Maybe try disabling hibernation - admin command prompt:
powercfg -h off

Then go into sysdm.cpl and manually set the swap file size so it remakes it on reboot. I usually do a min of 2GB and max of 8GB.

Then once rebooted, turn hibernation back on -
powercfg -h on

Although hibernation is a piece of trash and always has been and ends up causing weird issues EVERY SINGLE TIME after a while.

UCyborg wrote on 2026-03-14, 09:53:

... (it's Win10, keeps session in both RAM and disk) ...

FWIW, I still use mostly Win7 but I don't think this has changed much...

When Win7 goes to "sleep" it always writes it's current state to disk.

Often at night or otherwise going away from it for longer, if I see the system is already sleeping, I just cut the power (I have it, monitors, speakers etc. all on a power-bar)

Next time I go to use it and power-on again, it always comes right back to where I left it with never a problem.

I can tell when it's sleeping because the "System has power" LED is flashing - I think this LED is part of ATX spec....
Does your system behave the same way? - Is the LED flashing before you (unsuccessfully) try to wake it up?
If yes: Is it still flashing after you try?
If no: Watch when it goes to sleep ... does the LED start flashing right after?

I doubt there is anything with the OS specifically, same happens on older Win10 install (installed in 2021) as well as newer Win11 (installed late 2023). Yes, nothing special with Win10/11 in that regard, hybrid sleep has been a thing since Vista. Page file is automatically managed.

The issue used to happen once in the blue moon. Now appears more like 2-3 times a week. It seems a bit more likely to happen if I leave the external USB disk plugged in. But I also started using sleep bit more more often, eg. instead of normal (hybrid) shutdown for the night. I reboot every several weeks to clear things up, some kernel memory leaks accumulate over time.

The hardware is housed in a bit of special case (Aerocool V-Touch Pro), no traditional LEDs, a touchscreen that only has separate HDD activity indicator in place of LED, which is non-functional...or rather perhaps only works for IDE disks while I use SATA. I'd have to check to be sure, but I'm almost positive the HDD LED wire is plugged in the right place. Screen is just dimmer when sleeping / hibernating / fully shutdown. It's normally bright when in this frozen state.

Windows has been reporting corrected hardware error on AMD Northbridge for many years now though, always when resuming from either sleep / hibernation (which includes Win10's hybrid shutdown, which basically logs off users and hibernates kernel).

1A corrected hardware error has occurred.
2
3Component: AMD Northbridge
4Error Source: Unknown Error Source
5Error Type: 14
6Processor APIC ID: 0
7
8The details view of this entry contains further information.

Details tab has this to say about the event:

1ErrorSource 0 
2ApicId 0 
3MCABank 4 
4MciStat 0x80000010000e0c0f 
5MciAddr 0x0 
6MciMisc 0x0 
7ErrorType 14 
8Length 936 
9RawData 435045521002FFFFFFFF03000200000002000000A8030000053803000F031A140000000000000000000000000000000000000000000000000000000000000000BDC407CF89B7184EB3C41F732CB57131B248949139377F4BA8F1E0062805C2A3D983643B0A9DDB0100000000000000000000000000000000000000000000000058010000C00000000003000001000000ADCC7698B447DB4BB65E16F193C4F3DB0000000000000000000000000000000002000000000000000000000000000000000000000000000018020000800000000003000000000000B0A03EDC44A19747B95B53FA242B6E1D0000000000000000000000000000000002000000000000000000000000000000000000000000000098020000100100000003000000000000011D1E8AF94257459C33565E5CC3F7E8000000000000000000000000000000000200000000000000000000000000000000000000000000007F010000000000000002040000030000420F10000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000007000000000000000000000000000000420F10000008040009208000FFFB8B170000000000000000000000000000000000000000000000000000000000000000B3F8F31CB1C5A249AA595EEF92FFA63C01000000000000009E07C0800600000000000000000000000000000000000000000000000000000000000000000000000200000002000000AD8530A62FB4DC010000000000000000000000000000000000000000040000000F0C0E00100000800000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000 

I spotted this error randomly at some point in time, no idea when it started happening and if it's related. Well, it's not young anymore and no computer before it has been as reliable. Wouldn't be surprised if the extra slower stick of RAM, the Kingston, which was only installed 2 years ago, is a bit flakier. Guess I could run some stability test to see if things are still stable, I haven't noticed anything odd otherwise.

