Atomic I/O letters column #78Originally published in Atomic: Maximum Power Computing Last modified 03-Dec-2011.
Being the overclocking enthusiast I am, I've pondered the use of liquids that have a lower boiling point than what you'll find in most extreme tech overclocking articles, which is liquid nitrogen (LN2).
How about LHe, though? Can't anyone cool helium close to boiling point and use that as the cryogenic cooling liquid?
Wouldn't that make for a better and cooler (pun not intended) liquefied gas for CPU cooling experiments?
It's too expensive, and it wouldn't work.
Nitrogen is 78% of the atmosphere, so all you have to do to get impure liquid nitrogen is compress and super-chill air. If you want pure liquid nitrogen, you fractionally distil liquid air - like making whiskey, only colder - and can thus make the stuff in vast quantities at relatively low cost.
Helium, in contrast, only makes up about five parts per million (by volume) of air. It's functionally impossible to extract significant amounts of helium from air.
Instead, helium comes from fossil fuel deposits, where it's often found in significant percentages mixed with methane and nitrogen. Helium's the second most abundant element in the universe in general, but it doesn't hang around near (or under) the ground on planets if it has any choice at all, so we have to go looking for it. And if we run out of fossil helium deposits, we'll be screwed.
Once you've got your helium, liquefying it is the second problem. At sea level, nitrogen boils at 77 Kelvin, 196 degrees below zero Celsius. Cold, but not so cold as to be ludicrously difficult to make or transport.
Liquid helium is used as coolant only in applications where absolutely nothing else will do, like the superconducting magnets in MRI machines, which won't work at mere liquid nitrogen temperatures. That's because helium has a sea-level boiling point of only 4.2 Kelvin - four point two Celsius degrees above absolute zero.
It is really, really hard to chill things down to 4.2 Kelvin.
All of the simpler processes to do it require large amounts of liquid nitrogen to make small amounts of liquid helium, and the other couple of processes - which are what we use to make LHe in bulk - require an enormous amount of power per unit of produced liquid.
Storage and transport of liquid helium are a nightmare, too. Liquid gas tankers allow some of the gas to boil off to allow the rest to stay liquid (and prevent the tank from exploding...), but liquid helium boils off so fast even when surrounded by tons of insulation that you pretty much have to make it (from compressed gaseous helium) where you use it, not have it delivered in a liquid state.
And then, after going to all of this trouble, you'd probably find that your amazing super-overclocked system didn't bloody work anyway. The quantum weirdness that allows semiconductors to work also means that they're likely to stop working again below about 100 Kelvin (which, yes, means that liquid nitrogen rigs should be turned on before you start pouring the liquid in).
And, more prosaically, liquid helium would be likely to freeze components around the CPU that are better left unfrozen. You shouldn't, for instance, expect electrolytic capacitors that've explored the lower ranges of the Kelvin scale to work afterwards.
My slightly antique 3.4GHz P4 hit 100 degrees Celsius today. It tends to run extraordinarily hot all the time, often hitting 80-84°C under load.
The exact model of CPU is an Intel P4P-EM64T (Prescott) Pentium 4E 90nm. I had the same problem for months with the 3GHz version of the same CPU, and having tried everything else I replaced the CPU a few days ago with the one mentioned above, but still the heat problem exists.
Here's what I've done so far to try to remedy the problem:
Installed a better cooler. I ended up with a 120mm Zalman copper cooler, huge!
Replaced the power supply – I had indications the rails were not running at the voltages they were meant to and thought this may have fixed the problem.
Applied thermal paste – then removed it when it didn't help.
Underclocked my CPU down from 3.0GHz to 2.1GHz. This meant it topped out at 74°C, still far too hot from what I know.
Bought a new CPU.
Am I missing something? Could it be the thermometer in the motherboard that's dodgy?
What would you next steps be, aside from canning the whole computer and starting again from scratch?
100°C is about 30 degrees over that P4's official maximum operating temperature. The "maximum case temperature" listed on this Intel page is the temperature of the sensor in the middle of the CPU's heat spreader, not the temperature of the air inside the computer's case - so a CPU heat reading about 30 degrees above it ought to be impossible.
Some current CPUs can keep running at that temperature, but any P4 ought to crash hard if it gets that hot.
