Atomic I/O letters column #27
Originally published in Atomic: Maximum Power ComputingReprinted here 19-Nov-2003.
Last modified 16-Jan-2015.
How cold is too cold?
I'm buying a new HSF (Zalman 6000CU) soon and was wondering what to do with my old one. After reading the latest issue I've decided to whack it on my GeForce3 Ti500 and overclock until it bleeds.
Anyway, this extreme cooling got me wondering whether hardware can be too cold. We all know that a CPU fries when it gets too hot, but can a CPU freeze and refuse to run if it gets too cold?
Joel
Answer:
Yes, it's possible to over-cool a CPU, but only in a cryogenics lab. Even
then, some other important part of the computer is likely to die by misadventure
before you manage to freeze anything into nonfunctionality.
Ordinary computer chips, including CPUs, will function perfectly well at very low temperatures. Many extreme overclockers have used liquid nitrogen (boiling point at sea level: 195.8 degrees below zero, Celsius; that's 77.2 degrees above absolute zero) to cool CPUs, generally with the assistance of very temporary improvised rigs involving polystyrene foam cups and quantities of cloth tape.
This is no way to run a computer for any length of time, but it is a good way to determine what the maximum overclock the CPU can manage is when temperature is not a factor.
When cooled like this, CPUs aren't actually running at nearly-200-below; they're sitting in the middle of a storm of boiling nitrogen at some hard-to-determine temperature. No motherboard or in-chip thermal probe can be relied upon to give you a good reading this far out of their intended range, but it's safe to say that LN2 on a processor will keep it a long way below zero, provided you keep the coolant topped up.
In the real world, cryogenic computing becomes very difficult very quickly, for several reasons. You either need perfectly dry air around the computer, or you need to waterproof it thoroughly, because condensation is going to be a serious problem whenever the whole board isn't below zero. Condensation and ice crystals inside hard drives (which aren't sealed; they have vent holes with dust filters) are a Very Bad Thing; all of your drives need to be away from the cold area.
Various people have tried long-term cryogenic computing using foam coolers and converted fridges and mineral oil immersion and other tricks; something usually ends up rusting or fracturing or delaminating. The CPU generally survives.
Water cooling systems are much more practical. A decently implemented sealed pump, water block and radiator arrangement will cool a CPU a lot better than any air cooler can manage, won't need much maintenance, and should be pretty quiet, too. A good water cooling setup will also generally get you within about 10% of the maximum overclock the processor's capable of with LN2 cooling. The effort needed for really extreme cooling does not match the speed improvement you get.
Up a certain creek
One day I turned on my PC, and while I was waiting for it to start I got that "oh no" feeling, the feeling you get when the message "cannot find XXXX.XXX please locate the file" or something like that flashes up.
After many minutes of swearing I decided to put my hard drive in my other PC, and it found it OK, but when I explored it I only found 1 folder and 2 files. I then ran Norton Disk Doctor and did a scan; it found most of my files but they were all unorganised and had names like DIR00000 all the way through to DIR00136. Most of the files were 0kb.
I really want to get my files back as I have a Web site to maintain and about 15Gb of games and keys for my shareware that I have registered.
Is there a program that can fix it?
Justin
Answer:
Sorry, but you're boned. Your hard drive might just have had a major filesystem
conniption and not actually suffered a hardware failure, but your data is
still functionally unrecoverable. This is the experience that teaches you
to make backups.
Remember this mantra: If it's not backed up, it's not data you want to keep.
Tonight, on "Which Solvent?"
Much inspired by Atomic, I decided to cover the hole I was cutting in the side of my Lian Li PC-70 with a custom cut pane of 4mm glass. The window hole I cut was almost identical to the one in Atomic's "Heavy Water Project" - kinda in a "P" shape.
I also fixed this glass to the aluminium side cover with silicone as Atomic and my local glazier suggested and/or confirmed as the best idea for this particular adhesion. Fair 'nuff said I, until the panel lost balance while I was working on the PC and the glass broke.
I've tried all manner of spirits from acetone to petrol to soften the silicone, but it ain't gonna give before the expensive aluminium side panel does. The two surfaces are too close to get any kind of blade in under there, and even if I did, the hooks and rolled edges on the panel interior aren't going to allow much working room.
So, please, any suggestions? How the hell do I get that broken glass off my side panel that I put so much effort into cutting out?
Bacchus
Answer:
The usually recommended solvent for cured silicone adhesive is "sandpaper".
But you can do better, if you've got a bit of time and the materials involved
aren't too touchy - which they're not, in this case.
Ordinary silicone is not petroleum-proof. It doesn't dissolve instantly when petroleum products touch it, but it will come apart eventually. So if you put the case-side in a shallow tray of some sort and pour, say, petrol on it, then leave it for a week or two, the silicone ought to just fall apart. Briefer immersion would probably be adequate; I haven't done scientific tests, beyond accidentally destroying a silicone rubber exhaust coupler for a model car some years ago.
Pretty much any petroleum product should work for this. Kerosene or light oil, for instance, not to mention various thin solvents of varying levels of ferocity. Heck, Vaseline will eat silicone, but a thin liquid is better if you want it to penetrate.
Once you've removed the glass and the silicone, you'll need to give the panel-side a good scrubbing with detergent to get rid of the petrol residue, but then it should be ready for your next attempt.
There are some brand-name products for this purpose as well - DAP's "Silicone-be-gone" and 3M's "General Purpose Adhesive Cleaner", for instance - but they're not likely to be easy to use on a sandwiched film of silicone.
Temperature panic
I recently discovered that my Athlon XP (2000+) was running a little hotter than I would like it to. At restart it was about 54 degrees Celsius (this is the supplied cooler). Then I checked it with the supplied mobo monitor utility after playing UT2003 for an hour or two, only to discover that it was running in the 70 degree area.
