Atomic I/O letters column #51
Originally published in Atomic: Maximum Power Computing Reprinted here November 2005.Last modified 16-Jan-2015.
Canned cold
As I'm sure we both know, when something is cold, it shrinks. The atoms in it get closer together and move less. What would happen if you cooled something so that the atoms were close, and then locked it in a airtight container (very strong) so that it couldn't expand again? Would it stay cold?
I was thinking about it, and was wondering if you put liquid nitrogen into a very strong cube and stopped it expanding, could you plonk it on your CPU? Would it be forced to stay cold as a result of the compression of the walls of the cube on it?
Nathanael
Answer:
Actually, not all substances contract when they cool.
Water expands as it
freezes; so does
bismuth. Molten
bronze expands as it cools, then shrinks again when it solidifies
(this is why it's so good for making castings - it expands into the mould crevices, then shrinks and unsticks itself).
What would happen in the case you mention - and what does happen, whenever someone fills a container with compressed gas - is that the contents of the container will equalise temperature with the outside world, as normal, as heat seeps in or out. If the container's getting warmer, then the pressure inside will increase.
If you pump room-temperature gas into a cylinder, the gas (and the cylinder) will get hot as you compress it. But you certainly could just start out with an open container somewhere really cold, then seal the container and go somewhere warm. That'd even work as well as pumping the gas in, if your really cold place was, say, a liquid oxygen ocean on some gas giant moon, or something.
When you took your bottle of liquid oxygen back to your hot computer in Australia, though, it'd get warm and the pressure would increase (and, if it wasn't a tough enough pressure vessel, it'd pop).
If you then uncorked the bottle, the LOX would expand (very enthusiastically!) and escape, and both the bottle and the gas would become very cold, thanks to this expansion. You can feel this effect in a less dramatic form when a spray can gets cold while you're using it.
Find more information on the pesky laws of thermodynamics standing between you and perfect CPU cooling here.
That pr0n's all public domain, I swear
Just wondering if you could tell me the best way to ensure that my activities on the Internet are not snooped on. This especially applies to P2P activity. I have heard of programs such as GhostSurf, which claims to route data through anonymous servers. I doubt this, as the bandwidth requirements would be huge; I don't know how they could afford it with a one-off payment of $US50.
I would like to know if these programs are likely to protect me.
Rob
Answer:
The feasibility of Internet anonymity depends on what "activities" you're engaging in, and who's doing the snooping.
If the NSA wants you, the NSA's probably gonna get you, because they'll come
to your house and plant hardware and/or software bugs in your PC when you're not home.
If you're just Joe Average who doesn't want his credit card info stolen, though, then there's no real software solution, beyond keeping your computer up-to-date with security patches, and not buying stuff from Honest Yuri from the Ukraine who's only got an eBay feedback score of 2.
This especially applies to P2P activity.
Most P2P software is inherently insecure. Data-source peers in any ordinary file trading network have, by definition, to know the IP address of the data-destination peers, so all a snooper has to do is run software that says (truly or otherwise) that it's a source for various stuff, and they'll be able to see the IP addresses of the people requesting it.
There's more to it than that, of course; some degree of anonymity is possible, and developers are getting more and more interested in it as the movie and music companies get more and more enthusiastic about destroying their customers' lives.
About the only way to "patch" some degree of anonymity onto regular P2P clients, though, is by using a blocklist for IP addresses known to belong to snoop-y organisations, or by using Tor and hoping for the best, bandwidth-wise.
I don't know how well GhostSurf works. But at least it doesn't claim to protect you from prison rape.
Hot in the server tonight
I currently own a dual Xeon 2.8GHz with 6Gb DDR266 RAM, and a Supermicro X5DA8 motherboard. I am having serious overheating issues. The "Super Doctor III" monitoring program says CPU 2 is overheating; it reaches temperatures of about 85 degrees C, about 6 degrees hotter than CPU 1.
I have a Cooler Master case and standard Intel CPU coolers with the "P04" brackets. At a 21X multiplier with hyperthreading turned on, CPU 2 heats up within 5 minutes and the system freezes and alarms go off. I have disabled hyperthreading and turned the multiplier down to 15X and things are stable, but slow.
What cooling systems (air or water) are the best for a dual Xeon system? Is it possible to run one water cooling system for both CPU's?
