Slashdot Mirror
Liquid Nitrogen Beats Air Cooling (Again)
joe094287523459087 writes "some guys used liquid nitrogen cooling via a cardboard tube to get a 20,000 3D Mark score. you can see the frost forming on everything - wouldn't the moisture from the condensation kill the board?" The Muropaketti guys had already done this with their microprocessor. Apparently the next step was to speed up their graphics card to match.
I'm sure the soldered joints have no problem at all with the MASSIVE and abrupt temperature changes here.
When can we expect this on fuel cell powered laptops?
wouldn't any condesnsation freeze before it could short the electronics?
What's so special about him getting 20372 by overclocking his P4 to 3916 MHz?
The article has a picture showing that someone got 21504 by overclocking to 3998 MHz -- nine days earlier.
The boiling point of nitrogen is miles below the freezing point of water, so in the immediate vicinity you'd have some fast chilling and ice crystals. I'd be more worried about liquid condensation further away from the cpu, if it lowered the internal temperature of the case for example.
Unless they're venting vast quantites of boiling nitrogen into there though, it shouldn't be too bad. Plenty of hydrophillic stuff ariund should deal with that. A few sugar cubes in there, or some conc. sulphuric acid maybe!
Is it me, or is that guy pouring liquid nitrogen with his bare hands? Is he aware of the danger of getting that stuff on his hands?
Ok, so the Radeon has frost over most of its circuits. Is that dangerous for the said card? frost tends to melt into ... well liquid right?
I must be missing something, if someone could please fill me in...
how does one change his
So you can some awesome performance. What's the point? Lets say the guy does some UT testing and gets dick hardening results. But isn't that useless if when you're playing the game, and:
1.) The frost/condensation shorts out the board
2.) You run out of liquid nitrogen, and the board fries itself, locking up the game.
There is nothing like playing UT while worrying about filling up the liquid nitrogen. With my luck, I'd end up accidently pouring the liquid nitrogen directly on the board, causing it to fiercly boil, and I'd drop the dewar flask and get the shit all over my hands.
Besides, the blazing speed of the graphics will probably do nothing for me, so long as I have to deal with flaky lag on my broadband connection.
The limiting factor in the performance of modern GPU's seems to be memory bandwidth. They were able to overclock the GPU itself a good bit, but not much on the video RAM.
We really need to see more memory bandwidth saving technology on GPU's. ATI pushed ahead a lot of cool things (early Z, occlusion culling, Z-compression, fast Z-clear), but it's not far enough. The Kyro/Dreamcast use tile-based deferred rendering rather than immediate mode, and the GameCube's GPU (designed by ArtX, which is now a owned by ATI) uses a 2 MB on-chip Z-buffer cache which alleviates the need to go to video memory every time they want to do a Z-test (which is typically at least once per pixel). Given, the Cube doesn't ever have to deal with a frame buffer bigger than 720x480, so a fixed size Z-cache is much more useful there.
On another note, I'd really like to see support for geometry amplification schemes (n-patch tesselation, displacement mapping, etc.) that work properly with stencil-buffer volume shadows.
Lex orandi, lex credendi.
I've been doing some experimenting myself and I'm almost ready to post my 'steel is stronger than cardboard' article. Who would have guessed?
This is cool in some horribly over the top manner, but I think the last cooling using liquid nitrogen story a few months back covered it.
That aside, Liquid Nitrogen might be a bit of overkill here. When properly coated, I'm sure the parts are safe from the damaging effecets of melting, but everyone needs to remember one thing, heat is the enemy, it's obvious that it would take quite a bit to get too cold.
Obviously there isn't a future in Liquid Nitrogen cooled computers, but take the idea back from the "weird science" to the new liquid/radiator idea. I do believe silent machines running cooler with liquid cooling, will become a new trend.
Lastly, why does everyone brag about their 3DMark scores? If you suck at gaming the extra pixels sure ain't gunna help ya.
Ignore the "p2p is theft" trolls, they're just uninformed
Oohhh, let's hate the MPAA! Hey look, LOtR stuff!!!
No shit Sherlock.
Build a compressor/cooler.
If you could do it right, and can afford the proper parts to build it. Create a high-power compressor, and have it pipe super-cooled fluids continually through both processors. Therefore, you don't have to pour/buy liquid nitrogen everytime you want these results.
The problem is that it costs a lot and is quite difficult to build a fast high-power compressor/cooler. If you guys can accomplish this, then I'll be impressed. Pouring liquid nitrogen onto stuff to keep it cool isn't really that exciting/impressive in the long run, since it is far to manual, and doesn't require much thought to come up with the idea.
