It is not a instant fry. A 3,3 Volt 486 will work at 5 Volt, but die eventually.

Between my “instantly” and your “eventually” can we talk about “seconds”, “days”, “months”, “years”?

You can rule out seconds. I had a DX4 accidentally running at 5v for a few days. It’s still OK now at 3.3v years later.

If you have proper cooling, I guess they can run at 5V for quite a long time. It probably depends on the quality of the silicon, but I've read numerous reports of people accidentally running their DX4s at 5V for years.

Personally I wouldn't run much above stock voltage unless you really don't care about the CPU. From what I remember hearing from people overclocking 486 type CPUs is that overvolting for extended periods of time can change the doping in the silicon, and the CPU may not work reliably at stock voltage.

They are "tolerant" of +5V. Which means they will not instantly die. But could impact if continuously exposed.

You also have to handle the extra heat dissipation properly.

You need to check the "absolute max" DC ratings for the exact CPU you're testing.

One DX4 manual I'm looking at says 4.6V max. With "Extended exposure to the Maximum Ratings may affect device reliability."

MikeSG wrote on 2024-03-18, 09:32:

You need to check the "absolute max" DC ratings for the exact CPU you're testing.

One DX4 manual I'm looking at says 4.6V max. With "Extended exposure to the Maximum Ratings may affect device reliability."

as others have also written, the point I would like to discuss is "reliability". These computers would not be used as in the mid 90s turned on for 12 hours 4 days a week, but a few hours a week... I would not want reliability to have been exchanged for destruction within a few days of use over the years.

Mine stayed on back the day, day and night. Reliable back then, reliable now.

st31276a wrote on 2024-03-18, 11:26:

Mine stayed on back the day, day and night. Reliable back then, reliable now.

at 5v?

If it dies, presuming cooling is more than adequate, it's probably not technically the running at 5V that killed it, it was the using up of the "buffer" between what won't kill it and what will. Like a 1-2Volt spike over Vcc.

Therefore, take only one deliberate risk. You can run open air no shielding, you can run a questionable PSU, you can run your 3.3V CPU at 5V, but do all three together and it opens a portal for the fuckup fairy. Ensure you have a stable PSU and it's not putting out 5.5V on the 5V line, some people would call this "stout" I would call it risky if you want to run the CPU directly off it. Tune it down under 5V, I would be happy getting it down to 4.5. Then case this frankenstein, because you don't want any induced spikes pissing in your cheerios. Maybe use it on a surge protected outlet for extra peace of mind.

Edit: in this scenario, other things are risks that you might not count. Like hooking up a printer, having serial port connections to modems or other machines, having a cheapie USB card and connecting stuff to that... anything that's using a cable that acts as an antenna or pickup for induced current, so if you MUST do that, use good quality cabling, interface cards, and nothing else flaky, keep cables as short as you can.

It's also possible to use DOSIDLE (DOS) and WQHLT (WfW *.VXD driver) and have the CPU sleep while doing nothing.
It's meant to save power, but it also prevents the CPU from heating up (DOS era software uses quite some amount of IDLE loops).

Back in the day™, people would have used POWER.EXE or something similar.
The HLT instruction works with late 486 CPUs, APM BIOS or SMBIOS might be available on 386SL and other CPUs.

I had a DX4 100 for a while at 5v with the typical use of testing and retrogaming, the last time I used it I changed the voltage and the minimum it went to was 2.7v, so I would say that it is not easy for 3,3v 486 to burn at 5v as long as it has good cooling. I have a heatsink and fan on it.

Jo22 wrote on 2024-03-18, 14:08:

It's also possible to use DOSIDLE (DOS) and WQHLT (WfW *.VXD driver) and have the CPU sleep while doing nothing. It's meant to s […]
Show full quote

It's also possible to use DOSIDLE (DOS) and WQHLT (WfW *.VXD driver) and have the CPU sleep while doing nothing.
It's meant to save power, but it also prevents the CPU from heating up (DOS era software uses quite some amount of IDLE loops).

Back in the day™, people would have used POWER.EXE or something similar.
The HLT instruction works with late 486 CPUs, APM BIOS or SMBIOS might be available on 386SL and other CPUs.

The HLT instruction is present on any processor since the 8086, not just on "late 486 CPUs". Executing HLT likely will save power on all CMOS CPUs, as it's mostly the transitions that use power, and while the execution is halted, most stuff except the clocking is static. Using the APM "CPU idle" call in addition or instead of running HLT might be a good idea: For example, I had a Cx486SLC2 based laptop (don't get me started on that processor, though), that went into "non-turbo" mode whenever CPU idle was invoked.

mkarcher wrote on 2024-03-18, 17:38:

Jo22 wrote on 2024-03-18, 14:08:

It's also possible to use DOSIDLE (DOS) and WQHLT (WfW *.VXD driver) and have the CPU sleep while doing nothing. It's meant to s […]
Show full quote

It's also possible to use DOSIDLE (DOS) and WQHLT (WfW *.VXD driver) and have the CPU sleep while doing nothing.
It's meant to save power, but it also prevents the CPU from heating up (DOS era software uses quite some amount of IDLE loops).

Back in the day™, people would have used POWER.EXE or something similar.
The HLT instruction works with late 486 CPUs, APM BIOS or SMBIOS might be available on 386SL and other CPUs.

The HLT instruction is present on any processor since the 8086, not just on "late 486 CPUs". Executing HLT likely will save power on all CMOS CPUs, as it's mostly the transitions that use power, and while the execution is halted, most stuff except the clocking is static. Using the APM "CPU idle" call in addition or instead of running HLT might be a good idea: For example, I had a Cx486SLC2 based laptop (don't get me started on that processor, though), that went into "non-turbo" mode whenever CPU idle was invoked.

I see, thanks for the background information.

"Zwar umfasst der Befehlssatz aller x86-Prozessoren, beginnend mit dem Intel 8086, den Halt-Befehl.
Allerdings wurde er zunächst nicht zur Reduktion des Stromverbrauchs konzipiert. Dies änderte sich erst 1994 mit dem Intel DX4."
https://de.wikipedia.org/wiki/HLT_(Masc ... Geschichte

That's why I was thinking of late 486, I guess. . 😅
The programs I was using made use of this, I didn't know HLT was a thing before

Edit: DOSIDLE seems to support APM as an option, too.
The WQHLT driver doesn't, by contrast. But here, Windows 3.x has its own APM driver, anyway.
There's an entry in Windows Setup, I think. Somewhere in the PC type setting.