How true is it that low temperatures mean overvolting is safe?

[JonJ678]

Reasoning is fairly simple. Chipsets often need fairly high voltages to hold processors at high clock speeds. If kept cool, they need slightly less voltage for stability. However, if kept cool, can they also be run safely at voltages that would be unsafe if hotter?

I'm wondering if this idea justifies watercooling motherboard chipsets. Extending motherboard lifespan, and slightly improving overclock.



In a similar vein, but a different question, what on motherboards get warm/hot? Ignoring pci cards. So far Ive come up with
Processor
Ram
Northbridge (or whatever its been renamed to now)
Southbridge (ditto)
Mosfets

Have I missed anything? I'm thinking of running the motherboard in an airtight box, so would like water blocks on abolutely everything that generates heat. Otherwise it'll just get steadily hotter inside and it'll all end in tears. It seems likely that I'm employed at last, so can start making things soon. Free electricity & workshop => thermoelectrics => avoiding condensation using airtight box. So many projects planned

 

[ViRuS2k]

If a CPU or Chipset is COOL and you BUMP VOLTAGES.
and yet the chipset or cpu is still cool and within spec temps then bump again

why do you think you can select all those voltages in the motherboard bios
if it was unsafe they wouldnt put them there.

all you have to be carefull of is temps.

and when it comes to memory its advised to no overvolt even if the memory is cool as memory is very sensitive and can break quite easily if given to much volts.


UPDATE:

If you stick the motherboard in a airtight box, meaning no air at all, then you will be required to watercool everything

CPU/MOBO/GRAPHICS/

you dont have to watercool the mosfets LOL
but basicly everything else.

 

[PaulyD]

you dont have to watercool the mosfets LOL

Its the mosfets that gets hottest on i7 as its the voltage reg to the chip itself. The NB is now very cool on it.

However on 775 its reversed and the NB gets toasty with the overclocking as well as the chip

 

[Bubo]

In case you missed it, http://forums.overclockers.co.uk/showthread.php?t=18029742

The common things you've already mentioned obviously generate heat, but it also depends on your motherboard layout as there are always odd power transistors and chips which generate heat besides the chief culprits. On my motherboard there is clearly a patch next to my NB which generates quite a lot of heat, as well as a broad patch near the pci slots. So it really depends on areas like this on your own mobo as to what equilibrium temperature your sealed pc box would get to. However, in terms of guarding agaist true degredation affecting mobo life, then I would still think it is the main components which only need to be cooled in reality. Though to be honest I'm not sure what a totaly sealed box, with no airflow in/out, is going to gain.

 

[mmj_uk]

Keeping temps down will generally prolong the lifetime of a chip but too much excess voltage will have a damaging effect regardless of temperature.

 

[JonJ678]

Thank you. That's four solid replies already, and it's not even lunch time.

I'm not keen on overvolting ram, I think I've lost some ddr2 to this. Mosfets running hotter makes sense as the chip draws more power than the 775. Some boards have mosfets on both sides, which further encourages watercooling one side as it'll keep the board cooler, and this is the best the back chips can ask for.

IOn my motherboard there is clearly a patch next to my NB which generates quite a lot of heat, as well as a broad patch near the pci slots.

Though to be honest I'm not sure what a totaly sealed box, with no airflow in/out, is going to gain.

I definitely saw that thread, it's a large part of why I started this one. Thank you lots for the images. Sadly I don't have access to thermal imaging, so will have to go for informed guesswork. The patch near the northbridge worries me as I don't know what that is.

Totally sealed box with a small pot of dessicant means no condensation....

Keeping temps down will generally prolong the lifetime of a chip but too much excess voltage will have a damaging effect regardless of temperature.

I'm struggling a bit with the relationship between these. Will a given voltage cause the same damage at any temperature, or does keeping it cool reduce the damage done at a given voltage, so allowing more voltage?

Thanks again guys, i appreciate all the replies above

 

[Bubo]

Ah, you're thinking of a TEC/chilled water? I would still think you will need some active cooling to take out the ambient generated heat though as I imagine it will get pretty damn hot if it was totally sealed, maybe make one side of your sealed box as a water cooled panel? Water cooled lid would be best I guess as the heat would collect at the top of the box. Pheasible to make if its just two sheets of copper with convolutions.

 

[Cold_And_Miserable]

The safety of overvolting is not affected by temperature.
I suspect people have confused the dangers of overvolting with overheating as a result of overvolting.
Overvolting causes electromigration regardless of temperature and is far more destructive than overheating, at least as far as CPUs are concerned.
Capacitors will last longer at lower temperatures but these are mostly on the motherboard.

 

[JonJ678]

Ah, you're thinking of a TEC/chilled water?

Very much thinking of peltiers yeah, not given up on them yet. Construct a heat exchanger rather than strapping them to the chips directly, with many radiators cooling the hot loop. It will not be electrically efficient, and isn't exactly the standard first step into water cooling.

I fear I've come to the same conclusion, there needs to be something in the box to take heat out. I think I'm going to put a radiator inside the sealed box and pass the cold loop through this.


I'm confused by the common view of electron migration. It's electrons tunnelling to places you don't want them to be. So yes, voltage obviously increases the odds of this happening. But it's blatently proportional to current as well, and this is rarely mentioned.
I cannot see why it would be independent of temperature when everything else I know of that relies on this effect is not. Solid state diffusion is enormously dependent on temperature, and is very much quantum tunnelling in action. I need to find a materials scientist who overclocks computers, I'm sure theres at least on on here

Blacks equation is part of the answer, but not all

 

[R.C.Anderson]

Ok, well I have a quick read of the replies and didn't see this, but if I missed it and it's already been mentioned, forgive me

I've seen a few people who submerge their whole system in oil, and as it doesn't conduct, and is in direct contact with everything it will cool well. So rather than having an airtight container (which would take some doing) how about have it all submerged in oil? that way the whole motherboard is cooled, but you can still watercool the hottest components (CPU, Chipsets etc)

 

[Xelene]

High temperatures are a side effect of extra energy being pumped into the chip, you can mitigate the temperature with good cooling which will prevent thermal damage - but the energy is still there and it will agitate the electrons etc.

It's not a big deal if you are pumping power into a 10cm copper rod, but the circuits of a modern chip are tiny and it doesn't take much power to make atoms start jumping around - ruining the CPU.

Edit: I should add that cooling removes the energy (in the form of heat) from the chip.

 

[JonJ678]

Making the case airtight would be exactly as hard as making it oil tight, but if air leaks a bit it doesn't matter. Oil will seep along cable insulation and the like, generally a terrible idea. The pump itself stands a reasonable chance of being immersed in oil, but I don't think I'll do that with the motherboard. I still want to be able to take the lid off and work on it as normal, and I think chilled air would do a better job than trying to chill and recirculate oil. Thanks for the suggestion, I just don't think it offers any advantage over air and several disadvantages.

Thank you sldsmkd, but can you quantify this?