Those CPUs probably ran a decade without a heatsink/fan, so adding one now won't extend their remaining lifetime that much. CPUs get hotter with age and die eventually because of that anyway, but there's no way to prevent this, as it's part of CMOS degradation (which you can not stop). You may add a heatsink if you want to feel being on the safe side, but don't expect that this will make the CPU last for another decade. It simply won't make a difference.

fitzpatr wrote:

dionb wrote:

... Time for me to find a heatsink - and proof that this is not just an academic discussion.

It will be proof when that fixes your problem.

Very true. Will jury-rig a temporary solution this evening. If it does, I can look for something better than double-sided sticky-tape and a chipset heatsink on it.

derSammler wrote:

but there's no way to prevent this, as it's part of CMOS degradation (which you can not stop).

The thing is that the amount of degradation is very much dependent on the fabrication process. The reason why NASA stuck to the 8088 CPUs is that CMOS degradation is not that well-understood, and every type of CPU has to be studied individually for degradation. With the 8088 they knew from experience that these CPUs were extremely unlikely to fail. So far the 8088s can live for 40 years without any heatsink and fan, and nobody knows just how much life they have in them.
The 286, 386 and 486 CPUs from Intel are apparently also very unlikely to fail, since there are still so many working systems around, and not many reports of dead CPUs at all.

derSammler wrote:

Those CPUs probably ran a decade without a heatsink/fan, so adding one now won't extend their remaining lifetime that much. CPUs get hotter with age and die eventually because of that anyway, but there's no way to prevent this, as it's part of CMOS degradation (which you can not stop). You may add a heatsink if you want to feel being on the safe side, but don't expect that this will make the CPU last for another decade. It simply won't make a difference.

I don't know how likely it is that they ran a decade without a heatsink. There is always a chance that the one you have only ran for 3 years before it was upgraded and stored. There is really no harm in cooling it a bit just in case it would make a difference.

Scali wrote:

Not really, as I say, it's all about binning. AMD and other clones basically sold under-specced CPUs, think of them as factory-o […]
Show full quote

Baoran wrote:

There might be a difference in if cooling is required between if you want the cpu to last 5-10 years or 30+ years.

Not really, as I say, it's all about binning.
AMD and other clones basically sold under-specced CPUs, think of them as factory-overclocked. They simply CAN'T run at their graded speed without a heatsink and fan.
But Intel's CPUs were of better quality. They wouldn't label a 486 as a DX33 unless its die was pure enough to run at 33 MHz reliably without a heatsink and fan. They'd sell them as DX25 if they didn't make it.
What AMD and others effectively did, was take these 'DX25' CPUs, put a heatsink and fan on them, and run them on 33 MHz anyway.
Since the Intels never needed a heatsink and fan anyway, adding one doesn't have much use either. They just don't get very hot to begin with.
(of course there are also fake CPUs out there, which may have been re-labeled as a different brand or a higher speed grade... so your Intel DX33 may not actually be an Intel DX33).

All above 386 need at least heatsink, no matter who what is saying... And above 486 there is need of fan as well. Now on my 486 build i will put heatsink and fan as well, i want thing to be cool, not toasted..
That if intel processors or others does not need cooling, that does not mean they need to be left like that, that was back on 386, where all processors are exposed, but on 486 is not the same. At least i care for the parts, so cool them, as there one thing with computer parts the cooler = the better as the HDD are the only exception, as they don`t like too much cooling..

Radical Vision wrote:

All above 386 need at least heatsink, no matter who what is saying...

What are you basing this on exactly?
Take any early 486, of 33 MHz or lower, take a high-end brand-name machine if you like (IBM, Dell, Compaq, HP...). Open them up, and you will find no heatsink in most of them.
Why not? Because it wasn't part of the design specification from Intel.

See here, for example: http://www.totalgeekdom.com/?p=356

I am almost certain that the Zenith 486/66 I had back then did not have a heatsink...
Nowdays for piece of mind I but heatsinks even on 286 cpus as they get too hot for my taste. Is cooling needed? Doubt it since those machines work after decades but nowdays a piece of thermal tape and a heatsink are very cheap and easily found so why not? 🤣 🤣 🤣

I just think design specifications don't matter anymore because they don't design them to be used for decades when they design them. I am sure 33Mhz one would not need heatsink or fan to last as long as they thought it would be used for back in 90s.

Since when the design specification from the manufacturer is any important ? No matter they did not put cooling, i will put, as i don`t want to worry about the temperatures, no matter how tough the old hardware is.
For example 3Dfx cards early ones does not have even heatsink, but they are stoves, not to speak about the Voodoo 3 and Voodoo 5 they are all heaters, but as you state the manufacturer did engineered the whole thing, so we don`t need to check the temperatures, we will be just like the normal user that does not have a clue what is it is in the PC case and we will play it simple. For example i check on all my components that im using the temperature, and if is warm under stress/load it is ok for me, but if some parts are heaters under idle, then there is big problem, and it need some upgrade...

Also all manufacturers are cheaping out even on way more important things like japan caps, VRM elements, transistors, chokes and other, and you are thinking they will not cheap on the cooling really, as they cheap on that alll the time they can, to reduce cost and saving money.... That does not mean, the computer some OEM did make will fail, no it will work, but instead of working one more cool environment it will work in way wors one, that will reduce lifespan, will even cause some day the components just to die from heating up....

In the end i will cool my stuff, while you can use yours as heaters, after all it is winter in EU, so maybe you don`t need cooling, no matter the temperatures...

Baoran wrote:

I just think design specifications don't matter anymore because they don't design them to be used for decades when they design them.