Brian
Answer:
If you're running WinNT/2000/XP, it'll tend to hand system processes to the first CPU and applications to the second,
which can result in CPU2 being mildly warmer when you're doing stuff. Only mildly, though. Of course, if some heavy
duty single-threaded application has its affinity set to
the second CPU, then that could account for your machine's odd behaviour.
Alternatively, it could just be a lousier thermal contact between the second CPU's cooler and the chip. The thermal goop may have dried out, or there might be a hair between the chip and the cooler, or the cooler's fan may be on the way out, or there may just be a big dust clot in that CPU's heat sink. All of these problems are, of course, solvable without spending a whole lot of money.
If you've got the Intel Xeon coolers that come with the nice copper heat sink, you're unlikely to be able to find a better heat sink on any after-market cooler. Most third party Xeon coolers are made for slim rackmount servers, and have a lot less fin surface area than the stock units.
The Intel coolers have custom fans on them, but if you need to replace one with a standard square fan then you can use cable ties and tape and do a perfectly good job. If the idea of that offends you, then a 2U rackmount server cooler (like this, for instance) with a better fan on it should be more than adequate.
(A reader's now pointed out to me that he's encountered heat pipe coolers that've apparently sprung a leak, lost their heat pipe working fluid, and become hideously inefficient as a result. If you've got such a cooler, that's something to check for - see if the base of the heat sink's hot but the fins are cold.)
Pay attention to your case ventilation as well, though. Make sure the front fan filter (if there is one) isn't clogged with dust, make sure there aren't any ribbon cables blocking air flow to the CPU coolers, maybe put in a more powerful exhaust fan.
Water cooling for Xeons is perfectly possible, and yes, you can use two water blocks from one pump and radiator. You'd probably want to run the blocks in parallel, rather than daisy-chaining them, but in the real world either arrangement is likely to work as long as your pump's reasonably grunty.
There aren't many water blocks made specifically for Xeons, but various P4 models fit with a different mounting bracket, and you can get some excellent ones for non-ridiculous prices.
Add a standard 120mm-fan radiator and a decent pump (anything that doesn't have a nasty little brush motor in it should do; see this, for instance), and you'd be in business.
Double barrelled
Being a bit of a storage nut, I've got two PCs full of IDE disks (10 each). As you can imagine, 10 disks is stretching the envelope of a ordinary midi-tower PC case (I'm already using detachable 3.5 inch drive "boxes" from other cases screwed under the normal 3.5" enclosures in the two servers).
Now I'm thinking of building a great big server case (out of 1-2 cm thick wood preferably, to keep the noise down), slapping into it all the drives, a decent surplus dual P3 mobo (Intel 440GX+) with a gaggle (technical term) of PCI IDE controllers, two 350W ATX PSUs, plus assorted CD/DVD drives, a DDS-3 streamer etc. This would be easier to manage logically (RAID and LVM on Linux), probably better-cooled AND quieter (assuming I used 12V fans running at ~7V), and generally nicer-looking.
The question I have is: can I safely use two ATX PSUs (some drives on one, the rest + MB + other devices on the other) in one PC? They'd be powered from two outlets on the same UPS, so there shouldn't be any problems with different ground potentials... but am I missing something?
And another thing, would you advise to splice and join the PWR-ON wires on the ATX leads to allow both PSUs to be controlled by the motherboard/case switch, or would it be better to simply plug one into the MB as usual and to do the bridging trick on the other?
Piotr
Answer:
Yes, you can use two PSUs in one computer with no trouble, and you can indeed
rig them to power up elegantly by splicing the green pin-14 PS-ON wire
of the secondary PSU to the same wire on the ATX connector of the first PSU.
Or, as you say, you can switch the secondary PSU separately, or just stick ye olde paper clip into the secondary's ATX connector so it's on all the time.
The two PSUs should share the same earth reference if they're both screwed into the same chassis, but for completeness it's also a good idea to connect any earth wire on one of them to any earth wire on the other. Running a zero-gauge earth strap from one PSU casing to the other is generally considered to be overkill.
More demanding nutcases have extended this, and run three or more PSUs together. And yes, it really is this simple, as long as you don't mind them still working as separate current sources, and not truly sharing the load (to do that, something like this is needed).
It's even possible, though not easy, to buy a pre-built PSU-slaving cable. Cooler Master sell one with their CM Stacker case, which has two PSU mount locations.