I've read somewhere power up is supposed to be most stressful for computer. Not sure if resuming is supposed to be equal, no issue with normal power up, just when session is kept in RAM.

UCyborg wrote on 2026-03-15, 13:54:

... no traditional LEDs, a touchscreen that only has separate HDD activity indicator in place of LED, which is non-functional...or rather perhaps only works for IDE disks while I use SATA. I'd have to check to be sure, but I'm almost positive the HDD LED wire is plugged in the right place. Screen is just dimmer when sleeping / hibernating / fully shutdown. It's normally bright when in this frozen state.

Do you have another case or means of connecting the normal power-ON LED?

It would be very helpful to know what state the system thinks its in .. does it actually know it's sleeping?

The display "could" monitor the power-ON LED line and dim when it flashes, but I think more likely it's related to power supplies, as some things get shut off while sleeping.
But you can't tell if going to sleep completed correctly when going to sleep, and/or if it still knows it sleeping by the time you try to wake it up.

If you can find some way to monitor the LED output that might help....

Could this possibly have to do with the +5v sb rail being iffy? I seem to remember similar issues in systems once upon a time when the capacitor plague was a big thing. Since the +5v sb rail in PSUs is usually pretty low spec and low tech, it would usually be the first to get out of whack once it's filter cap started going bad.

I don't have any other case. But the bad +5V SB rail theory sounds logical. Leaving external HDD plugged in must add to the load, I remember hearing the disk in this case while computer was silent. So in theory, the easiest fix would be replacing the power supply and see how it behaves?

UCyborg wrote on 2026-03-15, 18:01:

I don't have any other case. But the bad +5V SB rail theory sounds logical. Leaving external HDD plugged in must add to the load, I remember hearing the disk in this case while computer was silent. So in theory, the easiest fix would be replacing the power supply and see how it behaves?

Yeah a PSU swap would be the easiest thing to try. Personally I'd open it and swap said capacitor just because it's fairly easy but I get not everyone may feel up for that. Iirc the +5v sb cap going bad would sometimes also result in a faint high pitched squeal when plugged in and off, may audibly change when you load the rail. This was often a dead give away to me when plugging in unknown/untested power supplies/systems - whine while off (or in standby) = almost always questionable power supply

bertrammatrix wrote on 2026-03-15, 15:24:

Could this possibly have to do with the +5v sb rail ...

Good point - I hadn't thought about the SB rail - most common PSU problems are with the high-power active rails, but yes it's going to be the SB rail used to begin restart after sleep.

If you have a multimeter I'd highly recommend check the SB rail for working voltage during sleep/restart-attempt, even better if you have a scope you could look for stability and noise.

But, even without those, definitely worth trying a known good PSU.

There are two extra devices that I may have plugged in at times when resuming, a smartphone and the external HDD. Sometimes just the phone, sometimes both, sometimes none. Last couple of resumes with none went through. And I guess another notable difference, the computer had 2 RAM sticks for most of its life, now it has 3. But I suppose flawless power supply should handle the case when you have every possible internal power consumer active that draws power during along with all possible external USB devices?

I have a multimeter, but figuring out where exactly to put the diodes would take some head scratching.

I'll be curious to know whether the situation improves with a different PSU. My retro desktop had been suffering from such occasional symptoms off and on. I had been suspecting the +5V standby rail, but in my case it almost always eventually ended up with other instability and crashes which clearly pointed to other components. However, even now, once in a while it fails to wake up from sleep. And the PSU in that system is probably 18 years old...

