Wednesday, May 4th 2005

NanoCoolers puts liquid metal in your PC

Most of you may not have heard of NanoCoolers yet. This company is a relatively young Startup(May 2002) who has been working on improving cooling technologies behind closed doors for quite some time.

Now the first details on their cool new products emerge:
The liquid metal has significant advantages over other single phase liquid solutions. The thermal and physical properties of the material give it the ability to cool extremely high heat fluxes. With its very low vapor pressure, the boiling point of the material is in excess of 2000°C. This provides the capability to cool extremely high power densities without the liquid-metal changing phase, removing power density as the limiting factor in cooling performance. The liquid metal is non-flammable, non-toxic and environmentally friendly. As a metal, the liquid is both highly thermally conductive and highly electrically conductive. The thermal conductivity makes it ideal for heat removal and dissipation. The electrical conductivity enables the use of electromagnetic pumps to propel the liquid.
We got our hands on a 16 page confidential presentation:

Add your own comment

86 Comments on NanoCoolers puts liquid metal in your PC

#1
sgi
Hi, I recently get confirm that the Nanocoolers liquid metal cooling will NOT appear in that video card. In the November 2005 version of MAXIMUM PC, an article of " Liquid-metal-cooler videocard is a No-Go" tells you all (page 11).

So, guys, the most important thing is using your brain first and then the mouth. If you don't understand, ask the experts before claiming.
Posted on Reply
#2
sgi
NOW, it's DEAD.

See this poster then you know everything is OVER.


http://www.theinquirer.net/?article=25164


BTW, the new topic of theirs -- thermoelectric cooling is JUST another topic of cheating money. It will NEVER succeed.

If you go back to visit here after ONE year, you will know what I say is true.
Posted on Reply
#3
ScottKin
One benefit from this so-called "Liquid-Metal" cooling system is this:

Q. How many of you have ruined your system because of a leak in your water-cooling system?

Everyone who has had this happen or have heard of it happening, please raise your hand!

With the proposed Liquid-Metal cooling solutions, the system is a closed-loop, within tubes simiar to what are used for heat-pipe solutions like those from Zalman. The liquid-metal can't leak out.

Whether this will be an actual solution, in light of NanoCoolers pulling their L-M product, has yet to be seen.

--ScottKin
Posted on Reply
#4
ericgrau
I'm a Mechanical Engineering student at UC Berkeley, taking a heat transfer class.

The limiting factor when cooling a concentrated heat source - like a tiny CPU - is conductivity, not specific heat. This idea shows great promise.

Heat is indestructable and must go somewhere. All heat sinks dissipate 100% of the CPU's heat; no more, no less. The amount of heat dissipated is a combination of efficiency and temperature. The amount of heat dissipated is fixed by the CPU. So efficient heat sinks dissipate the same heat while running at a lower temperature, period. Great heat sinks do so with low air flow, either from a slow fan or warm air rising naturally.

Common TECs take roughly 100-266W of power to cool 100W, not 1000W. TEC's have a trade-off between performance and efficiency. Refrigeration takes about 20-40W to cool 100W. Both are up to twice as efficient when underloaded (which advertisers assume :) ).

ALL nuclear power plants use heavy water (i.e., with deuterium in it). It's essential.
Posted on Reply
#5
CjStaal
In addition, I would like to point you to this information:

"Some early testing showed that prototype cooled Radeon X850XT PE card to just 12 Celsius.


yeah probably in north pole cos liquid cooling cannot cool the cpu below roomtemp.[quote]Not to belittle Canada any more than she already is...but the safest nuclear reactor title? Try the United States Navy. All those reactors on carriers and submarines, being operated by people as young as 21? No accidents? No nuclear-related deaths over the course of 50+ years? Sorry Canada.
--------------------------------------------------------------------------------

LOLOLOL no incidents.. If there even be an accident u think they'll tell it to public?..

Yeah and what about those two "nuclear reactors" that usa dopped to japan? Few hundrend thousand dead! Oh wait that wasn't an accident.
[/quote]A. Yes there has been deaths on submarines due to nucleur reasons... Ever hear of the infamous german subs that almost ALWAYS killed the crew, of course, not right away, gotta keep the sub. Who gives a shit about the german people? Of course, not Germany.

B. Those were Atom bombs buddy.
Posted on Reply
#6
CjStaal
Unregistered
I hate to be the one to inform you that the US is not without nuclear screw-ups. Three mile island ring any bells?
Exactly my point also.
Posted on Reply
#7
sgi
I am happy Ericgrau replied my post. I have two things to add:

1) thermal resistance If you read any textbook, papers or tutorials on electronics cooling, you can find the formula of thermal resistance. It has two parts, first one, 1/(hA), is connected with thermal conductivity of the fluid; and the second one, 1/(2mCp), is connected with the fluid capacity. Check electronics-cooling.com if want to know more. Liquid metal has very high thermal conductivity. However, the pump is not strong enough to provide enough mass flow rate (like nanocoolers). Therefore, the performance of liquid cooling is weak. Same in Cooligy (Kenneth Goodson at Stanford founded), their electroosmitc pump doesn't work either. That's the reason Cooligy was sold. These two companies collected over 20 million dollars for their project each. But they gave up, finally (you can check their website or contact them).