Again, what are you basing this on? These CPUs were aimed at a wide range of markets, including usage in embedded systems, military installations and whatnot. Many of which do require CPUs to last for decades.
We are talking Intel here, not some cheap random Chinese clone builder, that wants to sell the cheapest chips possible, at any cost.
486 CPUs were originally designed as extremely high-end server/workstation chips, and were incredibly expensive.

Scali wrote:

Again, what are you basing this on? These CPUs were aimed at a wide range of markets, including usage in embedded systems, milit […]
Show full quote

Baoran wrote:

I just think design specifications don't matter anymore because they don't design them to be used for decades when they design them.

Again, what are you basing this on? These CPUs were aimed at a wide range of markets, including usage in embedded systems, military installations and whatnot. Many of which do require CPUs to last for decades.
We are talking Intel here, not some cheap random Chinese clone builder, that wants to sell the cheapest chips possible, at any cost.
486 CPUs were originally designed as extremely high-end server/workstation chips, and were incredibly expensive.

Intel and is intel the understanding of quality really................
Just look the recent intel products, they are total trash, when it comes to build quality, and they are working close to the maximum temperature spec and other. The thermal goo/toothpaste they are using instead of soldering, also their box coolers are not very good, not to speak about intel junk thin CPU waffle on skylake, when they tend to bend and even brake........ No matter back in the 486 era intel did have way superior quality, not gypsy one like now, and still they are tend to cheap out, to reduce costs, no matter they are making tons of money...
Today Intel have also super expensive CPUs like i9-7980XE 18/36 cores, it cost a lot, and it is using thermal goo/toothpaste, so for what quality we are talking...

Scali wrote:

Again, what are you basing this on? These CPUs were aimed at a wide range of markets, including usage in embedded systems, military installations and whatnot. Many of which do require CPUs to last for decades.
We are talking Intel here, not some cheap random Chinese clone builder, that wants to sell the cheapest chips possible, at any cost.
486 CPUs were originally designed as extremely high-end server/workstation chips, and were incredibly expensive.

MIlitary grade 486 cpus were different than the ones sold to normal customers. Like these http://www.nj7p.org/Manuals/PDFs/Intel/271329-003.PDF

Here is a very non scinetific test. It's not like this thread is full of science so I guess it fits right in.

This is a 486 DX50 running at its rated speed looping PCP bench VGA mode. I guess PCP bench isn't a great CPU benchmark but the avarage game isn't either and PCP bench was the best (also only) loopable bench utility I happend to have on this DOS 8 start disk. As you notice the motherboard isn't in a case but on the other hand at least OEM systems tend to be designed so they have some airflow over the CPU area. On my "test bench" there is no airflow what so ever.

After looping PCP bench for about 45 minuts and seeing less than 0.5 degrees Celsius rise in temperature the last 15 minuts I took this picture. The CPU casing is probably a few degrees warmer than shown as only the tip of the probe is touching the CPU and the core is probably warmer than the CPU casing.

Even with the unscientific method in mind I would not be worried over running this Intel DX50 at it's rated speed bare in a case with at least some movement of air. This is coming from the guy that called out a former Vogoner (now living the "high life" on Youtube getting ad revenue, free hardware, nerd groupies and such) for running a DX4 bare in his Youtube show.

For anyone not understanding Celsius... the CPU is luk warm and using my finger as a probe I would guess the real temperature of the CPU casing is at most 5 degrees C over what the probe is reading.

An AMD DX33/40 is (like Scali already mentioned) probably running alot hotter than the Intel DX33 and even Intels DX33 could be running a bit hotter than this DX50 as it's made with an older process. I will test both an AMD and an Intel DX33 clocked at 40 MHz and report my findings.

Not sure how that result can be so different. Basically my intel 486dx 33Mhz one would definitely burn my skin if I would hold my finger there and cpu load doesn't seem to change that much because it becomes that hot in just dos without running anything else.

Baoran wrote:

Not sure how that result can be so different. Basically my intel 486dx 33Mhz one would definitely burn my skin if I would hold my finger there and cpu load doesn't seem to change that much because it becomes that hot in just dos without running anything else.

That sounds wrong. Are you sure you don't have a 3.3v CPU running at 5v?

As far as I know there is no 3.3V version of Intel 486dx-33

Because of it being so hot it could burn my finger, I did put that pentium heatsink + fan there just in case.

Baoran wrote:

As far as I know there is no 3.3V version of Intel 486dx-33

486-33.jpg

It's 5V alright, and it's not an early sample (1992 week 52).

The test with the Intel DX33 isn't fully done yet but I can already tell that the one I'm testing isn't running much hotter than the DX50 did.

I will test a few more Intel DX33 CPUs to see if there is much difference between CPUs

Baoran wrote:

As far as I know there is no 3.3V version of Intel 486dx-33

There are 486DX-S CPUs at 33 MHz, which are 3.3v.

Scali wrote:

Baoran wrote:

As far as I know there is no 3.3V version of Intel 486dx-33

There are 486DX-S CPUs at 33 MHz, which are 3.3v.

Aren't those quite rare? I don't think I have even ever seen a picture of one of those.

Some Intel DX33 CPUs seem to run hotter than others, one of the two extra I tested even locked up after 10-15 minuts or so at 40 MHz looping PCP bench.

The AMD DX40 (could not find an AMD DX33 eventhough I should have a few) was running like champ at 40 Mhz without heatsink looping PCP bench. It did get pretty hot! After I ended the test 20 minuts or so in the probe showed about 40 degrees but when pressing my thumb aginst the middle of the CPU I would say 50 degrees C is a better estimate.