I am curious though, does anyone know of more fancy coolers for the processor that work impressively but don't require constant manual addition of coolant?
~ kjrose
Checking out his 3dMark and you notice, fillrate, poly count, shader, spride speeds are missing. Also only 4x AGP, be nice to also see 8x AGP enabled, his motherboard might not support it yet.
His ATI driver is also 6.13.10.6159, he should upgrade to 6193, major performance increase. You can get it over at rage3d.com
Impressive thou, Double my 3DMark on a plain AMD 1800 with a ATI 9700.
Run the whole thing in an improvised glovebox filled with dry nitrogen gas at slightly higher than ambient pressure. Boiling liquid nitrogen results in large quantities of totally dry nitrogen gas.
:-)
Now the only problem remaining is how to avoid condensation on the glovebox itself so you can see what's going on inside
Stop worrying about the risks of nuclear power and start worrying about the risks of not using nuclear power.
Well--eventually, maybe. But what they've build is essentially the back half of a water distillation system. The water that condenses out of the air will be very pure, and have a very low conductivity. (The resistance of a 1 cm path through ultrapure water is on the order of 18 meg--that's ohms, not byes--so it probably conducts no more electricity than the plastic of the board.)
Yes, the condensate will eventually pick up contaminants, and at the edges of the cooled region where liquid water is free to flow you're likely to have problems. The solution would be to keep the entire mainboard in a dry environment. Seal it in a box with only an inlet for LN2. The little bit of water in the box will condense out (on the N2 fill pipe rather than the board if you remove a bit of insulation) and as the LN2 boils off, the box will be filled with dry, inert nitrogen. As an added bonus, this will help suppress fires.
~Idarubicin
Liquid Nitrogen Beats Air Cooling (Again)
In related news:
Ice is colder than Steam
Touching hot things will burn you
You can skate on ice
Lots of fire will make things melt
The speed of light is very fast
That's not frost you see, but chip outlines silkscreened onto the board. These are used by manufacturers to ease asssembly (eg, capacitor C32 goes here, resistor R98 goes here, chip IC42 goes here, etc.).
Put my fist through my alarm clock with its ding-dong death inside my ear. - The Blackjacks.
More like common sense.
"And like that
Have the guys on the ISS (or indeed any space mission/station) ever used the 'cold hard vacuum' of space to get their systems running cool and fast. You don't have the problem of condensation, you could make the sink a great big copper array (being very, very sure to always keep it out of the sun) and you'd enjoy even better performance (it's only about 3 Kelvin out there...).
I guess you'd run into problems with cosmic radiation - but nothing a good dose of shielding wouldn't fix. Placement would also be an issue, couldn't just pin it to the back of the station...
Anyone know of any experiments along those lines?
It's not that I'm Anti-American - I'm Pro-Freedom
Darwin award candidate, surely?
"And like that
Why is this 'high school physics.'?
I can say that....
Air is mainly nitrogen.
Gas at 20 deg c has 26 moles per decimiter cubed, I don't know what it is for Nitrogen vapor at -100 - degs.
I don't know the spacific heating capacity of air or nitrogen vapor (the amount of energy required to raise x moles of y by 1 deg) 'not quite high school physics, though I could work it out'
Ok so is fairly obvious, but there's a lot more than plain old 'high school physics' to think about.
thank God the internet isn't a human right.
Obviously there isn't a future in Liquid Nitrogen cooled computers,
Live today, because you never know what tomorrow brings
Chemistry is just a mislabelled branch of physics
11*43+456^2
The moisture forming on the board is distilled water. It is about as clean as possible and has a very low electrical conductivity. It is mildly corrosive, but given that the only exposed metal surfaces are tin or gold, that is not a problem if the exposure is not too long.
I would think that you need to immerse a board in distilled water for some weeks or months to get actual damage. Fans, HDDs and other moving parts are a different story.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted and ignored otherwise.
OK, maybe i'm missing something, but at that temperature, wouldn't some parts be smaller than they should be and some larger? I'm pretty sure that the coefficient of thermal expansion of plastic != that of silicon != that of copper. And besides, if you consider silicon, its resistivity is negative, meaning it generally increases with temperature increase. (it may be positive, depending on how your teacher taught it.) So although cooling it would result in higher conductivity for the copper, wouldn't that put the silicon to sleep, effectively?
I'm the Devil the Windows users warned you about.
