Are you sure the PSU is beyond any doubt?
Ageing / instable PSUs can cause all kinds of issues.

majestyk wrote on 2022-10-20, 05:16:

Are you sure the PSU is beyond any doubt?
Ageing / instable PSUs can cause all kinds of issues.

I don't think PSU is at fault. At present I'm using a brand new one.

Though the board has only a 20 (not 24) pin main connector and a 4 pin ATX12V connector for power.

A little bump on this one.
- The system is POSTing fine for the time being and I haven't encountered that dreaded beep for a while, either because I haven't moved the case much, or that I haven't booted any Linux kernels since then, though I cannot be really sure about this.
- Actually I'm still mostly bound by traditional IRQ limitations with this board, even though it does have APIC. As a result almost all PCI devices (onboard or external) ended up using IRQ9 which at one time created a non-reproducible disk corruption that borked a Windows ME install, forcing me to start over. Disabling most onboard devices that I'm not using, especially 3 onboard serial port headers, onboard audio, and the secondary NIC (100Mbps), freed a good amount of IRQs that some devices are now assigned to separate IRQs and this appears to have reduced the risk of further corruptions.
- Probably the backported NVMe driver (stornvme) for WinXP is not perfect. On this board I'm getting 0xC5 and 0x7E BSoDs with that driver in normal mode, though it loaded fine in safe mode. I also tried other backported drivers, but they either cannot detect anything or simply cannot start. On the other hand, the NVMe drive is working fine with Hiren's BootCD PE (Win10 based), and I could even write some files into it after formatting the disk there.

Another bump and a preliminary conclusion:
- I think there's something really bad going on with the CPU and the motherboard. Since a few days ago it simply wouldn't boot anymore. This time there's only the alternating high and low beeps and no more long beeps (RAM) mixing into it. Either the CPU has become faulty, or the motherboard is not really comfortable with 130W TDP CPUs, despite it was officially considered supported, given its limited power connectors. (NOTE: The highest clocked P4 processors appear to be rated around 115W TDP). I did not notice any oddities when looking at the BIOS' PC Health Status, however, as everything (temperature and voltage readings) appeared normal there.
- I think the motherboard's BIOS simply can't handle things like NVMe after all. Even though I could use the NVMe drive under Hiren's BootCD PE, after some writing (maybe about a few hundred files), the drive becomes inaccessible (a device which does not exist was specified), and some files are lost. The NVMe drive itself is okay as I had no problem accessing it on other devices especially on real PCIe environments.

So for now I have to end experimenting with that system. Guess with the limited power connectors on boards of these ages those TDP numbers can really be problematic, unlike today's motherboards which provide quite a few 12V connectors to allow uses of very high TDP CPUs and peripherals.

When's the board from? 2006/7ish? Think that's the tail end of the particularly bad capacitor era so have you had a close look at any big electrolytics on the board for any sign at all of bulging/leaking/leaning over? I've got a socket 939 from (I think) 2005 that had all the big caps go bad and leak.

snufkin wrote on 2022-11-07, 12:37:

When's the board from? 2006/7ish? Think that's the tail end of the particularly bad capacitor era so have you had a close look at any big electrolytics on the board for any sign at all of bulging/leaking/leaning over? I've got a socket 939 from (I think) 2005 that had all the big caps go bad and leak.

The BIOS seems to indicate a time around 2007-2008, but I'm not sure when such boards were actually made. It's an industrial motherboard that I got in around 2014-2015 as far as I could remember.

I remembered having that alternating high-low mixed-with-RAM beep issue the very first time I powered on the board. At that time I had to use a weak 533FSB Pentium 4 to boot it initially and make necessary setups before switching to the actual Pentium D CPU. While I didn't use that Pentium 4 on that board for too long, the board never had any issue with it. So either my Pentium D was somewhat faulty or the board was, again, not really good at handling high TDP CPUs (533FSB Pentium 4 CPUs were usually rated at around 84W TDP).

Anyway I'll have to take out that board and CPU when I have time to check what's really going on...

A very very late bump on this. I replaced the old CPU with another similar Pentium D (950 SL95V) and tried booting that board again, after a long hiatus.

It seems the board is very picky about memory. After some intensive testing I eventually found a pair of memory sticks that could reliably POST with this board.

Still, it remains the question what could be making the board this picky about memory sticks, and I wonder if it was possible to somehow "fix" a board's pickiness about memory to some extent...

The memory sticks themselves appear fine as whenever it POSTs successfully, the system would not misbehave at all, and I can even run MemTest86+ for many passes just fine.

And there's another thing that I intentionally left out previously: The board I'm currently running is my second one. The first one (which I got in around 2014-2015 as I mentioned in my previous post) died a few days before I started this whole thread: Its ATX12V (4-pin) header finally burnt out after many years of uptime (it's now effectively a short circuit as I have tested with a multimeter). It had this similar memory-related POST issue I'm having with the current one throughout the course of its life.

I bought this second (used) one which is almost identical to my first one shortly after its death, and it's having the very same issues as the first one, so I did not previously mention this back then, as I considered the issues generic to this particular board model and this (mostly if not entirely) ruled out the possibility that I was simply having a defective board.

PS: Considering the circumstances around the death of the first board, I think some boards of these eras are indeed not designed for running CPUs rated for higher TDP values, and care must be taken when using such CPUs there. It's also not an issue that can be addressed by using higher quality and higher rated PSUs, as old systems draw power in a very different way than today's standards.

EDIT: I'm still struggling with getting the board boot reliably. Looks like the board's memory slot is a bit loose and I could even feel it while installing and adjusting the memory sticks. Right now, with some good enough sticks, the board is in a state that, if I power it on while standing, the memory beeps, but if I lay it sideways, it could boot fine, and I can then stand it up until next reboot. Will see if tightening the memory slots using some long enough cable ties would help...

EDIT 2: Managed to get the memory slots tied with cable ties so it would stay in place, but it doesn't really solve the problems as the memory stick itself can still be moved though not as much as without the ties, and the board still refused to POST as usual... until I plugged in a debug card to inspect what might be causing it to fail, and it POSTed, and could continue POSTing fine even after I removed the debug card. I've put back the system to the place I intended to put it while it's still powered on, as I'm testing the integrity of the memory sticks I've tied in place. Most likely the thing would stop POSTing again after restart...

EDIT 3: An update on this... The cable ties do appear to work to some extent. The system still can't POST reliably, as I still have bad days that it could not boot, but the success rate has indeed improved a bit.