I panicked. This was far too hot for my (shamefully) un-over-clocked 2000+.
I have now investigated the possibilities of cooling and considered buying a Peltier and a Volcano 11. Is this insanity? What do you think the most effective solution is, and what risks are involved (this is of course excluding the luxury of VapoChill and water cooling, mainly because I cant afford them)?
Alex
Peltier devices can make ice easily, unless they're loaded up with a CPU's
heat...
Answer:
My standard answer to questions of this type is: Does your computer crash?
If the answer to that question is "no", then you don't have a problem.
Not yet, anyway.
It's possible for cheaper thermal transfer material to deteriorate, resulting in steadily climbing CPU temperatures that can eventually leave you with a flaky PC. If the temperature when you're doing some particular task isn't getting higher and higher as the weeks go by, though (remember to account for ambient temperature - if today's 25 degrees C and yesterday was 20 degrees, you can expect your CPU temperature to be five degrees higher today...), then there's probably nothing to worry about. If the CPU temperature is climbing, then taking off your CPU cooler and re-greasing the contact patch, preferably with a premium thermal grease that won't dry out, will probably deal with the problem.
Numbers from motherboard temperature sensors are unreliable. Under-chip sensors commonly vary from board to board, depending on how firmly the sensor's touching the chip, if it's touching at all; the sensors inside current processors are more accurate, but can still be badly miscalibrated. BIOS updates can and do change CPU temperature readings, by changing sensor calibration. So your "70-plus degree" CPU temperature readings aren't necessarily accurate, anyway.
Again - if it ain't crashing, don't worry about it.
If you're using a stock cooler at the moment, and you'd like to cool your CPU a bit better, then Thermaltake's current chunky Volcanos are perfectly good options for the money. Steer clear of the heat pumps, though. Peltier devices are very difficult to get working properly.
Since your processor can output about 70 watts when working hard at stock speed, you'd need a Peltier device that can shift that much heat to a CPU cooler, while maintaining a decent temperature differential between its two sides. This means a high power Peltier, or an inelegant parallel-Peltiers arrangement with a giant cold plate.
You'd probably need a separate power supply for the Peltier(s), and you'd also need to condensation-proof your CPU socket and the motherboard behind it, or use a thermostatic control that'd keep the CPU above the dew point when it wasn't working hard. And then you'd probably want some automatic shutdown system or other, to save the CPU from burning up if the Peltier or its power supply should fail. It's all just too much trouble, if you ask me.
NTLDR not found. Use NTCRSHR?
I was running Win98SE and wanted to upgrade. I brought another hard drive, installed it as a slave, formatted with Seagate DiscWizard, and then installed XP (NT File System) on it, and transferred the files I wanted to keep from the 98SE disk.
I want to run the XP drive as the master, and format the 98SE disk to use as a backup disk. But have found that if I change the jumpers on the drives, the BIOS settings, or disconnect the 98 drive, I get an "NTLDR error".
Is there any way of fixing this boot error without reformatting my XP disk?
Iain
Answer:
NT-series Windows versions, including WinXP and Win2000, can't really
be installed on any disk other than the one you boot from. Well, not without
help from outboard boot manager software, at any rate.
If you tell XP to install on D:, practically everything will be installed there, but the loader - NTLDR - will be on the boot drive. Remove or re-letter the boot drive and NTLDR can no longer be found.
If you swap the drives over and then boot from the XP CD, you should be able to use the "repair" install option to make the existing installation work, without having to nuke it from orbit and start again.
Duron de-dotting
I have a 1.3GHz Duron. Its usual temp is around 50 degrees + on a normal day! I know AMD processors run warm, but the fan is quite loud and frankly I'm going insane!
I have come to the conclusion that I need to improve the cooling in my case and on the CPU as much as possible. I'm wondering - is it OK to remove the little foam pads on the processor? Will this increase the contact with the CPU causing more efficient cooling, or will I cause irreparable damage and be left with "an expensive paperweight"?
Douglas
Answer:
Point one: Dead CPUs make very poor paperweights. They are, however, popular
as key-chain
ornaments, and if you push hard enough you may be able to use them to hold
a document to a corkboard.
Point two: Don't remove the rubber dots.
The dots on the top of Socket A CPUs (AMD calls them "Compliant Load Support Pads"...) are there to stop the CPU cooler from crushing the core. You're most likely to damage the core by tilting the heat sink while you install it; the rubber dots stop it from tilting too far. They don't greatly reduce the contact pressure between the CPU and the heat sink.
Some people remove the dots and install a metal shim instead, but the only vaguely coherent reason to do that is so that you can use an extremely high pressure bolt-mount cooler.
All other things being equal, swapping the dots for a shim makes no difference to anything but your bank balance.
RAM + RAM = HANG
Hi. I have just purchased 512Mb DDR-266 to add to my current 256Mb (which is also DDR-266). OK so the 256 works fine by it self, the 512 works fine by itself, however when I try them together to give 768Mb there is major system instability. The 256 is a Synnex stick, the 512 is an Elixir. The motherboard is an Asus A7S333 with an Athlon XP 1600+, running Windows XP Professional.
Simon
Answer:
Welcome to the wonderful world of memory incompatibility. It's common
enough, especially when you're trying to make two budget memory modules
like these work together.
It's also possible that one or both modules is damaged (probably by static electricity), but problems like this arise quite often with cheap RAM, even when it's never been mis-handled.
You can try installing the modules in different memory slots, but that's not very likely to help. If you try to get a refund, you may find it difficult, since the retailer can just insist the old module must be the source of the problem, or that any damage must be your fault.
To be fair, it often is the customer's fault; many people don't realise that just taking two steps can be enough to build up a tiny static charge that'll damage a RAM module.