It would appear that muropaketti.com's server is using the same technology to avoid the slashdotting meltdown that I fully expected.
/.'ers. Brrrr. Is that all you've got? Fsck it's cold in here. All your hits are belong to us. God Dammit!!!, when I was shipped here from California I knew Finland was cold, but this is just fscking ridiculous."
I can hear it talkin' smack right now - "Bring it on you
__ Someday, but not this morning, I'll finally learn to use the preview button.
It would suck to have wasted all of that time with ln2 and still not have the top score. http://holicho.lib.net/top/020912/020912.htm http://gamershq.madonion.com/hardware/halloffame/
It's really, really cold.
Bibo Ergo Sum.
Wouldn't what they're doing exceed environmental specifications for the chips?
I looked around ati.com and intel.com, but couldn't find any specifications on what the upper and lower bounds are as far as temerpature is concerned. I recall seeing in most product specifications for electronic devices temperature limits, and I thought the lower was usually around -15 degrees celcius.
Or does the temperature of the chips ever get that low? Do they hit some kind of equilibrium that keeps them from reaching their lower limit?
-kidlinux.
Shoes beat top barefoot running speed
Telephone more effective than tin can and string
Umbrella proven more effective at staying dry than dodging raindrops
Mark
And I just got a Pentium II 233MMX IBM ThinkPad and was hella excited about that. But seriously, why doesn't someone try out these experiments on a Linux system? I'd be curious to see the differences in performance. I'd do it myself, but as I noted earlier I just got a PII 233 Thinkpad so we know what my financial situation is.
Also, if I remember rightly, the actual drain currents of the transistors goes UP because the resistance is going down (which is why you can overclock, of course.)Although the lowered temperature means the tracks will not be damaged, there may be other effects of the increased current density in longer term degradation of the die. If there is track necking anywhere, this might be a potential failure point.
You might also expect damage to the epoxy cladding of the graphics chips, as the contraction pulls the epoxy away from the filler. This could result in the epoxy eventually becoming porous and the system failing due to moisture penetrating the cladding, just like 6502s etc. used to fail before anyone realised that glass fibre filler could wick water in to the die.
The answer is to follow Seymour Cray and sink the entire system in cold fluorinert, using the total loss nitrogen system, or much cheaper dry ice, to keep the temperature at a sensible -45C or so. But that wouldn't be nearly so spectacular, would it?
This is all a bit like our local hot rodders who can't safely make it to the next town and back for fear the engine will blow up on them. Even so, it would be nice if Intel would release some of the data they doubtless keep on this sort of thing.
Panurge has posted for the last time. Thanks for the positive moderations.
At least it's not another "Let's beat the dead horse about Microsoft!" article with 600 stale monopoly jokes.
Kind of to go with this, I found a (Finnish, sorry, but you can still look at the purdy pictures) page that has some interesting avi clips. These people got an infrared camera to do a few DivX clips on how the heat is distributed across the components. I like the first one best; they actually destroy an uncooled 1.4GHz Athlon Thunderbird. The CPU temperature rises to 300 deg C in about six seconds!
You can't shut us down! The Internet is about the free exchange and sale of other people's ideas!
When using rigour ways of cooling like just pouring LN over your board usually the capacitors give out first. Since they always attract moisture like crazy (remember those old radio's used to die because the capacitors were leaking electricity due to moisture in them?) there's always a small ammount of moisture in them. Now you don't want to have icecrystals forming in there. But still it's amazing what those boards can take. I have seen an old BX motherboard bent at an odd angle by someone pouring LN over it and still working. (Due to the immense temperature differences in the upper and lower side of the board the upperside folds inwards as material shrinks when the temperate is decreased) I guess the soldering is really done well. The same thing must have been happening at this chip btw. The underside of the GPU must have been warmer (a lot warmer) than the upperside so i don't want to even think about the massive ammount of material stress going in. ;) Well.... just keep up tweaking those GPU's and CPU's, it's always fun for us to watch.
but it seems like there should be some sustainable way of acheiving this kind of performance from these processors... could better cooling be acheived by increasing the pressure? anyone know how many atmospheres of pressure a P4 can handle?
There are plenty of reasons why this would not have worked.
1) The specific heat of LN2 is much lower than water. This means that although you can get it cooler, you might not be able to get away enough heat to make a difference, or it might form a vapor barrier that prevents the LN2 for cooling efficiently. My guess would have been that you needed to actively pump it or agitate it. In fact, you probably can get quite a bit more cooling this way. If you've ever pour LN2 on your hand, you know that it doesn't really do very much (unless you cup your hand). Try this with boiling water and see if it hurts or not.
