naujoks wrote on 2023-08-31, 15:41:

I have a Pentium II/III board with an i440EX chipset that's stuck at the boot screen.
It doesn't recognise the PIII CPUs as such. I have one PIII 500MHz, which registers as a PII MMX 333MHz, and a PIII 550MHz, with which the board doesn't post at all.

333MHz is OK. A multiplier is 5 , a bus is 66MHz, then 333MHz.

Try to clear CMOS and check CMOS battery.

P.S. IMO, the model is Shuttle HOT-671

P.P.S. Your PIII-500MHz is Katmai, but PIII-550MHz maybe is Coppermine - too new and the mainboard definitely doesn't support the latter

Excellent job at mainboard spotting! The board is indeed the Shuttle HOT-671.
I flashed the latest BIOS and the board is posting now.
With the FSB max being 83MHz and the multiplier, I guess I can't get the CPU set to be any faster than 415MHz?

naujoks wrote on 2023-08-31, 17:10:

Excellent job at mainboard spotting! The board is indeed the Shuttle HOT-671.
I flashed the latest BIOS and the board is posting now.
With the FSB max being 83MHz and the multiplier, I guess I can't get the CPU set to be any faster than 415MHz?

Correct, and that's an overclocked setting. The highest FSB the chipset officially supports is 66MHz.
The 440EX is a lower cost version of the 440LX. It was only intended for 66FSB Pentium-2 based CPUs. 75FSB would probably be reliable, not sure about 83.
It's likely that the fastest officially supported CPUs would be the P2 333MHz and whatever fastest Celeron that came in Slot-1, and unofficially, a slocket adapter with the Celeron 533MHz (Mendocino version, the black and silver one that looks like a Pentium MMX - *not* the Celeron "A" which looks like a P3).

Depending on the voltage regulator chip on your board, or if you can manually jumper it to 1.8V, then it might be able to boot Coppermines. Your experience with the P3-550 implies that your board can't do this, but maybe it only needs a BIOS flash or the 1.8V manual voltage setting.
If it can be convinced to boot a Coppermine CPU, then in theory it could run a Coppermine P3 based Celeron (like say a Celeron 733 for example) which runs a high multiplier on 66FSB. That's what internet folklore always said, anyway. Personally I've had a significant issue attempting that kind of combo. I tried using Coppermine and Katmai P3 CPUs with 3 different 440LX boards and all of them locked up whenever I tried to launch 3DMark. They worked perfectly with a Mendocino Celeron or P2 though.
I was collecting data to make a post about this back then but I waited because I wanted to try a PCI video card. It might only be an AGP issue. I've never had any luck buying PCI 3D accelerators from that era so that stage of the experiment never happened, and I forgot about it.

I'm still in the process of testing it, at 83 * 5.
I only have a non-3D PCI graphics card at the moment, and playing Thief with it is smooth and stable (at 800x600).
Today I should be receiving a 3D AGP card, I'm curious how that will perform.

naujoks wrote on 2023-08-31, 17:10:

Excellent job at mainboard spotting! The board is indeed the Shuttle HOT-671.
I flashed the latest BIOS and the board is posting now.
With the FSB max being 83MHz and the multiplier, I guess I can't get the CPU set to be any faster than 415MHz?

Well done! It's also excellent job of theretroweb guys.
I think you will be able to help them if you make a clear shot of the mainboard and send it them to update that lowres photo on the site

High resolution picture of the board would be lovely!
83MHz FSB is fine as long as your HDD and cards can keep up. Back in the day this fsb wasnt all that popular due to 1/2 PCI divider resulting in PCI overclock to 41MHz. Some hard drive/PCI IDE combinations silently corrupted data at that clock.

I'll definitely submit hires pictures.

Meanwhile the graphics card has arrived, an Elsa Gladiac (Geforce 2) MX from 2002, so quite a bit younger than the board, and it's not working right.
During the install of Win 98 SE everything is working fine, but when I install the drivers off the original driver CD, 3 seconds after the booting, the computer freezes completely.
I returned to 66MHz FSB for this, so no overclocking, the CPU is running at 333MHz.
Is the combo of too new CPU and too new gfx card the problem or is there something else I could try?

