Voltage and frequency have a big say in power draw / heat / noise.

Modern CPUs don't need to be under-clocked. They all lower the voltage and reduce the multiplier.

Depending on where you live / weather you also need to consider the environment.

CPU life should be fine as long as it is not overclocked and not overheated.

I don't think underclocking older hardware will really help prolong its life. If anything undervolting the cpu to tested stability will be the only benefit against electromigration and deterioration.

As for your K6-2. If running it on a older OS (dos...Win95/98/ME) all never really issues good HALT commands (basically a pause command by system idle to help cooling). In later days there were later alot of programs that assisted in advanced features to cool K7's/P3's down. Like CPUCool, Rain, Waterfall. I'm not really sure if any of those helped K6's down. Also note when looking at temperatures in the BIOS, there are NO Halt commands so cpu runs full cycles. This is why temps always look higher in the BIOS 40+c usually.

Pfew, lots of questions, I will try to answer all of them as best as I can 😉

Malik wrote:

Will the heat be reduced if "underclocked"?

Definitely yes, the amount of heat produced by an underclocked CPU is always less

Malik wrote:

Can the CPU life be "prolonged" if it is underclocked?

In theory yes, but the life of a CPU is determined by many factors.

The reason why CPU's can die a slow death is when they are ran at higher temperatures. As already has been pointed out, this is caused by electromigration.

Malik wrote:

Are the results of CPU life inversely proportional when overclocking and underclocking are compared head to head?

To give a good answer to this we'd need to take into account what causes electromigration.
Wiki has an excellent article about electromigration.
In short, CPU's will die faster when the voltage they are run at gets past a certain point. This point is determined (for a very large part) by the CPU's manufacturing process (the nanometer thingy). Their deaths are further accelerated by the temperature the CPU is being run at and this is exactly why good cooling is so important when overclocking.

Malik wrote:

People say that the cpu life expectancy is reduced when overclocked. But by how much? (let's keep the heatsink-fan temperature control a constant, so that we can find out the effects of extra heat on the cpu).

And so, when we remove the constant, i.e. we enhance the cpu cooling to keep the temperature down and equal to that of stock speed level, will the life expectancy of an overclocked cpu be increased to that of a stock, non-overclocked running cpu?

There isn't really a constant. It's kinda like an inverse "L" shaped graph.
Suppose we start running a CPU at very low voltage and clock speed. If we start increasing it's voltage and clock speed while keeping temperatures constant (for instance by throwing a HUGE big-azz cooler onto it), life expectancy decreases by very little.
But at some point (the 'bend' of the "L") life expectancy will start decreasing by a lot suddenly.

Malik wrote:

My Super Socket 7 Epox MVP3C2 runs a K6-II 550 MHz cpu underclocked at 366MHz (66MHz x 5.5). In BIOS, if the sensors are to be believed, there's no difference (substantial at least) when running the cpu at stock speed and when it's underclocked. The temp lingers around 38 - 42 degrees celcius.
I'm using a cooler master Socket 370's heatsink-fan combo.

Ok, suppose your readings are all correct, then I can assume 2 facts from your info:
1)Underclocking your CPU makes the CPU temperature drop by basically nothing
2)You're using a relatively HUGE big-azz cooler on your CPU.

This phenomenon can only be caused when your ambient temperature (the temperature within your case) is already very close to the CPU temperature. I'll bet that when you, for instance, put a good fan case inside your rig, temperatures will go down significantly 😉

If you're measuring from outside the case and temps keep being as high as you state while the ambient temps are much much lower, then the temp sensor must simply be wrong.

Malik wrote:

And finally has anyone made a temperature comparison - when either the multiplier is increased or when the FSB is increased - are there any differences in heat production between these two?

Thanks for your information.

I've done some unofficial testing with this by running several CPU's as various clock speeds and watching the temps from within the BIOS for like half an hour.
I do this with the motherboard outside a case every time to check if I installed the HSF properly (and to check if the cooler I picked is able to put up with the task of cooling the CPU). Sometimes I simply try out a cooler without really knowing for sure if it's adequate and in such cases I'll have it run for like half an hour and monitor the temps from within the BIOS to see how long temps keep increasing

And one final hint:
You can mount a big HUGE touch-azz heatpipe cooler onto an underclocked DX4, but if your case temperature is like 60c, theres no way the air cooler can cool the CPU below that 60c.
This is why good case cooling is so important (and also why AT cases are so crappy compared to ATX cases when it comes to cooling).

