Atomic I/O letters column #32
Originally published in Atomic: Maximum Power Computing Reprinted here April 2004. Last modified 16-Jan-2015.
Mad-scientist department
I have been studying some chemistry, and have found out about exothermic reactions and endothermic reactions. Exothermic gives out heat and endothermic takes in heat. What if a custom liquid cooling rig was built which used an endothermic reaction? If suitable chemicals were found, could this be a very easy and cold method of removing that unwanted thermal problem?
Gabe
Answer:
Sure, this'd work, for as long as your chemical supply lasted. You'd need
a couple of buckets of reagents, and you'd need to keep topping them up,
and you'd need to do something with the waste, too.
One simple endothermic reaction - actually, it's not technically a reaction at all - is the one used in commercial "instant cold packs" for treating injuries. The packs get cold when you squeeze them, because you've ruptured a seal inside that lets some ammonium nitrate (more often used as fertiliser or explosive...) dissolve in some water. Many things will cool water when you dissolve them in it, but ammonium nitrate's especially good at it, absorbing 26.2 kilojoules of heat per mole. One mole of ammonium nitrate weighs 80 grams.
One joule is one watt-second, so 26.2 kilojoules is 26,200 watt-seconds, which is the amount of energy a 100 watt heater (like an overclocked CPU) will emit in 262 seconds. An ammonium nitrate and water cooling system for a 100 watt load would, therefore, go through about 18.3 grams of ammonium nitrate per minute.
Of course, you could set up an evaporator arrangement to reclaim the ammonium nitrate (and distil back the water, as well, if you were feeling really clever). This is left as an exercise for the reader.
If you're going to use something consumable to cool a CPU - presumably because you're shooting for a ludicrous overclocking record - you might as well do it the traditional way and use liquid nitrogen. No messy mixing and metering required; just seal a foam cup to the top of your processor, and pour in more liquid N2 whenever it looks like boiling dry!
Insufficient entropy
How do computers (which can only calculate formulas) generate random numbers? They can't cycle through a list and stop at a random point in the list, because that requires a random number first. Is there a formula that can generate a random number?
John
Answer:
Computers don't generate random numbers (well, almost all of them don't).
Instead, they generate "pseudo-random" numbers. Pseudo-random numbers may
have a random distribution (generate a million pseudo-random ones
and zeroes and you'll get pretty much the same number of each, as you should)
but there will always be non-random patterns to the output - repeating sequences
that wouldn't exist in truly random data.
This is because a normal computer has no source of true randomness to use as the input to a pseudo-random number generator (PRNG) program. PRNGs can twiddle "seed" data into superficially-random-looking numbers as big as you like, but they can only produce output as random as the seed. Given the same seed, a PRNG will always give the same output.
The quick'n'easy way to seed a PRNG is to feed it some time-based number. If you seed a PRNG with the number of milliseconds that've passed since midnight, for instance, you'll get output with a repeating 24 hour pattern.
This doesn't matter if the PRNG's just generating "random" events in a computer game (though the very non-random spraying of bullets in games like Counter-Strike can be used to players' advantage), but it's important for things like cryptography. If the random numbers used for an encryption key aren't actually very random and an attacker knows it, they can restrict a brute force keyspace search to only those areas of the keyspace that the encoding computer could have actually used. This can result in surprisingly easy decryption of apparently highly secure data.
A PRNG seeded by something genuinely random will do a much better job. SGI's famous but now defunct "lavarand" project used cameras looking at lava lamps as the randomness source, for instance. It's mutated into "LavaRnd", which is less hilarious but more useful.
Via's Nehemiah-core C3 CPUs also have a proper random number generator (RNG) built in. Some software is now starting to support it.
Pixel fever
Why is it that my laptop (Dell Inspiron 8200) has a native resolution of 1600 by 1200 with only a 15 inch screen, and yet I can't find a single LCD desktop monitor that can handle a resolution this high? Even the best 19 inch Sony screens will only go as high as 1280 by 1024. I've been thinking (dreaming?) about replacing my 21 inch Sony Trinitron (also running at 1600 by 1200) with a couple of LCD flat screens - but not at the cost of a reduced desktop area.
Simon
Answer:
1600 by 1200 desktop LCDs exist - they're all at least 20 inchers, though,
and they ain't cheap.
For instance, at the time of writing, just off the top of my head (he lied, after a brief burst of research), there's five 20 inch NECs, one 21 inch NEC, a couple of Sonys (but I'll only link to this one, to punish Sony for their even-worse-than-NEC URLs), and IBM's ThinkVision L200p 20 incher (a cheapie, at only $US1349!), Hitachi's CML200B, and Samsung's SyncMaster 213T.
Apple have widescreen displays that fit the bill, too; the 23 inch Apple Cinema HD Display is 1920 by 1200, and the cheaper 20 inch Apple Cinema Display still manages 1680 by 1050.