In BIOS settings should be set AGP aperture at least 64mb. There's also Nvidia driver - detonator. You can download different versions here from driver's part of the forum. Try to start from those early versions that already supports GeForce 2. Good luck!

naujoks wrote on 2023-09-03, 06:26:

Elsa Gladiac (Geforce 2) MX ..... 3 seconds after the booting, the computer freezes completely.
I returned to 66MHz FSB for this, so no overclocking, the CPU is running at 333MHz.

Its working fine at 66fsb? Afaik GF2MX had no problem working on AGP overclocked to 89mhz on 440BX boards with 133MHz FSB, so in theory shouldnt be a problem here, but you never know how Intel crippled BX to get EX.

I used one of the Detonator drivers now, it's difficult to say which version is the latest one that still has Win 98 SE drivers.
The card now gets identified as Geforce 2 MX and boots into the desktop fine, I can change resolutions and colour depth as I like.
However, as soon as a 3D application of any sort starts, whether it's a game or even one of the 3D screen savers, the computer freezes immediately.
I did have AGP aperture set to 64MB.

naujoks wrote on 2023-09-03, 22:29:

I used one of the Detonator drivers now, it's difficult to say which version is the latest one that still has Win 98 SE drivers. […]
Show full quote

I used one of the Detonator drivers now, it's difficult to say which version is the latest one that still has Win 98 SE drivers.
The card now gets identified as Geforce 2 MX and boots into the desktop fine, I can change resolutions and colour depth as I like.
However, as soon as a 3D application of any sort starts, whether it's a game or even one of the 3D screen savers, the computer freezes immediately.
I did have AGP aperture set to 64MB.

It looks like your system may have a similar problem discussed in the topic
Problems with Geforce 4Ti 4200 on retro build

issue with later (1999 and up) AGP cards that draw too much power

Try your GeForce card in an another more modern AGP-system or try a less power hungry AGP-graphics card in this system

It seems to me that my problem is a different one, since all 2D applications, including Windows desktop, installers, Word etc work just fine, but as soon as I start a 3D app, the computer freezes instantaneous.
It might still be a driver issue. I'll see if I can find yet another one.
Unfortunately I don't have another AGP board to try.

naujoks wrote on 2023-09-04, 10:26:

It seems to me that my problem is a different one, since all 2D applications, including Windows desktop, installers, Word etc work just fine, but as soon as I start a 3D app, the computer freezes instantaneous.
It might still be a driver issue. I'll see if I can find yet another one.
Unfortunately I don't have another AGP board to try.

It's worth to try.
You can find different versions of Nvidia Detonator here
http://www.vogonsdrivers.com/index.php?catid= … &menustate=14,1

naujoks wrote on 2023-09-04, 10:26:

It seems to me that my problem is a different one, since all 2D applications, including Windows desktop, installers, Word etc work just fine, but as soon as I start a 3D app, the computer freezes instantaneous.

2D draws minimal power compared to any 3D rendering.
Power draw might be a good lead. I see EZ1084CT near the AGP slot. Completely stupid and unnecessary regulator on ATX board https://eu.mouser.com/datasheet/2/761/SEMT_S_ … 5_1-2575352.pdf If its wired to AGP you should mod this board. 5A = 16W and thats divided between graphic card, chipset and ram.

3.3V power was a problem on this generation of motherboards, but the Geforce2 MX doesn't pull much power. It's hard to think of a 3D accelerator that draws much less.
If you switch to a Pentium 2 or Mendocino Celeron (which is P2 based), I bet it will solve it. This is the same problem I had with the 3 440LX boards I mentioned earlier. Yours might be case #4.