If you have any further questions, feel free to ask 😉

Thanks a lot guys!

And Tetrium, thanks a lot for the exhaustive, patient explanations!

Yep, the AT casing was never designed with proper ventilation in consideration, mostly due to the associated cpus and other components available at that time, which do not require aggressive cooling.

I added some fans inside my system. I'll post some pictures tomorrow.

Malik wrote:

And finally has anyone made a temperature comparison - when either the multiplier is increased or when the FSB is increased - are there any differences in heat production between these two?

When I run a P-III coppermine at 150MHz/50FSB with a big heatsink (no fan) it stays cool, hardly above room temp. When I raise the FSB to 133MHz, the CPU goes to 400, then both the CPU and the motherboard chipset (northbridge) get warm.

When one uses throttle to slowdown the CPU, then the CPU will run much cooler, depending on the amount of slowdown.

Malik wrote:

Thanks a lot guys! […]
Show full quote

Thanks a lot guys!

And Tetrium, thanks a lot for the exhaustive, patient explanations!

Yep, the AT casing was never designed with proper ventilation in consideration, mostly due to the associated cpus and other components available at that time, which do not require aggressive cooling.

I added some fans inside my system. I'll post some pictures tomorrow.

Np mate 😉

I've got 1 thing to add though, which actually does work like a formula.
You can set a CPU at a certain frequency and you can run it at some given voltage.
If you increase it's frequency by 50% (or by a factor of 1.5), it will also increase heat production by 50%.
But if you were to increase voltage by 50% (btw, plz don't do it! 😜) then heat production is increased by 1.5 square!
This means that if you want to run your CPU as cool as possible, it may be more rewarding to find out at what lowest voltage a given chip remains stable.

I'm currently considering a Slot A project. I've been seeing a lot of people saying that they can get those to overclock like mad. One guy says he had his 500mhz Athlon running at 900mhz on water. Those posts are all ancient, but I'm still curious to see for myself how far they can be pushed.

On topic, my guess is that any life extension from underclocking is going to be far less than any life shortening from overclocking ie: underclocking a CPU by 100mhz won't save as much as life as overclocking it by the same amount takes away.

Also, if you're going to underclock a CPU, why not just get one that is rated for that speed in the first place?

sliderider wrote:

Also, if you're going to underclock a CPU, why not just get one that is rated for that speed in the first place?

Because it's easier to get a faster CPU and underclock it then it is to buy a crappy CPU and overclock it

Tetrium wrote:

Np mate ;) […]
Show full quote

Malik wrote:

Thanks a lot guys! […]
Show full quote

Thanks a lot guys!

And Tetrium, thanks a lot for the exhaustive, patient explanations!

Yep, the AT casing was never designed with proper ventilation in consideration, mostly due to the associated cpus and other components available at that time, which do not require aggressive cooling.

I added some fans inside my system. I'll post some pictures tomorrow.

Np mate 😉

I've got 1 thing to add though, which actually does work like a formula.
You can set a CPU at a certain frequency and you can run it at some given voltage.
If you increase it's frequency by 50% (or by a factor of 1.5), it will also increase heat production by 50%.
But if you were to increase voltage by 50% (btw, plz don't do it! 😜) then heat production is increased by 1.5 square!
This means that if you want to run your CPU as cool as possible, it may be more rewarding to find out at what lowest voltage a given chip remains stable.

Just to add to your post, when one lowers the voltage the CPU runs at, the amperage that said CPU will draw from the VRM's on the motherboard will go up if you drop the voltage. I know some boards are fussy about how much amperage they can supply on the VRM's too, so it's a consideration.

I think the whole overclocking thing is a myth made by manufacturers. 😁
I've overclocked every CPU I have ever owned (well except a couple AMDs that refused to budge) and they never died or slowed down. How long does it actually take to see the effects of overclocking, 20 years at full load for 24 hrs a day?

