Are TECs even worth using?

[-Metachronos-]

From reading this and a few other forums, I've gotten the impression that TECs really don't cool that great. Can someone tell me otherwise?

 

[Stugots]

From reading this and a few other forums, I've gotten the impression that TECs really don't cool that great. Can someone tell me otherwise?

They can but I think its risky. You could have a 226W peltier right on the CPU and it will cool it damn good. But, if it fails you will almost surely burn up the CPU. I would rather use peltiers to cool water then pump that through a typical water cooling system. Temperatures won't get out of control as quickly that way.

Swiftech has a waterchiller that uses 2x226W peltier. The bad part is that they recommend that you use 2 seperate loops to cool the peltiers alone. So thats at least 3 watercooling loops running one system. Not to mention the power supplies you'd need. You'd probobly need at least 2 high capacity power supplies, one for the waterchiller and its associated pumps, fans, peltiers, and the other for the rest of the system and its water cooling parts. The only practical use for that I can think of is a overclocking work bench to test out CPU's, I would never use that in a system I use on a regular/daily basis.

 

[DFI Daishi]

up until recently, i would say that pelts were worth using within a certain price/performance/noise range. if you are a pretty typical gamer, you want to overclock your GPU in addition to your CPU. in a price range of around $500 US you could watercool the GPU and TEC cool the CPU, using the same water loop to deal with the heat from both.

a phase change solution for the CPU and a water solution for the GPU would perform better, but cost a good bit more.

with OCZ's new unit priced for less than $300 US, and likely offering better performance i think that there is now very little that pelt has to offer. add the additional cost of a water loop for the GPU and you're in the same price range as the pelt solution, with better performance.

pelts aren't THAT prone to failure, and if they do you BIOS should be able to shut off your computer in time to avoid damage. under load, pelts aren't going to keep you at super-low temps. the thermal conductivity across a pelt with no load across the terminals is also not too bad, meaning that you're temps won't shoot up much faster than other potential cooling failures would account for.

for the chiller that acascianellig put up, you need two dedicated 12V PSUs, most likely meanwell 320 watt units. you would also need to insulate all of your chilled water hoses. one loop or two to cool the hot sides doesn't make a whole lot of difference, but you need a lot of pumping power and radiator real-estate to do the job.

i run my system daily, and my own catastrophic failure came not from my cooling solution shutting down with my computer on, but rather my computer shutting down and the cooling solution staying on. i'm not in the mood to give details, and i have taken precautions to make sure that it doesn't happen again, but if you are doing any type of extreme cooling: you should have saftey shut offs to make sure that everything turns off if any one element in the system stops working normally.

 

[zer0signal667]

the thermal conductivity across a pelt with no load across the terminals is also not too bad, meaning that you're temps won't shoot up much faster than other potential cooling failures would account for.

Each side of the peltier is a ceramic plate... what's the range of thermal conductivity that you consider "not too bad"? Most ceramics, including what these are most likely made of, are considered bad thermal conductors. "Good" thermal conductivity for a ceramic material isn't even close to that of aluminum.

 

[x_dreamer]

well, how about the artic silver ceramic....... sure its not as good as metal, but it still works decently.

 

[DFI Daishi]

Each side of the peltier is a ceramic plate... what's the range of thermal conductivity that you consider "not too bad"? Most ceramics, including what these are most likely made of, are considered bad thermal conductors. "Good" thermal conductivity for a ceramic material isn't even close to that of aluminum.

i fully argee with what you have said, as stated.

however, based on some "not so precise" testing that i did with a dead pelt, a soldering iron, a chunk of copper with a hole in it and a compunurse probe: although the CPU may get too hot to be stable when overclocked, several minutes after a failure, the proc is unlikely to jump into the "kill" zone before the motherboard is able to shut the proc down. with a heat load of 80 watts from the soldering iron, applied to a 9 cm^2 area of the pelt, it took about 3 minutes for the compunurse to start reading temps of 65C at the edge of the copper block. that's with the other side of the pelt not being cooled by anything other than an Al heatsink i had it pressed against with no thermal compound and the fan off.

it's not a precise test or anything, but it did confirm for me what some users over at XS had commented on anecdotally: even if the pelt dies and the system becomes unstable, the temps are unlikely to kill the proc, at least before the saftey shutdown triggers.

personally, i have my shut down temp set at 30C..........i'm not all that worried.

