Another round of testing done. CPU is now the slot-1 PIII@500 (100 Mhz x 5) with 2 x 64MB PC100 SDRAM.

Installed Windows 2000 and the results are as follows:

• TNT2 M64, FX 5200 Ultra and FX 4000 are 100% working at any resolution and color depth.
• Geforce4 MX440 first time booted OK and was able to set 1024x768@32, after reboot I got scrambled graphic but Windows didn't crash, it was still usable. Reverting to 640x480@16 didn't fix the issue even after a reboot.
• FX5600 and Quadro4 same issue of Win98. Black screen after Windows boot one.

So is it a power issue? Because the only two cards that worked are the less power hungry of them and the only one with a molex for additional juice. While the GF4MX440 is right on the edge of the power limit.

I was feeling confident and I put back the Win98 drive and... nothing, issue still present even with the three cards that are working under Win2k.

AlexZ wrote on 2023-09-19, 18:15:

Reinstall Windows. FX 5600 works fine on my 440BX.

Ehm.. trust me I reinstalled Win98 basically after every test 🤣 I wrote a batch script that formats the drive and launches the unattended Windows setup.

shevalier wrote on 2023-09-19, 18:50:

Check the position of the Vio jumper. And hardware monitoring to. AGP8X cards react very poorly to voltages greater than ~3.5V w […]
Show full quote

Check the position of the Vio jumper.
And hardware monitoring to.
AGP8X cards react very poorly to voltages greater than ~3.5V when running on AGP2X.
Most often they hang under load.

Sure, I will definitely check the voltages.

mockingbird wrote on 2023-09-19, 17:28:

Wait a moment -- don't be so quick to draw conclusions from posts by strangers on internet forums...

I mean sure but I opened this post to brainstorm ideas, so any opinion is welcome. For example, I hadn't really thought about installing Windows 2000 before they advised to do so.

Now that does look more like a power problem, still weird that w98 outright refuses to load driver while w2k is more forbidding agreeable.

Asus P2B boards have a provision to source 3.3V from ATX connector (Diode D33), but its left unpopulated and instead separate onboard power supply is used.
plus: You can bump 3.3V to 3.45V for better ram OC
minus: limited power output

P3BF seems to be continuing this design, but I cant verify without diagram/boardview. You can check by measuring resistance between AGP 3V pins (for example pin 60, or just count 6th 7th pins from the side near blue switch block) and AGP pins 1/2/11.

First thing to try would be switching VIO jumper in upper right corner of the board.
Ideally you would setup oscilloscope on 3.3V supply line and run one of the 3D tests hanging computer to see if voltage sags/rippless excessively, but I suspect you dont have one.
Hardcore time tested solution is setting onboard 3.3V supply to normal 3.3V setting and soldering a thick wire between ATX 1/2 pins and AGP 9/71.

In my last post the page I linked said in the paragraph above what I quoted:

One practical matter which must be considered is the fact that some of the original AGP 1.0 motherboards do not provide enough power to operate some newer AGP video cards reliably. For example, some of the original motherboards using the first chipsets which supported AGP (like the Intel 440LX and 440BX) can become unstable if you install video cards which draw lots of power through the AGP slot. The motherboards can't always supply the necessary current for the newer video cards. So if you're adding a video card to an AGP 1.0 motherboard then it's a good idea to install a video card which doesn't consume very much power.

But if its power then its even weirder that some cards work under Win2k and not Win98, maybe your board is possessed. I know old hardware seems that way at times.... just like the fun we had back in the day but with more years of wear and tear.

Sometimes we make it harder on ourselves by mixing hardware that was never designed to be together.

shevalier wrote on 2023-09-19, 18:50:

Check the position of the Vio jumper. And hardware monitoring to. AGP8X cards react very poorly to voltages greater than ~3.5V w […]
Show full quote

Check the position of the Vio jumper.
And hardware monitoring to.
AGP8X cards react very poorly to voltages greater than ~3.5V when running on AGP2X.
Most often they hang under load.

VIO jumper was on default, but I started experimenting with three different motherboards because I could not get consistent results... To be continued...

In mine the VIO jumper is in the correct 1-2 position for 3.5v.

Later today I will try to measure the actual voltages.

Grem Five wrote on 2023-09-20, 00:14:

Sometimes we make it harder on ourselves by mixing hardware that was never designed to be together.

Yeah definitely this is true too... But sometimes this is the fun part, well at least until something break!

rasz_pl wrote on 2023-09-19, 23:09:

Asus P2B boards have a provision to source 3.3V from ATX connector (Diode D33), but its left unpopulated and instead separate onboard power supply is used.