There seems to be a serious hardware issue with P3 on these chipsets. The thing that leaves me in doubt is that the whole internet always said Coppermine Celerons were viable on the 440LX/EX. Is it possible that nobody who said that ever tried it?
I wonder whether it's an AGP specific problem. I didn't have any PCI cards that were suitable to try.
Perhaps it's a configuration issue that some late BIOSes might correctly handle, but most don't?

If you can find any Beta BIOS that's newer than what you already have, you could try updating it, but if you don't have an external chip programmer you might not want to be too adventurous.
Also check if there's any AGP feature options that can be disabled in the BIOS settings.

rasz_pl wrote on 2023-09-04, 16:27:

naujoks wrote on 2023-09-04, 10:26:

It seems to me that my problem is a different one, since all 2D applications, including Windows desktop, installers, Word etc work just fine, but as soon as I start a 3D app, the computer freezes instantaneous.

2D draws minimal power compared to any 3D rendering.
Power draw might be a good lead. I see EZ1084CT near the AGP slot. Completely stupid and unnecessary regulator on ATX board https://eu.mouser.com/datasheet/2/761/SEMT_S_ … 5_1-2575352.pdf If its wired to AGP you should mod this board. 5A = 16W and thats divided between graphic card, chipset and ram.

I agree.
I don't like that LDO too because it can be a weak link. But there's no a schematics to know for sure what load is for it.
If load is a graphic card, a CPU, a chipset and ram then power budget calculation of the 3.3V LDO rail may be:
GeForce 2MX can draw up to 10W (typically a LDO 3.3V->2V US1150 - max 4A output current)
PIII-500 L2 discrete cache - up to 3.3V*1.2A=4W from the datasheet
440EX - up to 3W from the datasheet
SDRAM - ?

Also if C47 lost its capacitance the LDO output voltage would be unstable under high load

shamino wrote on 2023-09-05, 20:37:

3.3V power was a problem on this generation of motherboards, but the Geforce2 MX doesn't pull much power. It's hard to think of […]
Show full quote

3.3V power was a problem on this generation of motherboards, but the Geforce2 MX doesn't pull much power. It's hard to think of a 3D accelerator that draws much less.
If you switch to a Pentium 2 or Mendocino Celeron (which is P2 based), I bet it will solve it. This is the same problem I had with the 3 440LX boards I mentioned earlier. Yours might be case #4.

There seems to be a serious hardware issue with P3 on these chipsets. The thing that leaves me in doubt is that the whole internet always said Coppermine Celerons were viable on the 440LX/EX. Is it possible that nobody who said that ever tried it?
I wonder whether it's an AGP specific problem. I didn't have any PCI cards that were suitable to try.
Perhaps it's a configuration issue that some late BIOSes might correctly handle, but most don't?

If you can find any Beta BIOS that's newer than what you already have, you could try updating it, but if you don't have an external chip programmer you might not want to be too adventurous.
Also check if there's any AGP feature options that can be disabled in the BIOS settings.

The BIOS is the latest.
I suppose we discuss here PIII-500 Katmai not Coppermine. So maybe is not so bad about compatibility of the chipset and the CPU

Roman555 wrote on 2023-09-06, 10:44:

The BIOS is the latest.
I suppose we discuss here PIII-500 Katmai not Coppermine. So maybe is not so bad about compatibility of the chipset and the CPU

One of the video cards I tried on my LX boards was a Geforce2 MX AGP.
They locked up with Katmai and Coppermine, but worked fine with Deschutes and Mendocino.

shamino wrote on 2023-09-06, 23:32:

Roman555 wrote on 2023-09-06, 10:44:

The BIOS is the latest.
I suppose we discuss here PIII-500 Katmai not Coppermine. So maybe is not so bad about compatibility of the chipset and the CPU

One of the video cards I tried on my LX boards was a Geforce2 MX AGP.
They locked up with Katmai and Coppermine, but worked fine with Deschutes and Mendocino.

Yes, I agree about hangings of PIII on 440Lx mainboards. I've read it on forums. Supposedly the problem is SSE instructions which is feature of exactly PIII (not P2). 440Ex was presented later and the error was corrected