So you've actually just got a CPU that's misreporting its temperature, or a motherboard that's misinterpreting the temperature the CPU reports. Since you've seen the same symptoms with two CPUs, I'd bet on the mobo being the culprit.
If the computer doesn't crash all the time, then the CPU is not actually too hot and you have no problem to solve.
You mention, though, that you applied thermal paste and then removed it again. That's bad. There should be a thin smear of paste on the top of the CPU, just to fill the tiny air voids between the metal of the CPU's heat spreader and the bottom of the CPU cooler.
In reality, lots of people put a big old toothpaste-commercial blob of grease between their CPU and the heat sink, and it works well enough. It's really not that critical if you're not going for absolutely optimal performance. But there still ought to be some grease in there. A dry joint between the CPU and heat sink could well give you a genuinely overheating CPU in the future.
Can you use magnetic-tip screwdrivers in (relatively modern) computers without risking damage to components?
I've seen warnings with regard to possible corruption of PROM chips, and I suppose the magnetic tip could induce small currents in devices in ideal conditions. But I've also seen people warning about damage to floppy and hard disks (the latter seems unlikely), which makes me wonder if this is still relevant to modern computers. I've seen these screwdrivers used by PC repair shops, and haven't heard of a computer being destroyed.
Yes, they're safe. You probably couldn't even damage data on a floppy disk with most magnetic screwdrivers, and hard drives have much higher coercivity than floppies, and are thus much harder to magnetically damage.
An actual PROM is a write-once device that can't be erased by any means, though you could corrupt the data on one by applying the same voltage to the programming pins that was used to program it in the first place. That'll be at least 12 volts, which you're not going to generate by waving a magnet around over the board.
EPROMs can only be erased by UV light, so they're safe too.
EEPROMs are another relatively high-voltage device, so magnetically induced voltages aren't likely to be anywhere near enough to corrupt them, either.
I want to add a smoke machine to my case, but most smoke machines use an additive in water right? Is there any smoke machine that I can safely put in my computer and not kill it?
There are two basic kinds of modern theatrical fog machine (as distinct from dry ice, which gives you a great low-lying spooky fog, but isn't really a switchable effect).
Smaller smoke machines use, as you say, a water-based fluid, usually with glycol in it to make the actual smoke. I think it's almost always triethylene glycol, which is a low-toxicity substance. Ethylene and diethylene glycol will work too, but both of them are pretty poisonous; they're the glycols that kill you if you drink that sweet, sweet antifreeze.
It's pretty much impossible to poison yourself with triethylene glycol smoke. Some people feel sick when they breathe it for a while (especially if it's the "flavoured" variety that some fool invented), but most people are fine in even ludicrously thick glycol smoke. I certainly am.
The smoke as it comes out of the machine is mainly hot steam. That's obviously not a good substance to spray on electronics. The smoke will also leave a thin layer of condensed glycol on surfaces, and the glycol is hygroscopic - it attracts water. So over time there could be a corrosion risk, too.
Just filling a room with glycol smoke a few times won't do computers in that room any harm, though. Well, not based on my experience, anyway.
The other kind of fog machine, sometimes called an "oil cracker", is basically just an atomiser, like a perfume sprayer only more so. It blasts light mineral oil into ultra-fine droplets. Oil fog is not very healthy to breathe - mineral oil isn't particularly toxic, but hydrocarbons in the lungs are generally inadvisable. Oil fog is completely fine for electronics, though, and will never hurt anything unless you manage to fog up the lens in your optical drive with it, or something.
Unfortunately, oil crackers tend to be big heavy loud machines that're made for fogging very large areas. I don't know how easy it would be to make a small one.
It's quite easy, in contrast, to make a tiny glycol fogger. All you have to do is boil the smoke fluid somehow, and you get smoke. The Zero Blaster smoke ring gun I reviewed a while ago has a little tiny glycol fogger in it that runs from AA batteries.
If I were going to put a smoke machine in a PC, I'd make it a glycol fogger. It wouldn't be hard to make one that'd run from the 12V rail of the PSU, with a bit of nichrome resistance wire and hardware-store parts. If you left it running for ages, though, you'd probably end up with corrosion problems in the PC, plus more subtle problems like a flat CMOS backup battery (because moisture would bridge the terminals of the battery, which has very little capacity).