Sorry, a diode with a voltage drop of 0 volts has not yet been invented, this is a diode for soft start DC/DC. This doesn't affect anything.

mrfusion92 wrote on 2023-09-20, 06:09:

In mine the VIO jumper is in the correct 1-2 position for 3.5v.

Correct position - no jumper.
Re: Voodoo 3 with P3 933?
and all topic.

This is nonsense from Asus, like for better overclocking of very old video cards, whose core and memory are directly powered by the AGP.
In AGP8X cards, only AGP interface are directly powered, and with AGP2x (3.3V signals) they hung at increased voltage.
For me, 3 out of 3 cards hung under load at 3.6V.
At 3.5 they don’t hung, but this doesn’t really give anything, alas.

my P3B-F does not work stable when jumper set to default position 1-2. only on 2-3 (3.65v) being stable. ( i use 4 sticks of 256mb ram). same voltage goes to agp.

dm- wrote on 2023-09-20, 07:38:

i use 4 sticks of 256mb ram

50/50
Sometimes this works, but it's better to find 4 sticks with CL2.

shevalier wrote on 2023-09-20, 06:52:

Sorry, a diode with a voltage drop of 0 volts has not yet been invented, this is a diode for soft start DC/DC.

this is a diode for nothing because its unpopulated 😀 but you could desolder L16 and solder a wire in place of D33

dm- wrote on 2023-09-20, 07:38:

my P3B-F does not work stable when jumper set to default position 1-2. only on 2-3 (3.65v) being stable. ( i use 4 sticks of 256mb ram). same voltage goes to agp.

I think 4x unbuffered 256MB was officially unsupported on 440BX. There are funny boards with buffer chips between DIMM slots. P2B-D has buffers in front of all DIMM slots. It makes sense you would need bumped voltage to run such setup.

I had time to do a few tests during my lunch break.

Removed the VIO jumper and the FX 5600 started working... but plugged the jumper back in and it still works now?! Maybe the motherboard is truly possessed... I really didn't change anything else.

Without the jumper, same black screen with the Quadro4 and corrupted graphic with the MX440. And of course under Win98 nothing changed at all with the jumper removed, all the cards just go to black screen.

About the AGP voltages, there is a safe way to sample them from the top side of the motherboard?

EDIT: Oh by the way, if you are wondering if the MX440 is just faulty, I tested again in another system and it works okay.

mrfusion92 wrote on 2023-09-20, 12:32:

Removed the VIO jumper and the FX 5600 started working... but plugged the jumper back in and it still works now?! Maybe the motherboard is truly possessed... I really didn't change anything else.

P3-series motherboards have a fairly powerful 3.3V power supply.
It’s not that it can powering a monster like the Radeon 9800 XT, but these motherboards power all sorts of GF 5200/5600 without any problems.
This is some kind of madness with a specific exemplar of the motherboard.
If you have a soldering iron and three decent capacitors, I would recommend replace these 3.

Well I figured out my issue after testing with several motherboards... I extricated the JMicron SATA to PATA adapter and used the Marvell which I had previously accused of being incompatible with the PIIX4 southbridge which is definitely not the case... When I traced my issues some time ago to a memory stick, I did not test again with the Marvell. The gist of it is that the JMicron misbehaved at 133Mhz FSB on the PIIX4 and that's what was causing my issues.

Just to clarify: The issue I was having was that after going into Quake and setting the graphics options, it would crash when it restarted itself to apply the settings.

With the Marvell adapter, the ti4200 8x works fine on my P2B (1.04) at an 89Mhz AGP clock frequency. I didn't test again with the P3B, but I'm certain I'll get the same successful result.

By the way, the resulting AGP frequency is not shown anywhere in the Asus BX BIOS.
Or what divider is used - 1:1 or 2:3.
It might be worth moving the jumper to the “manual” position and setting it to 2:3 explicitly.
It may easily turn out that the AGP frequency is 100 MHz, instead of 66.

mockingbird wrote on 2023-09-20, 14:46:

Well I figured out my issue after testing with several motherboards... I extricated the JMicron SATA to PATA adapter and used the Marvell which I had previously accused of being incompatible with the PIIX4 southbridge which is definitely not the case... When I traced my issues some time ago to a memory stick, I did not test again with the Marvell. The gist of it is that the JMicron misbehaved at 133Mhz FSB on the PIIX4 and that's what was causing my issues.

hmm I dont know, P2B/P3B-F have 1/4 PCI divider so IDE is not overclocked. I would go back to basics, 66MHz fsb, 1 dimm, no other cards.