2) Since none of these components were designed to withstand low temperatures, there's a chance that something could have cracked. If there was something that had a vastly thermal expansion coefficient in the packaging and/or became very brittle (e.g. plastic), it could have just split open. Apparently that's not a big problem either. My bet is that they cooled it down pretty slowly before they turned it on.
I think the condensation issue is not as bad as one might think. It takes quite a bit of ions in order for water to become very conductive. It might be worthwhile to rinse your board in distilled water before trying this, but you probably don't have to bother. Just don't pour any salt on your board and it should be OK.
Just wait till they hold the air cooling part of the contest on Neptune. Liquid Nitrogen won't be so smug then, I can tell you!
You see? You see? Your stupid minds! Stupid! Stupid!
from the no-shit dept.
The speed of light is found to be "really fast"
I don't have any good hard figures on this, but it seems to me that if you want to cool something down that near absolute zero, you're going to need something a bit more active than a big radiator. In labs here on earth that sort of cooling can only be done on a nanoscopic level, generally using lasers to slow down atoms.
-
from my knowledge, water created this way - ie condensation isn't yet ionised an doesn't have the ability to conduct electricity.
if many of you know the celeron [tm somewhere], the ppga grid is visible from the top of the chip and I myself have had water covering every pin and moving it's way down the board just to test this very idea of water damage.
[peltiers - not NO2] = P
fyi no damage was incurred and I was damn suprised.
- We seek not the answers, but to understand the question.
In all actuality, water is not an electrolyte. It's the minerals and elements in water which conducts electricity. This means that distilled water(condensation) does not conduct electricity, and therefore would not short out electrical equipment.
What you reap is what you sow
Chemistry is just the physics of electron shells (among other things). Biology is just the chemistry of organic molecules (at its heart anyway). Psychology is just.. well you get the picture. Physics is about finding the fundamental theories of the universe. Everything is physics by definition.
Of course, sometimes the link between physics and a higher level abstraction is poorly understood. For example, human behaviour is just the firing of neurons and so should be explainable by physics but we aren't even close yet.
I'm not suggesting that we get rid of these higher level sciences. Anyone who's programmed in assembler will realise that abstractions are important.
Sure, pure water is a bad conducting media, but that is offset by all those dust particles and soluble chemicals on the components. You can perform an experiment: Get a bottle of RO-water (reversed osmosis, much cleaner than distilled water) and pour in on your moderboard while running the computer. Se it short out. This particluar experiment has been carried out a few years ago. It is possible that the motherboards have become much cleaner since then, but I doubt it.
The last word means that convection and conduction are no longer options for getting rid of heat.
Vacuum happens to be an excellent insulator...
And when the station is in direct sunlight it heats up VERY quickly. You can expect temperature swings for something in orbit to go from freezing cold to burning hot - Often in the course of a few minutes.
I was involved at school in designing a small satellite (See cubesat.org for a general description of this class of satellite) - I was (fortunately) handling the radio board, and wasn't one of the thermal guys. I felt sorry for them...
During dark times, the satellite had to have small heaters to keep the batteries from freezing.
20-30 minutes later in the orbit, the satellite's problem would be overheating.
retrorocket.o not found, launch anyway?
To get accurate temperature info from thermal images, you need to paint everything you're imaging. (Usually black - Get it as close to an ideal black body as possible). Otherwise, some colors are better radiators than others and will appear warmer on the camera at the same temperature.
That said, even unpainted, a thermal camera can tell you, "Something is really heating up quickly... That isn't right..."
retrorocket.o not found, launch anyway?
I used to program in assembler, and I wish more people still did. Actually my ideal programming language is now "C++, but don't use templates, polymorphism, or multiple inheritance - dont bother with the STL unless it's a very clear win - and profile your code and recode any super-hot-spots in asm". I guess what I really want rather than C++ is C with just a few C++ features - like member functions for structs, private declarations, operator overloading, and constructors/destructors.
My point is, abstractions sometimes go a bit too far.
11*43+456^2
Maybe you should have a look at some of the embedded subsets of C++. For example, drivers for Mac OS X are written in a subset that removes most of the stuff you mentioned. IIRC, all that is left is single inheritance and member functions. Constructors and destructors are also left out. new just allocates. You then have to call init() on the method to do the setting up. This is in line with Objective-C, and it isn't so bad. It reduces some of the complexity too.