 

[DFI Daishi]

well, how about the artic silver ceramic....... sure its not as good as metal, but it still works decently.

first off, you're comparing apples and oranges: AS ceramique is a thermal compound, and that's all that it has to do.

there are more design constraints placed on the ceramic solid used in peltiers.

second off, our mod is a specialist in composites.........i would imagine that he knows his ceramics pretty well.

 

[zer0signal667]

i fully argee with what you have said, as stated.

however, based on some "not so precise" testing that i did with a dead pelt, a soldering iron, a chunk of copper with a hole in it and a compunurse probe: although the CPU may get too hot to be stable when overclocked, several minutes after a failure, the proc is unlikely to jump into the "kill" zone before the motherboard is able to shut the proc down. with a heat load of 80 watts from the soldering iron, applied to a 9 cm^2 area of the pelt, it took about 3 minutes for the compunurse to start reading temps of 65C at the edge of the copper block. that's with the other side of the pelt not being cooled by anything other than an Al heatsink i had it pressed against with no thermal compound and the fan off.

it's not a precise test or anything, but it did confirm for me what some users over at XS had commented on anecdotally: even if the pelt dies and the system becomes unstable, the temps are unlikely to kill the proc, at least before the saftey shutdown triggers.

personally, i have my shut down temp set at 30C..........i'm not all that worried.

Gotcha... so, you were heating the copper with the soldering iron, as if the copper chunk was a coldplate and the soldering iron was the CPU core? It would be more accurate to measure the temp of the soldering tip in that case. The copper chunk's temp change is going to be heavily governed by chunk geometry/interface with the heat source.
I don't mean to crap on your experiment, I like to try stuff like that too, but I wouldn't be so sure that it applies to a general situation so much as your specific system.

well, how about the artic silver ceramic....... sure its not as good as metal, but it still works decently.

Right, it works decently- it's all relative. Air<silicone grease<ceramic. Therefore Air<(Silicone grease filled with ceramic). Still, (silicone+anything)<metal bonds, but it's not possible to form a perfect metal interface between the parts. Liquid metal, however... could, in theory, be worlds better. I'm still waiting for some more people to test that stuff that's been turning up on ebay

 

[Logan321]

"Liquid metal"? So what, use mercury as a thermal interface? lol AFAIK, that's the only metal that's liquid at room temperature... and you REALLY wouldn't want that running off your heatsink and on to surrounding conductive circuits.

Any solid metal particulate in a non-metal suspension is no different than arctic silver, in terms of how liquid the metal is. Advertizing gimic?

 

[robberbaron]

"Liquid metal"? So what, use mercury as a thermal interface? lol AFAIK, that's the only metal that's liquid at room temperature... and you REALLY wouldn't want that running off your heatsink and on to surrounding conductive circuits.

Any solid metal particulate in a non-metal suspension is no different than arctic silver, in terms of how liquid the metal is. Advertizing gimic?

Gallium, I believe, can be melted by the warmth of electronics.

 

[zer0signal667]

"Liquid metal"? So what, use mercury as a thermal interface? lol AFAIK, that's the only metal that's liquid at room temperature... and you REALLY wouldn't want that running off your heatsink and on to surrounding conductive circuits.

Any solid metal particulate in a non-metal suspension is no different than arctic silver, in terms of how liquid the metal is. Advertizing gimic?

Try gallium alloys, as mentioned above...
Link
I don't know what you're trying to say, but there's definitely a difference between solid metal particulate and liquid metal.

 

[Logan321]

Fair enough. Guess I missed that... still "Please do not use any aluminium coolers!" is a warning that needs to be very carefully followed, if I recall.

http://www.frostytech.com/permalink.cfm?NewsID=46586

Who reads warning labels anyway though?

Seriously though, I'd be worried about trace amounts being accidentally applied to other heatsinks in the computer which are made of aluminum, such as mosfets or passive northbridge heatsinks.

 

[zer0signal667]

Seriously though, I'd be worried about trace amounts being accidentally applied to other heatsinks in the computer which are made of aluminum, such as mosfets or passive northbridge heatsinks.

I wouldn't, because I usually try to be conscious of my actions...