The lesson here is, check it carefully before starting.

2) TEC efficiency You are partially correct. I cited the number (10%) is too small (from nanocooler guy's paper). The COP of TEC is generally can be order of one or less, depending on a lot of factors. Among those factor, temperature rise (delta T) and Tref are two dominated factors besides the material properties. Commercial TEC products have no impressive performance in refrigeration cooling. That's the reason many companies are trying to improve it (like Nextreme, Nanocooler, JPL etc.). But my view is, before ZT is greater 3.0, TEC cannot compete with other technologies.

The lesson here is, don't believe it before thinking carefully.
Posted on Reply
#8
ericgrau
sgi
I am happy Ericgrau replied my post. I have two things to add:

1) thermal resistance ... 1/(hA), ... 1/(2mCp),


...
my view is, before ZT is greater 3.0, TEC cannot compete with other technologies.
the h in 1/(hA) is related to fluid conductivity (~10-20 times better for liquid metal vs. water) and heatsink geometry & conducivity (~same in either case). In my thermo textbook they actually derive values of h from this. So liquid metal cooling wins hands down here. The second equation, which should be 1/(m'Cp), depends on the mass flow rate (m' = density times flow rate) and specific heat (Cp) of the fluid (~3 times better for water). Water wins hands down here. For a standard 180gph pump: m' = 12.5g/s, Cp = 4J/g, so R = 1 / (12.5 * 4) = 0.02C / W. So an 80W cpu would lose only 1.6C from water or ~5C from liquid metal. Not much. Also keep in mind that heat transfer temperature. Most cooling methods do nearly all of their cooling at the hottest parts near the tiny CPU, and not much everywhere else. Using liquid metal (or a copper base plate, for example), helps spread out that heat a great deal, dramatically reducing the 1/(hA) term. So unless the flow rate is below, let's say ~60gph (which might be the case), liquid metal should win overall.

COP 3.0, huh? Well a TEC could be advertised with a COP of 2.0 assuming low load (which is how they normally rate COP). I think TECs are still by far the best solution for cooling CPUs below room temperature, because refrigeration is just too big and expensive. I'm working on a setup to overclock my next 2 video cards at between -57.5C (3 stage) and -20C (2 stage), depending on the card's power consumption (40W-80W). Over 8 years I'll supposedly get my money back plus ~$500 from not buying more expensive video cards. These estimates come from a close look at TEC performance curves, the thermal resistivities of my heatsinks, the contact resistance of thermal compound, etc. The supposed money saved is still up in the air, and I'll have to take a close look at some actual video cards to be sure.

Here's the secret: Run your TECs at 6V instead of 12V to get the higher efficiency. To get 6V, just wire 2 in series. That will make it easier to stack multiple stages and easier to cool the final stage. For a more conservative setup, I'd recommend running 4 72W TECs in parallel to cool an 80W load, or 2 in parallel to cool a 40W load. Don't supercool the CPU directly. Use a water block on the CPU and cool the fluid. This keeps your heat spread out more. 2-4 TECs may seem like a lot, but remember at 6V you don't get as much performance - your "72W" TECs aren't 72W anymore. They're roughly equivalent to 40W TECs, except for their higher efficiency. But it's worth it because power hungry 12V TECs are next to impossible to cool. Why use multiple 72W TECs instead of 1 powerful TEC? Again, it's to spread out the heat more. 4 times the area (and 4 times the heatsink size) means 1/4 the losses. By my numbers (which could be wrong), this setup should get your chip down to 17C with cheap (but decent) heatsinks & fans cooling the TECs, or as low as 12C with expensive heatsinks & fans cooling the TECs. Power consumption is only 20W per TEC, so you don't need a seperate power supply like most setups.

The common way to do this is to get a 212W TEC, a $48 heatsink, air cool it with a massive 60dB fan (!), get a very powerful case fan, get an auxillary power supply to run the TEC, get a relay to automatically turn on your auxillary power supply and then hopefully you'll be down to 20C. If that's not enough to spend by itself, your electricity bills will make you wonder why you didn't just buy a better PC. Not to mention the difficulty you'll have hearing your games with your fan running, unless you replace the fan and heatsink with a very expensive water cooling setup (your ordinary water cooling setup won't cut it). You want 2 stages to run it even colder? Okay, let's get 3-4 more 212W TECs...
Posted on Reply
#9
sgi
I am happy someone wants to discuss :)

the h in 1/(hA) is related to fluid conductivity (~10-20 times better for liquid metal vs. water) and heatsink geometry & conducivity (~same in either case). In my thermo textbook they actually derive values of h from this. So liquid metal cooling wins hands down here.