1600 by 1200 desktop LCDs could indeed be made cheaper - maybe considerably cheaper - if they used laptop panels instead of the larger, coarser-pixeled desktop ones. But desktop monitors are typically viewed from further away than laptop screens, and the monitor makers don't want to bet that people will be willing to pay for pixels they can barely perceive.
Imperceptible pixel grain is a good thing; you've got to have it if you want a truly photorealistic display. Today's operating systems just aren't ready for it, though.
Display devices will have invisibly small pixels in the fullness of time, but right now it's a pain to make text and window gadgets and such big enough to be legible on a small, hyper-resolution screen at desktop distances. Heck, 1600 by 1200 on a 15 inch screen can be pretty eye-straining even when the screen's sitting on your lap; many laptop users would love a bigger screen, if only there were some way to fold it into an easily carried package.
17-inch-screen laptops exist, but they're all widescreen units (usually with resolutions like WXGA, 1366 by 768 pixels), and even then they're aircraft carriers. Until we can make the screens roll up, we're just not going to fit anything bigger than a 17 inch widescreen into a laptop that doesn't fall into the old "transportable" category.
On a small 1600 by 1200 screen, you can tweak many text and graphic elements to be big enough that they won't be set in one point Myopia when viewed from a couple of feet away, but the ones you can't fix are irritating enough to those of us in the 20/20-or-worse vision category that the monitor manufacturers don't want to try to sell small hyper-res screens to just the 20/10-and-better population.
Even at laptop distances, 1600 by 1200 on a 15 inch screen can require a lot of squinting when you're viewing un-scalable things that assume they're going to be shown on a classic 72 dot per inch monitor. Many people run their monitors at more than 72dpi and are happy enough with it at desktop viewing distances - 1280 by 960 on an 18-inch-diagonal "19 inch" CRT is 89dpi - but 1600 by 1200 on a 15 inch LCD is 133dpi. That'll make anything that can't be scaled only 54% of the size it was probably meant to be.
Highly squintworthy, in other words.
Earthy inquiry
Power leads have 3 prongs on the plug - positive, neutral and earth. My understanding is that the earth is for safety reasons. Most metal appliances have the earth attached to the casing to let wayward current go to ground instead of through the user, however there is no such connection on metal computer cases. Is there one inside the power supply?
Thomas
Answer:
Yes, there is. The earth contact on the power cable connects to the PSU
casing, and thence to the computer chassis via the PSU casing's connection
to the chassis, and several other paths. All of the black wires coming out
of the PSU are connected to the same earth, and normally end up connected
to the chassis via motherboard mounts, drive casings and so on.
It's just possible to foul this up if you've got a fancy lacquered PSU, mount it with low-conductivity anodised screws, put fibre washers on top of the standoffs under your motherboard, and mount drives in plastic fan cages or rubber-suspended noise isolators. If you've got little or no path to earth from the computer chassis, then a PSU failure that puts 240V on some output wire or other can result in a nasty belt if you touch the computer or a device attached to it.
Live-case problems are much more likely to be caused by a defective power lead or mis-wired wall socket, though.
Woes * 2
I can't transfer files over MSN Messenger. I have Messenger 6.0 running on WinXP Pro, on Optus cable. I have two computers networked; this is not the Internet computer. I have Windows' built in firewall turned off, so I don't think that's the problem.
Also, a couple of months back, I bought a new motherboard, CPU and RAM on a trip overseas. In excited haste at getting my new computer up and running, I got it out as soon as I got home, but after my 25 hour trip with no sleep, my brain wasn't exactly working. As I was attaching the heat sink to the CPU socket, my screwdriver slipped and made a tiny gash in the motherboard. The gash is only 5 mm long, as wide as the corner of the screwdriver, and half a millimetre deep, but it knocked off a little component between the CPU and the RAM. All I ever got from the new system was a memory error beep, so I was forced to buy a new mobo.
Well, time has passed and I am now wondering if I can repair my Asus A7V8X-X.
Tim
Answer:
Your Messenger problem is probably because the Internet sharing computer
is using Network Address Translation (NAT), through which MSN Messenger's
fancier features
don't work.
Microsoft's solution is that you stop using NAT and start using an old-fashioned proxy server, which is not convenient.
ICQ and AOL Instant Messenger have similar problems, though you may be able to get ICQ to work.
You may also be able to use Trillian.
Or you can do it the hard way. Set up your network ye olde static IP address way (no DHCP), configure your instant messaging software to use some particular port for file transfers, then open that port on your connection sharing/firewall machine for the IP address of that computer. You can even keep DHCP, if you can make the DHCP server always assign a particular IP address to the MAC address of the computer you want to file transfer from.
Regarding the missing component - if you can find a proper circuit diagram or, more practically, just another A7V8X-X, then you could identify the missing widget and replace it. It's probably just one surface mount resistor or capacitor, and can easily be replaced by anyone who knows how to do surface mount soldering.
This assumes, however, that the gash hasn't damaged any tracks on the board, or at least only a surface track. If you've damaged more than one layer of the board, it'll probably be extremely difficult for anyone to fix.