I have a Cyrix PR200 that went faulty after about 5 years of use (from friends rig I fixed). It was never overclocked or ran in a harsh environment. I have a Celeron 300A @ 450 and a Celeron 533A @ 800 that ran for just as long, and still is operational today.

bushwack wrote:

I think the whole overclocking thing is a myth made by manufacturers. 😁
I've overclocked every CPU I have ever owned (well except a couple AMDs that refused to budge) and they never died or slowed down. How long does it actually take to see the effects of overclocking, 20 years at full load for 24 hrs a day?

I have a Cyrix PR200 that went faulty after about 5 years of use (from friends rig I fixed). It was never overclocked or ran in a harsh environment. I have a Celeron 300A @ 450 and a Celeron 533A @ 800 that ran for just as long, and still is operational today.

It's not a myth, it's a very real problem if you jack the voltage up too high, but you can help offset this by keeping things cool to some extent.

It can kill motherboards though. I hosed a CUSL2 running a PIII 800EB @980mhz (jeez that chip didn't want to go very far). I tried the chip a few years later in another board and the cpu was fine, but obviously the board was dead in the water.

SavantStrike wrote:

bushwack wrote:

I think the whole overclocking thing is a myth made by manufacturers. 😁
I've overclocked every CPU I have ever owned (well except a couple AMDs that refused to budge) and they never died or slowed down. How long does it actually take to see the effects of overclocking, 20 years at full load for 24 hrs a day?

I have a Cyrix PR200 that went faulty after about 5 years of use (from friends rig I fixed). It was never overclocked or ran in a harsh environment. I have a Celeron 300A @ 450 and a Celeron 533A @ 800 that ran for just as long, and still is operational today.

It's not a myth, it's a very real problem if you jack the voltage up too high, but you can help offset this by keeping things cool to some extent.

It can kill motherboards though. I hosed a CUSL2 running a PIII 800EB @980mhz (jeez that chip didn't want to go very far). I tried the chip a few years later in another board and the cpu was fine, but obviously the board was dead in the water.

Did overclocking actually kill it though, for a fact? I've had a Jetway mobo die on me just 2 months after I bought it, no overclocking involved what so ever. Damn thing just wasn't capable. 😜

SavantStrike wrote:

Just to add to your post, when one lowers the voltage the CPU runs at, the amperage that said CPU will draw from the VRM's on the motherboard will go up if you drop the voltage. I know some boards are fussy about how much amperage they can supply on the VRM's too, so it's a consideration.

Thanks for bringing this up, I forgot to do so myself 😉
It's indeed a consideration, especially with some particular motherboards (the TX boards come to mind).

bushwack wrote:

Did overclocking actually kill it though, for a fact? I've had a Jetway mobo die on me just 2 months after I bought it, no overclocking involved what so ever. Damn thing just wasn't capable. 😜

Jetway dead after 2 months, that's why I and a friend of mine (who is also computer savvy) call them "Throwaway" 😜

As I understood it - and happy to be corrected - cpu manufacture involves making a vast number of cpus at the same time or from the same process, and then assigning speeds to them according to how well they came out, presumably through some kind of testing. I can only think they run some sort of fairly quick stability test and I'm guessing they mainly look for how close to the steep part of that L-curve it's safe to take a cpu.

Ontop of that there's the various badging strategies that we hear about from time to time where faster cpu's are badged as slower ones etc.

All of which means that there's going to be a lot of "noise in the data" when you start to look at whether - in practice - overclocking shortens the lifespan of a cpu. I infer that:

- a lot of heavy overclocking almost certainly will reduce lifespan, especially once you get your cpu into unstable regions for significant periods.

- it's going to be hard to distinguish the effects of low-level, stable overclocking because setting the cpu speed in the first place isn't an exact science.

So unless you overstress your cpu, you might as well not worry about it any more than you worry about usage reducing the lifespan of a non-overclocked cpu.

ratfink wrote:

As I understood it - and happy to be corrected - cpu manufacture involves making a vast number of cpus at the same time or from […]
Show full quote

As I understood it - and happy to be corrected - cpu manufacture involves making a vast number of cpus at the same time or from the same process, and then assigning speeds to them according to how well they came out, presumably through some kind of testing. I can only think they run some sort of fairly quick stability test and I'm guessing they mainly look for how close to the steep part of that L-curve it's safe to take a cpu.