You are right! But this part isn't important in total thermal resistance :(


The second equation, which should be 1/(m'Cp), depends on the mass flow rate (m' = density times flow rate) and specific heat (Cp) of the fluid (~3 times better for water).
Water wins hands down here.

For a standard 180gph (A pump for liquid metal with this number has not been developed yet and it is expected no this product in the future) pump: m' = 12.5g/s, Cp = 4J/g, (for liquid metal, cp=0.365 J/g) so R = 1 / (12.5 * 4) = 0.02C / W. So an 80W cpu would lose only 1.6C from water or ~5C from liquid metal. Not much. Also keep in mind that heat transfer temperature.

Most cooling methods do nearly all of their cooling at the hottest parts near the tiny CPU, and not much everywhere else. (Wrong, if you check carefully for any liquid cooling system, an ambient heat exchanger is always there. You must guarantee heat can be taken away by air finally. Otherwise, the system will die immediately. Therefore, total thermal resistance must includes this ambient heat exchanger. Unfortunately, liquid metal performs badly there.)

Using liquid metal (or a copper base plate, for example), helps spread out that heat a great deal, dramatically reducing the 1/(hA) term. So unless the flow rate is below, let's say ~60gph (which might be the case), liquid metal should win overall. (Check the 1st page of this post and see the performance of Nanocooler's devices. Then think carefully and ask why its performance doesn't match your estimation. Another easy way is to call nanocoolers and tell them you have a better way to save liquid metal cooling technique. :))




COP 3.0, huh? Well a TEC could be advertised with a COP of 2.0 assuming low load (which is how they normally rate COP). (NOT COP, It is ZT) I think TECs are still by far the best solution for cooling CPUs below room temperature, because refrigeration is just too big and expensive. I'm working on a setup to overclock my next 2 video cards at between -57.5C (3 stage) and -20C (2 stage), depending on the card's power consumption (40W-80W). Over 8 years I'll supposedly get my money back plus ~$500 from not buying more expensive video cards. These estimates come from a close look at TEC performance curves, the thermal resistivities of my heatsinks, the contact resistance of thermal compound, etc. The supposed money saved is still up in the air, and I'll have to take a close look at some actual video cards to be sure.

Here's the secret: Run your TECs at 6V instead of 12V to get the higher efficiency. To get 6V, just wire 2 in series. That will make it easier to stack multiple stages and easier to cool the final stage. For a more conservative setup, I'd recommend running 4 72W TECs in parallel to cool an 80W load, or 2 in parallel to cool a 40W load. Don't supercool the CPU directly. Use a water block on the CPU and cool the fluid. This keeps your heat spread out more. 2-4 TECs may seem like a lot, but remember at 6V you don't get as much performance - your "72W" TECs aren't 72W anymore. They're roughly equivalent to 40W TECs, except for their higher efficiency. But it's worth it because power hungry 12V TECs are next to impossible to cool. Why use multiple 72W TECs instead of 1 powerful TEC? Again, it's to spread out the heat more. 4 times the area (and 4 times the heatsink size) means 1/4 the losses. By my numbers (which could be wrong), this setup should get your chip down to 17C with cheap (but decent) heatsinks & fans cooling the TECs, or as low as 12C with expensive heatsinks & fans cooling the TECs. Power consumption is only 20W per TEC, so you don't need a seperate power supply like most setups.

The common way to do this is to get a 212W TEC, a $48 heatsink, air cool it with a massive 60dB fan (!), get a very powerful case fan, get an auxillary power supply to run the TEC, get a relay to automatically turn on your auxillary power supply and then hopefully you'll be down to 20C. If that's not enough to spend by itself, your electricity bills will make you wonder why you didn't just buy a better PC. Not to mention the difficulty you'll have hearing your games with your fan running, unless you replace the fan and heatsink with a very expensive water cooling setup (your ordinary water cooling setup won't cut it). You want 2 stages to run it even colder? Okay, let's get 3-4 more 212W TECs...

There is some comparison among TEC solution and Non-TEC solution. I am sure you can find somewhere. The baseline is, you must have a heat sink with very low thermal resistance so that you don't need too much increase in temperature to dissipate the extra heat from TEC itself (electricity). Otherwise, it will get worse.
Posted on Reply
#10
ericgrau
A detailed thermal analysis is needed to discuss liquid metal further :) . I don't want to bother. But I can say any good textbook would do for those interested. Basically liquid metal will in fact do a better job if the pump flow rate is very high. But apparently that's difficult to accomplish. Any time I mentioned "cooling" I meant heat transfer from CPU to the closest fluid (water or liquid metal, not air except in a pure air cooling setup). I assume a good radiator is feasible with any liquid cooling system.

Said TEC complications involved a TEC setup with TECs run at full power. I plan on using low power and common high thermal resistance CPU heatsinks.
Posted on Reply
#11
PVTCaboose1337
Graphical Hacker
Yes, YES! now we can overclock more and that means FAST!
Posted on Reply
Add your own comment