How low can ya go ?

BrainEater

[H]ard|Gawd
Joined
Jul 21, 2004
Messages
1,189
I'm sitting here thinking about full system cooling.

Regardless of method , whether gas or liquid cooled , the same question keeps popping up.How cold can a mobo/cpu/ram/video card go ?....My initial research is showing mixed results below -40c......Now, this is 35 ish degrees below spec for most parts .....

I'd like to open a discussion on this , but we'll need to set some initial conditions.

Condition 1) water and/or condensation will not be an issue.Consider it a closed system.It's obviously not water based.

Condition 2) The cooling to these low temps will be accomplished slowly to minimize thermal shock.

Condition 3a) While overclocking is implied, it should be stable without.

Condition 3b) Lets assume the high heat loads from cpu/gpu are dissaped to same levels as the rest of it ; I.E the whole system is at -40 .
---------------------------------

Idea's /comments ? :D
 
I cool my CPUs to about -53'C with no issues thus far. I have done this on several AMD CPU's, (opt150,4800+,6000+). I have cooled my system memory to 3'C, and I have only cooled my GPU's to 0'C.

hope this helps.
 
Any issues ?

The current extremes I've see indicate the cpu and gpu will tolerate liquid nitrogen @ -196c.

The other hotter components will do -78c ish (dry ice)

The components I see failing are a)Battery b) Capacitors and then c) any physical connections i.e. connectors , solder joints, etc.......

The battery can be moved out.The rest is an issue.........

I'm in the process of setting up some tests.

:D
 
Hmmm... Quite interesting. I would be wary of the Plastic parts becoming brittle and possibly cracking due to thermal contraction...

Obviously the hotter components can handle temperatures in the far negatives (LN2) but I strongly agree with solder joints and (especially) liquid caps popping. The CMOS battery could be replaced with power from the 3.3v line, or just externalize it. Caps could be externalized as well... Fairly easily if they're the electrolytic kind too.

I wonder if you could avoid cooling the components directly, by attaching the board to an anodized aluminum plate covered in cheap thermally conductive padding... Just cool the coldside of the plate, and have it radiate (can cold radiate lol?) into the board.
 
It's fortunate motherboard builders are moving away from electrolytics...From what I'm seeing conductive polymer capacitors wont suffer the same issues.

There's another major problem I realized : The crystal oscillators.They are very definitely temperature sensitive.They are usually spec'd @ 25c....The colder they get the slower they'll oscillate.

-----------

I think cooling the back is probably the same as direct cooling when we're talking <-40c...

-------------

I can get dry ice from the place I get my welding supplies , so I intend to do some basic tests with dry ice / alcohol . That'll get down to the -60 to -70 c range.

Some tests I intend to try ;

-physical connector integrity ; If the pci card comes loose in the slot at -70c , it probably won't work.

-What does happen to electrolytic caps .

-I've got a scope , so i'll check the oscillator issue.


Anyone think of any other tests ?
 
Ok......My victim is ready !!

---------------

I must preface this by saying this is very dangerous.

Do not try this type of experiment unless you know what you are doing.

I have 20+ years of 'work environment safety' experience.This shit can explode if you are sloppy.

-------------------

This is an old rig I salvaged outta my parts bins.
It's an asus p5a-b motherboard , an AMD k6-2 450 , 64 Mb ram and a 3dfx voodoo3000 AGP.

It's fully submersed in 99% Isopropanol , and running just fine ;


liquid1.jpg




----------------

I'm ready for the dry ice I think...........Mebbie tomorrow. :cool:


Last chance for test idea's.

:D
 
That's awesome. Nothing pleases me more than the sight of a submerged computer.

Edit: Are you using the Midel, again?
 
Nothing pleases me more than the sight of a submerged computer.

LOL :p

well indeed, once the plastic polymers are below the vitreous transition temperature they will become brittle,


The current extremes I've see indicate the cpu and gpu will tolerate liquid nitrogen @ -196c.

where do u get that? to actually corroborate this you'll have to run a liquid nitrogen refrigerated PC 24/7 for some time, but that would super expensive. I've seen some LN2 refrigerated CPUs that stop working right afterwards the LN2 experiment ended. After microscopic evaluation of the dead cpus, the cause of its failure was broken connections, due to thermal contraction and dilatation. You have to be careful when u shutdown your pc, the temperature has to raise slowly
 
:D

I get that from the many OC experiments that take place.LN2 OC runs are commonplace.

You are very correct.Stuff tends to shrink to fail when dealing with sub-cryo temps.

