar problems with few P5A-boards here (1.04 and 1.06) - they are stable at FSB of max. 112 MHz (or less) only.

Over 112 MHz FSB, the picture in 3DMARK 99/2000/2001SE freezes at different times and in different tests (sometimes after few seconds, sometimes few minutes later).

Theoretically, all of my boards boot into Windows at FSB of max. 124 - but are no stable. No restarts in 2D, no other problems. SuperPi and Hot CPU Tester run very fine! The only problem is 3DMARK.

All of my boards does not have any visible damages on the caps...

I'm using a K6-2+ modded to III+, and the 1-kohm-mod on the 1.06-board.

Tried different VGAs (Radeon 9500 and GF 3 Ti 500) and 5 different PSUs - no difference in (un)stability.

Disabling "Internal Page Detection" in BIOS has no effect.

Using the ALi-AGP-Utility in Windows even in "safe"-mode does not help, too.

I haven't done that cap replacement yet 😀

By the way, the super strange problem with my P5A was surprise-reboots when idling on desktop. And instant reboot if you are playing midi file. Try to play canyon.mid and see what happens?

I'm quite sure all this weird unstability is caused by epic ripple in voltage and bad caps is the reason. If I only had oscilloscope to prove this.

gonzo wrote on 2023-06-19, 10:45:

dj_pirtu, does the replacement of the caps help you? […]
Show full quote

dj_pirtu wrote on 2022-11-21, 12:11:

EDIT: I was too optimistic, 120FSB wasn't fully stable, whatever I did. 115FSB needs 3,5V IO-voltage and Internal Page Detect: disabled in bios. K6-2+ takes 633MHz (115x5.5) but needs 2,4V 😁
EDIT2: And seems like stability is getting slowly worse... So it's the damn caps.

dj_pirtu, does the replacement of the caps help you?

I have similar problems with few P5A-boards here (1.04 and 1.06) - they are stable at FSB of max. 112 MHz (or less) only.

Over 112 MHz FSB, the picture in 3DMARK 99/2000/2001SE freezes at different times and in different tests (sometimes after few seconds, sometimes few minutes later).

Theoretically, all of my boards boot into Windows at FSB of max. 124 - but are no stable. No restarts in 2D, no other problems. SuperPi and Hot CPU Tester run very fine! The only problem is 3DMARK.

All of my boards does not have any visible damages on the caps...

I'm using a K6-2+ modded to III+, and the 1-kohm-mod on the 1.06-board.

Tried different VGAs (Radeon 9500 and GF 3 Ti 500) and 5 different PSUs - no difference in (un)stability.

Disabling "Internal Page Detection" in BIOS has no effect.

Using the ALi-AGP-Utility in Windows even in "safe"-mode does not help, too.

What speed RAM are you using when overclocking? Is your external (mobo) cache disabled? Also, you can try jumper VIO1 set to 2-3 for +0.1V to RAM, AGP, chipset, and i/o buffer (see manual), which may offer more stability.

I don't have a 1.06, but for the record, my P5A rev 1.03 with K6-2+ w/cache mod is very stable at 633MHz (115x5, 2.3V, mobo cache disabled, 256MB PC133 RAM @7ns, VIO1 default). I could not get 120MHz FSB working with any stability even with a somewhat reasonable amount of extra voltage.

wiibur wrote on 2023-06-27, 07:46:

gonzo wrote on 2023-06-19, 10:45:

dj_pirtu, does the replacement of the caps help you? […]
Show full quote

dj_pirtu wrote on 2022-11-21, 12:11:

EDIT: I was too optimistic, 120FSB wasn't fully stable, whatever I did. 115FSB needs 3,5V IO-voltage and Internal Page Detect: disabled in bios. K6-2+ takes 633MHz (115x5.5) but needs 2,4V 😁
EDIT2: And seems like stability is getting slowly worse... So it's the damn caps.

dj_pirtu, does the replacement of the caps help you?

I have similar problems with few P5A-boards here (1.04 and 1.06) - they are stable at FSB of max. 112 MHz (or less) only.

Over 112 MHz FSB, the picture in 3DMARK 99/2000/2001SE freezes at different times and in different tests (sometimes after few seconds, sometimes few minutes later).

Theoretically, all of my boards boot into Windows at FSB of max. 124 - but are no stable. No restarts in 2D, no other problems. SuperPi and Hot CPU Tester run very fine! The only problem is 3DMARK.

All of my boards does not have any visible damages on the caps...

I'm using a K6-2+ modded to III+, and the 1-kohm-mod on the 1.06-board.

Tried different VGAs (Radeon 9500 and GF 3 Ti 500) and 5 different PSUs - no difference in (un)stability.

Disabling "Internal Page Detection" in BIOS has no effect.

Using the ALi-AGP-Utility in Windows even in "safe"-mode does not help, too.

What speed RAM are you using when overclocking? Is your external (mobo) cache disabled? Also, you can try jumper VIO1 set to 2-3 for +0.1V to RAM, AGP, chipset, and i/o buffer (see manual), which may offer more stability.

I don't have a 1.06, but for the record, my P5A rev 1.03 with K6-2+ w/cache mod is very stable at 633MHz (115x5, 2.3V, mobo cache disabled, 256MB PC133 RAM @7ns, VIO1 default). I could not get 120MHz FSB working with any stability even with a somewhat reasonable amount of extra voltage.

Thanks for the tips. I already tested any possible options of them (disabling onboard-cache, increasing VIO etc.).
I am using different modules (but only one of them at the same time) of double-sided 256 MB SD-RAM PC133, CL2, Infineon-chips.
Maybe both of my VGAs do not support high AGP-frequencies...
I do not think, it's a CPU-problem, because the problem occurs even at lower CPU-frequencies (< 600 MHz) using a FSB > 112 MHz.

dj_pirtu wrote on 2023-06-27, 04:34:

I'm quite sure all this weird unstability is caused by epic ripple in voltage and bad caps is the reason. If I only had oscilloscope to prove this.

You are maybe right about a voltage-problem, at least for particular boards.
In Power Management inside of the BIOS, one of my 1.04-boards shows every few seconds a flacking "error"-message for any VIO > 3,6 V (tested with different PSUs).
On the other hand, all of my P5A-boards have the same kind of problem as shown above at FSB > 112 MHz, so it cannot be caused by the voltage only.
The very disappointing thing for me is, that sometimes most (but not all) of the 3DMARK-tests run pretty fine.
Also, I did a full maintenance of my VGAs (cleaning, new themal compound for the GPU etc.), so they are in perfect conditions.
But, anyway, I am still happy to use this board at least for a FSB of 112 MHz 😉