 

[DFI Daishi]

Gotcha... so, you were heating the copper with the soldering iron, as if the copper chunk was a coldplate and the soldering iron was the CPU core? It would be more accurate to measure the temp of the soldering tip in that case. The copper chunk's temp change is going to be heavily governed by chunk geometry/interface with the heat source.
I don't mean to crap on your experiment, I like to try stuff like that too, but I wouldn't be so sure that it applies to a general situation so much as your specific system.

actually, since the heat density of the soldering iron tip is much higher than that of a proc, a better approximation still would be to sink in a temp well about .5cm horizontally away from the hole with the soldering iron in it, and vertically as close to the surface in contact with the pelt as the metal allows for.

i just found that chunk of copper in the scrap bin of one of the labs and used it "as is" since i don't have a drill press or even a vise here at home. holding the metal and drilling by hand would be a little hazardous for my liking so........

regarding the liquid metal TIM, i was kind of suprised talking to one of the profs in the microelectronics heat transfer lab at the university. he mentioned, among many other things, that when they tested AS3 a while back it had piss-poor thermal properties as compared to most of the products that they have gone through. i'm thinking that about the only reason that AS products do so well is because of their spreading and curing properties.

all the same.....i'm now wishing that i could fork out for some of the indium foils that they have tested.......about 13x the thermal conductivity ot AS3, as tested on their equipment. even a graphite sheet product that they had handy put up like 5x the conductivity of AS3.

 

[Logan321]

I wouldn't, because I usually try to be conscious of my actions...

So you've never had the telltale sheen of arctic silver dusted on your finger after cleaning a heatsink while preparing for reapplication? I don't believe it.

 

[zer0signal667]

So you've never had the telltale sheen of arctic silver dusted on your finger after cleaning a heatsink while preparing for reapplication? I don't believe it.

NEVER!
Seriously, it's like (hopefully) you know to wash your hands after you take a leak and before you eat... Be aware of cross-contamination and prevent it!

 

[Dark_fire]

doesnt mercury have the same effect on aluminum? why not just fill your WC with mercury and pray it doest leak onto some really expensive component

 

[Erasmus354]

doesnt mercury have the same effect on aluminum? why not just fill your WC with mercury and pray it doest leak onto some really expensive component

Because mercury is highly toxic

 

[Vertigo Acid]

And because the density is such that pumping it would be extremly impractical

 

[masher]

Fair enough. Guess I missed that... still "Please do not use any aluminium coolers!" is a warning that needs to be very carefully followed, if I recall.

http://www.frostytech.com/permalink.cfm?NewsID=46586

Who reads warning labels anyway though?

Interesting. My vote is that it's simply mercury, as I know it will react with aluminum in exactly the way described in that link.

Because mercury is highly toxic.

Mercury in metallic form isn't terribly toxic. Not that I would advise ingesting any or breathing hot mercury fumes, but its the organic mercury compounds that are the real killers.

 

[Vertigo Acid]

Interesting. My vote is that it's simply mercury, as I know it will react with aluminum in exactly the way described in that link.

And what would that reaction be? Aluminum will defintily displace mecury in compounds, but elemental mercury + elemental aluminum or aluminum compounds shouldn't generate a reaction

 

[DFI Daishi]

And what would that reaction be? Aluminum will defintily displace mecury in compounds, but elemental mercury + elemental aluminum or aluminum compounds shouldn't generate a reaction

WRONG!

or rather, it's not a displacement reaction.

metallic Al is highly unstable under normal atmospheric conditions. the only thing that allows Al to exist normally is that a thin crust of Al oxide forms on the surface, and proctects the metallic Al undreneath.

metallic Al oxidizes within a minute when exposed to air. the oxide coating on the surface of the Al is all that prevents the solid mass of Al metal from spontaniously rotting away into Al2O3 powder.

Hg is able to dissolve Al2O3 from the surface of the metal, and O2 from the atmosphere, allowing any Al in contact with the Hg to be exposed to a continous stream of O2, with the oxide being continously carried away rather than forming a protective oxide coating. the net result is that solid Al metal in contact with even a small volume of liquid Hg will rapidly be reduced into a pile of Al2O3 powder.

incidently, don't carry any Hg onto an airplane unless you have are a certified meterologist with permits, or you'll get branded a terrorist and subjected to this: http://abcnews.go.com/WNT/Investigation/story?id=1356870

god bless america, just make sure that it's a christian god.

 

[masher]

Daishi is mostly correct...mercury doesn't dissolve Al2O3, it forms an amalgam with the aluminum itself. When exposed to the oxygen in the air, th amalgam separates into mercury and aluminum oxide. Unlike the tightly-bound oxide layer on normal aluminum, however, it easily flakes away, and the elemental mercury is free to attack more aluminum. The mercury isn't itself consumed in the reaction; its merely a catalyst.

The characteristic feathery appearance of the oxide flakes is seen in the earlier link, which is why I suspect this "mystery cooling paste" to be no more than mercury.