mrfusion92 wrote on 2023-09-20, 12:32:

About the AGP voltages, there is a safe way to sample them from the top side of the motherboard?

sure, on the inductor in the corner near VIO jumper, but measuring only makes sense if you have oscilloscope

shevalier wrote on 2023-09-20, 18:20:

By the way, the resulting AGP frequency is not shown anywhere in the Asus BX BIOS. Or what divider is used - 1:1 or 2:3. It migh […]
Show full quote

By the way, the resulting AGP frequency is not shown anywhere in the Asus BX BIOS.
Or what divider is used - 1:1 or 2:3.
It might be worth moving the jumper to the “manual” position and setting it to 2:3 explicitly.
It may easily turn out that the AGP frequency is 100 MHz, instead of 66.

There is a DIP switch and no auto setting, just x1 (on) or x2/3 (off) and it always has been in the off position.

But I think I concluded it's not a freq issue, yesterday I swapped the CPU back with the PIII@933 (133x7) and under Windows 2000 all the cards behaved the same.

I have also checked the AGP voltage, I don't have the oscilloscope but apparently, at least in my motherboard, there is no difference between VIO jumper in 1-2 position or no jumper at all. The reading was always ~3.5V. Instead in 2-3 position was >3.6V as the manual reports.

shevalier wrote on 2023-09-20, 14:02:

If you have a soldering iron and three decent capacitors, I would recommend replace these 3.

I do have the replacement caps (in another complete dead P3B-F on which I replaced all the caps as last hope... do you see that I'm right in saying that I'm unlucky?). So I might try that too.

rasz_pl wrote on 2023-09-21, 03:45:

hmm I dont know, P2B/P3B-F have 1/4 PCI divider so IDE is not overclocked. I would go back to basics, 66MHz fsb, 1 dimm, no other cards.

I will test again with the P3B-F at some point to confirm my diagnosis, but everything was working perfectly on the P2B 1.04, so I'm fairly certain.

mrfusion92 wrote on 2023-09-21, 06:20:

do you see that I'm right in saying that I'm unlucky?).

When there are unexplained problems with a video card, they usually start testing it.
But from personal experience, it’s worth starting with testing the RAM and CPU.
When RAM is buggy, there are all sorts of glitch. Including the video card, because the video driver is located in system memory and is executed by the CPU.
Although, may be unlucky with the motherboard too

mrfusion92 wrote on 2023-09-21, 06:20:

I have also checked the AGP voltage, I don't have the oscilloscope but apparently, at least in my motherboard, there is no difference between VIO jumper in 1-2 position or no jumper at all. The reading was always ~3.5V. Instead in 2-3 position was >3.6V as the manual reports.

This means, that the bios still says ~3,44V in the Hadware Monitor without any VIO Jumper? Same as position 1-2?
3,50V is only interesting for FSB133 on the i440BX, but for FSB 66/100 is 3,30V more than enough.

The P3B-F is produced for FSB133 marketing on the BX. So no 3,30 default VIO settings.

I re-tested with my P3B-F, everything is stable at 133Mhz FSB in Quake3 1.32C. I prefer the P2B though -- the P3B has detection issues with my SSD. Not sure if it's only this SSD or SSDs in general, but I have to keep rebooting until it gets detected.

At first I experienced that same problem again of Quake3 hanging when restarting after applying graphics settings, so I erased the Quake directory and re-installed.

My tests were done with a Slot1 P3 866. If you're going to run a Coppermine or Tualatin on this, use a Slot 1 CPU or make certain you have a good slotket. With my lin-lin and a generic slotket, it boots and seems to run fine, but crashes during the last stage of Windows setup with a P3-S 1.4Ghz. I might re-test that with a good slotket at some point later on.

The only other thing I can think of that I did special for this is use Rob Lowe's PTCHSATA... I suspect that this patch is also useful for SATA drives that are connected to an IDE port through a SATA to PATA bridge.

NostalgicAslinger wrote on 2023-09-21, 17:01:

This means, that the bios still says ~3,44V in the Hadware Monitor without any VIO Jumper? Same as position 1-2?
3,50V is only interesting for FSB133 on the i440BX, but for FSB 66/100 is 3,30V more than enough.

Are you a wizard or something? Because that's exactly the value it says. And there is no difference between no jumper and 1-2 position.

Yesterday I received two "new" graphic cards, a Nvidia Quadro2 MXR and a Matrox G400. Both 4x universal, and guess what card doesn't work under Win98? Well of course the Nvidia!

I really don't know what else try, I guess I will keep the motherboard with the G400 which is at least period accurate.