UCyborg wrote on 2026-03-16, 19:30:

There are two extra devices that I may have plugged in at times when resuming, a smartphone and the external HDD. Sometimes just the phone, sometimes both, sometimes none. Last couple of resumes with none went through. And I guess another notable difference, the computer had 2 RAM sticks for most of its life, now it has 3. But I suppose flawless power supply should handle the case when you have every possible internal power consumer active that draws power during along with all possible external USB devices?

I have a multimeter, but figuring out where exactly to put the diodes would take some head scratching.

if you have a powered USB hub you could try that before trying another PSU.
Connect your smartphone & hard drive to the powered USB hub and see if the resume still works ok.

UCyborg wrote on 2026-03-16, 19:30:

There are two extra devices that I may have plugged in at times when resuming ...

Typically some USB ports are powered by the SB power supply (so things like mouse/keyboards can be used to wake up the system), but often some are not (for things like hard-drives etc.) .. you need to refer to the mainboard docs to be sure.

Usually the "always powered" ones are a different color, but best way to be sure it use a simple "USB tester" which shows the voltage and current drawn at a USB port. If it still shows 5V when the system is powered-OFF, don't use it for higher power things that you don't want powered when the system is OFF.

Even if you don't have a USB tester, some devices (flash drives, mice, keyboards, phones) may have a indication when powered - make sure this indicates power and not "initialized" - a good way to do that is to plug it into a USB battery and see if it lights.

And do be aware that originally USB had a 500ma maximum current, this has gone up with wanting to charge newer phones in less than a week... but those powered by the SB rail typically follow the original spec. of 500ma (so having an HD or phone plugged into them might overstress / cut-power)

And... also make sure you've not moved your mouse/keyboard to one that is not SB powered. (but you should always be able to resume from sleep with the power-button)

weedeewee wrote on 2026-03-16, 21:11:

if you have a powered USB hub you could try that before trying another PSU. ...

Not entirely sure this would be valid - the concern is the the SB rail may be "unstable", possible by bad-caps, overstress from high-current devices ...

Removing external loads may not make it "work like new".

UCyborg wrote on 2026-03-16, 19:30:

... I have a multimeter, but figuring out where exactly to put the diodes would take some head scratching.

Not sure what you mean by "diodes" .. assuming you mean test-leads/probes.

The 5V SB rail is on the ATX power connector - google "ATX power pinout" .. you want to measure between that pin and ground (also on the ATX connector, but also easy to get on a floppy/IDE power cable)

With the system plugged in, but powered OFF - you should see +5v on the SB-power pin relative to ground.

You're right, probes is what I meant.

Welp, today it failed without external disk or phone plugged in. 😒

I have to check if manual says anything about powered USB ports in standby, I don't recall that it does.

UCyborg wrote on 2026-03-17, 10:17:

... Welp, today it failed without external disk or phone plugged in. 😒

There's many reasons old PSUs can fail, bad capacitors being a main one .. so it might not be actually related to things being plugged in during power-OFF., and even if it is, PSU can still fail in ways that lack of load wouldn't "fix".

Still worth trying a new PSU. Recover from sleep is a very small amount of "extra stuff" on the mainboard, and the fact that it works fine other than this makes me suspect that the board doesn't have a serious problem.

I wonder what to choose. Quite few names to choose from; ASUS, Corsair, Gigabyte, Be Quiet!, Chieftec, whole bunch of less known brands. Shopping today is PITA. Though LC Power wasn't particularly reputable and yet the PC made it this far. But it's said these things last longer when you don't work them so much.

There are horror stories out there about some power supplies catching fire / exploding!

Seasonic.

Make sure you get at least 80+ Bronze certified.

I will trust no other brands for the most part unless I can verify the OEM that made them and can find actual testing done on them in regards to efficiency and ripple.

All others should be treated as unreliable.

Agreed with Seasonic ... I've had good lock with Corsair as well.
(and I've still got some old cheap generic AT supplies from the 1990s that are still going strong- they really don't make then like they used to)