Ontop of that there's the various badging strategies that we hear about from time to time where faster cpu's are badged as slower ones etc.

All of which means that there's going to be a lot of "noise in the data" when you start to look at whether - in practice - overclocking shortens the lifespan of a cpu. I infer that:

- a lot of heavy overclocking almost certainly will reduce lifespan, especially once you get your cpu into unstable regions for significant periods.

- it's going to be hard to distinguish the effects of low-level, stable overclocking because setting the cpu speed in the first place isn't an exact science.

So unless you overstress your cpu, you might as well not worry about it any more than you worry about usage reducing the lifespan of a non-overclocked cpu.

That's basically correct 😉

CPU dies are made from silicon wafers from which the dies are cut out.
Depending on the quality of the wafer and whether the die is from the center or the edge of the wafer (the nearer to the center, the better the die will scale) are the 2 main reasons in deciding how well any particular die will scale 😉

Tetrium wrote:

sliderider wrote:

Also, if you're going to underclock a CPU, why not just get one that is rated for that speed in the first place?

Because it's easier to get a faster CPU and underclock it then it is to buy a crappy CPU and overclock it

So if you need a CPU that runs at 500mhz, you'd happily spend more money for a 1ghz CPU and underclock it? That's silly. Whether it's rated for 500mhz or 1ghz won't matter if you're going to underclock it to 500mhz anyway. The performance will be identical.

sliderider wrote:

So if you need a CPU that runs at 500mhz, you'd happily spend more money for a 1ghz CPU and underclock it? That's silly. Whether it's rated for 500mhz or 1ghz won't matter if you're going to underclock it to 500mhz anyway. The performance will be identical.

Yes, but you may feel somewhat snuggly and satisfied that you can "increase" the speed anytime you want to make it run at it's full potential.

And it's like uh..."you-have-the-more-powerful-stuff-to-run-a-simpler-system" kind of feeling...something like you have a killer beast, that is under your control, and despite the fact that it can ram through anything, you jyst keep that power in your leash....and you can always release it when required...kind of feeling...hard to describe...

Am I even making any sense here?... 🤣

Alright, the pictures :

The system - my trusty P133 casing with add-on stickers :

Ever seen an AT casing with USB? 😉 :

(I removed the bracket from a USB connector and fixed it to the LPT1/25-pin connector opening.)

The ventilation :

There is a slot cooler behind that huge fan - the only outlet of air from within the case.

The temp. with the case opened at 20mins runtime :

And the temperature after closing the case, 15mins later :

EDIT : System contains :

EPOX MVP3C2 SS7 Motherboard
256MB PC-100 SDRAM (128MBx2)
AMD K6-2 550 @ 366MHz (66MHz FSB x 5.5 Multiplier)

This board does NOT have a 2.3v setting (as printed on the cpu heatsink) for the cpu. Only 2.0, 2.2 and 2.4v and above. So I set it to the lower 2.2v. It's stable and I'm happy that it runs on a lower voltage, which I prefer.

Other contents :

Diamond Stealth 3D 3000 Pro 4MB
Diamond Monster 3D
SB AWE64 Gold
SB Live! 5.1
Music Quest PC Midi Card ---> CM-500.

sliderider wrote:

Tetrium wrote:

sliderider wrote:

Also, if you're going to underclock a CPU, why not just get one that is rated for that speed in the first place?

Because it's easier to get a faster CPU and underclock it then it is to buy a crappy CPU and overclock it

So if you need a CPU that runs at 500mhz, you'd happily spend more money for a 1ghz CPU and underclock it? That's silly. Whether it's rated for 500mhz or 1ghz won't matter if you're going to underclock it to 500mhz anyway. The performance will be identical.

The thing with these retro CPU's is that faster doesn't always equal more expensive.
If I can buy a P3-1000 for only $1, then why would I want to buy a 600Mhz Celeron for $1 and waste extra time overclocking it?
The 500Mhz and 1Ghz examples you speak of in your example are often not more expensive the higher the clockfrequency is.

And theres another reason for getting a higher rated CPU, it's generally of better quality silicon then a lower clocked variant, thus can often be run at cooler temperatures then a similar CPU running at it's rated speed, another small advantage.