That's what this thread is about.

I'm testing the whole system to failure or -78c.......whichever comes first.

:D

-------------



neg Mark..........thats Isopropanol.
 
Careful with the splashing! That much alcohol...have towels on hand and ready.

In case you (or anyone else) are curious, t takes 30 minutes to an hour to get 198 proof (rubbing ;)) alcohol cold enough to start becoming slushy in parts. Afterward, you can reuse the isopropanol, just be careful bottling it up because there'll be a ton of CO2 that will continue to come out...for a few days.

Take more pics :cool:
 
oooooooo k.

Looks like I have an answer.-52c appears to be too much. :D :D


di2.jpg


Last shot before failure ;

di1.jpg



The temp sensors maxd' at -40c and -48c , so I have had to estimate the final temp based on it's cooling rate, which works out to about -52c or -53c.

---------------------------

I'm letting the computer and iso warm up to room temp slowly , and I'll try to ascertain what failed....should be interesting.

I didnt bother taking out the battery , might as well test it too......

I have some spare bits that'll fit this mobo , so if I can , i'll be repeating this experiment.

:D
 
The temp sensors maxd' at -40c and -48c , so I have had to estimate the final temp based on it's cooling rate, which works out to about -52c or -53c.

---------------------------

I'm letting the computer and iso warm up to room temp slowly , and I'll try to ascertain what failed....should be interesting.

I didnt bother taking out the battery , might as well test it too......

I have some spare bits that'll fit this mobo , so if I can , i'll be repeating this experiment.

:D
Awesome :D
 
Very interesting.

The machine still totally works , so whatever caused the failure wasnt permanent.

/me plans the next course of action.

:D

--------------------

From what I've been reading , it's probably failed by 'freeze-out' of the electrolytics or the battery , as was predicted.

As far as the other components go , there should be no issue getting to -78c.
Silicon doesnt fail till -40k.......
Entire computers have been submerged in ln2.....

guess I have some soldering to do ! :D
 
Ok.......Update and a question.

------------------

At this point I'm looking for free bits/ stuff at the scrapyard to build a 2 tank ,circulating coolant setup.I'm seriously going to try hard to make it a vacuum jacket system ( similar to dewars. )

The setup is going to be designed to allow fully cryogenic cooling ( -120 C and below ).

:D

-----------------

I have a question about RAM and cold tho.

I've seen the story about 'freezing ram to break encryption'.....

Is this something I should be worried about ? This is basically ram failure....

Also....same thing goes for bios ....


I'd hate to have to install RAM heaters to run liquid argon.... :p
 
I've seen the story about 'freezing ram to break encryption'.....

Is this something I should be worried about ? This is basically ram failure....

My understanding is this is done immedaltely AFTER power down in an attempt to slow down the normal loss of memory contents because the electrical signals that keep the memory cells "alive" has been removed. Memory cells are a lot like little capacitors in that they store a charge for a small amount of time (a logical 1) but then have to be refreshed to continue to hold the voltage charge representing the 1 (or zero). Without the recharge the 1's are lost. Supercooling the stick slows the loss of charge and allows the memory to be read. And frankly that article sets up an unlikely set of events. The user leaves the computer, unsecured with an encryption key still stored in memory and someone rushes in, disassembles the case, freezes the ram stick removes it, and reads it somehow most likely with a specialized piece of equipment. I am not sure what the supercooling does to the normal operation of the memory when the proper electrical signals are present. It is a concern but I dont think its the same concern.
 
BrainEater, Nice work man, keep the info.coming.

This thread is on the verge of becoming educational and I fear I just may learn something if I'm not careful :)
 
Thx ! :D

I appreciate the input everyone !


If anybody knows where I can find dented/damaged LN2 dewars , pls let me know.I'm going to my local metal scrap-yard twice a week , but no luck so far.....
 
Thx ! :D

I appreciate the input everyone !


If anybody knows where I can find dented/damaged LN2 dewars , pls let me know.I'm going to my local metal scrap-yard twice a week , but no luck so far.....

I would try a local college (preferably one with a decent sized chemistry department). There's probably a few old ones around that profs might not mind unloading.
 
Woot !


I have my next victim !!

I have replaced my Striker Extreme , with a Maximus Formula.

Rather than smash this 680i POS to sand , Let's see what it thinks about -80 c .

Hot chipset ? No longer. :p

I'll prolly chuck in this 1 gb of mushkin redline I have.......Get a cheep core 2.....

---------------

I've also got a new memory card for my camera , so I can shoot some short videos now .

:D
 
Back
Top