My system, circa late 1999, consisted of a Western Digital 1.6gb hard drive, PowerMan 300 watt power supply, and K6/2-400 processor with a giant GlobalWin heat sink and fan combo. Around this time, I noticed I was having trouble sleeping at night, and began looking for moderate ways to make my computer quieter. At the time, there were maybe three sites on the web that dealt with quiet computing, and I was fairly soft core about making my computer quieter.
At the time, the main sources of noise in my system was the power supply fan. To reduce the noise, I found a Panasonic Panaflo fan, and replaced the original noisy fan with the Panasonic. It was a bit quieter, but my hard drive was still pretty loud. At the time, I needed a bigger hard drive anyways, so I bought an IBM 37.5gb, since IBM had a reputation for being quiet. In hindsight, I probably should have bought a Fujitsu, which are much quieter, or even an IBM with fewer platters, but at the time, I was pretty soft core.
The next biggest source of noise was my CPU fan. As an experiment, I tried removing it. Since the CPU was directly under the power supply fan, it ran fine during a burn-in test even at 400MHz/2.2V, but just to be safe, I clocked it down to 300MHz/2.1V. The computer was much quieter, and being low on time, I ran in this configuration for the rest of the school year.
Over the summer, I decided the power supply noise was still a bit too much, and set out on a quest for a fanless power supply. I had downloaded the ATX specs from Intel's web site, and was wavering between building my own and modifying an existing one to run fanless, when I discovered a company called TK Power made a fanless ATX PSU good to 125 watts. I ordered the PSU, but got a PSU with an ATX connector that was not electrically ATX compatible. After much more time on the phone and a bit more money, TK Power finally sent me a fanless ATX supply (the new one good to 150 watts). I ran another burn-in test on my CPU running fanless, and found it ran fine at 400/2.1V. Again, to be conservative, I underclocked it to 300MHz, and set up a warning clock when the CPU came within a fairly wide margin of overheating.
Since my system had no fans, and so only convection airflow, I ran it caseless. Just to be safe, I also placed small heat sinks on the motherboard chipset and other power-hungry components. I configured my hard drive to spin down after 15 minutes of inactivity, so over the night, or when reading, or when doing anything other than actively using my box, my system was absolutely silent. When I was using it, it only made some fairly low level hard drive noise, which I did not mind too much. Life was good.
At this point, I set up a web site to document what I did. I really did not expect much traffic, and set up the web site as much as notes for myself as to help others.
Over the next couple of months, the traffic to my web site began to increase. Each page hit meant a log write, so my hard drive would spin up. My hard drive was spun up a fairly significant portion of the time, and again, my system was a bit noisier than I liked. As the drive aged, it also gradually fell out of calibration, and the noise levels picked up. I looked into setting up RAM disks and similar. In the end, decided there was no clean software solution that gave me all of the functionality I desired, and I bought a SilentDrive hard disk enclosure. With this, my hard disk was inaudible about a yard away from the computer. Since my bed is two or three yards away, I was fine. This was my configuration for about a year. I was pretty happy, and the processor was sufficient for most everything I did (occasionally, compiles took longer than I wanted, and I could not play DVDs, but otherwise, the system was fine).
In the summer of 2001, for the first time, the CPU overheating alarm went off. As time went on, the CPU went from being able to run burn-in tests fanless at 400MHz and 2.2V to overheating after Netscape crashed and left an idle process at 300MHz and 2.1V. At first, I watched CPU usage closely, though realizing this was not a sustainable solution as the CPU eventually reached a point when it could not run fanless idle, and by the winter, could not run within the bottom end of AMD's specified 2.1-2.3 voltage range.
I reviewed my books on device physics, and found the source of the effect to be most likely related to hot electron effects. This is unrelated to silent running or reduced core voltage (despite the name, hot electron effects actually decrease slightly with temperature, and would probably also decrease with the lower core voltage). This was due to either a design or manufacturing flaw on AMD's end, but AMD technical support was unwilling to replace the chip.
For the time being, I installed a fan on my CPU. It was a Panaflo fan (my favorite model, purely because one of the supply rooms here has a huge surplus, so I can get them free). I built a suspension mount around it, first out of rubber bands, then out of more durable hair bands, so it would not transfer vibrations to the case. I ran it at 9V instead of 12V. I also placed it a couple inches away from the heat sink, so it ran unloaded. It makes little enough noise that I cannot hear it from my bed, but it is a poor solution, as the CPU is at the end of its lifetime. I've also been running fanless for over a year now, and would prefer not to stayed fanned if even purely for pride.
For a while, I had already been planning to upgrade to either a mobile or Cyrix CPU. I had contacted Intel and AMD about buying mobile CPUs about a year back, but with no luck. Neither company sold them for desktop use, and in the case of their more modern chips, some of the nicer models were often socket-incompatible. The only place where I could acquire an individual unit were EE catalogs with a couple hundred dollars of mark-up. Instead, I was patiently waiting for the Cyrix units to reach a speed where they could reliably play DVDs under Linux. In reviews, the models before the C3 had offered a negligible speed improvement over my K6/2.
Also, for some time, I had been working on a major revamp of this site, including a reference silent system built out of the best parts available, as opposed to one scrounged together based on the noisy system before I got into fanless (the K6/2 was never a very good chip for fanless operation). All three came together. Via/Cyrix was shipping a new line of chips claiming to be comparable in performance to the Intel Celeron at the same clock speed. These chips are very low power, and can run fanless by design. Since my sister's 700MHz notebook could play DVDs under Linux, I figured a C3/866 would probably be able to as well. I did a couple of web searchs, but I could not find any resellers in the US. I e-mailed the folks at Cyrix for a sample unit. They wrote back saying that they would gladly ship me one, and I just received the unit. I assembled the machine in early January, and am currently finalizing the review.
Copyright © 2000. 2001. Piotr F. Mitros. Questions? Feedback on the site? Feel free to contact me.