Atomic I/O letters column #37
Originally published in Atomic: Maximum Power Computing Reprinted here September 2004. Last modified 16-Jan-2015.
Clan analogue
After watching a not so recent episode of Doctor Who ("The War Machines") in which an intelligent computer tries to take over the world (think Skynet with only 2k of RAM) I started to think about analogue computers. While working at RMIT some time ago I came across parts of an analogue computer in a display cabinet. My question is, how does an analogue computer function compared to a digital computer? Are there any working analogue computers in existence? Would it be feasible to build an analogue computer today?
Jason
Answer:
Analogue computers represent information with variable quantities - positions
of components, voltages, or other things that can be set to different values
to represent different input values. Kids learning about maths with those
coloured wooden Cuisenaire
rods are doing analogue computation.
Electronic analogue computers work by using thermionic valves not as simple relays, but as variable amplifiers, so that the widely variable output values of different valves can be added and subtracted and multiplied according to the values of other valves.
Digital computers, in contrast, do calculations with a "finite state machine", in which values are represented by things that have distinct possible value levels, with nothing in between.
Mechanical analogue and digital computers are still in pretty common use around the world. An abacus is a digital computer, and a slide rule is an analogue computer.
I don't think anybody uses electronic analogue computers for anything any more. They all date from the valve days and thus take up lots of space and power, are absolute monsters to keep in working order, and are quite easy to emulate on current hardware anyway. Sure, you could make one today (and make it much smaller, by using solid state hardware), but it wouldn't be useful for much.
Australia's first computer, CSIRAC, is now "restored" and on display in Melbourne Museum, but the restoration didn't extend to actually getting it to run. An expert's been quoted as saying that powering the 30 kilowatt monster up today would probably cause it to catch fire.
Slot selection
After reading Atomic issue 38, I've decided to make a RAM drive for my page file. Is there a way that I can assign the RAM disk to a particular memory slot (in which I would have a cheap yum-cha 512Mb stick) and use my two sticks of good RAM for everything else?
Malcolm
Answer:
No, you can't do that. Nice idea, though!
Big discs
I'm thinking of purchasing a DVD burner, but I am still unsure about media compatibility issues. For example, when DVD technology came out, it was said to have been capable of storing up to 17Gb, but cheap DVD burners still only support only 4.7Gb discs.
Gareth
Answer:
The capacity of one side of a basic single layer 12cm DVD is 4.7 billion
bytes (not the usually quoted "4.7 gigabytes"; a gigabyte is 1,073,741,824
bytes, and 4.7 billion bytes is 4.38Gb). Single layer recordable and rewritable
DVDs have that much capacity. You can copy the entire contents of a full
single sided, single layer DVD to one DVD+/-R or RW disc.
Double layer DVDs have a capacity of 8.54 billion bytes (7.95Gb; those "17Gb" DVDs are double layer, double sided, and hold less than 16 real gigabytes), writable DVDs that offer that much capacity have been around for a little while now, but aren't good value yet.
Dual layer discs should be readable in everything that can read regular DVD-Rs and DVD+Rs; the next generation of 12cm disc will hold a lot more, and need a whole new kind of drive.
Given that single-layer DVD writers are very cheap now, though, it's not completely daft to buy one anyway. They do DVDs, they do CDs, and you can use them to back up even dual layer discs, if you're willing to accept some quality loss. A transcoding utility like DVD Shrink will do the job.
Top secret files
I downloaded a nifty tool that tracks HDD usage, and while browsing to see where my disk space had gone, I noticed a very peculiar thing. In my Windows 2000 "winnt" folder, the program said that there was a folder called CSC which was using 2Gb of space. Explorer on the other hand said that there was no such folder, although I have "show hidden files and folders" checked.
I entered the address manually (c:\winnt\csc), and the folder showed up. There are lots of weird files in here with no extensions. Can you tell me what this folder does, and why it's so hidden?
Ognen
Answer:
That's the Client Side Cache folder, created when you tell Windows to make
network files available offline. It can indeed be quite enormous, if you've
got a bunch of stuff cached in there.
It should have both the Hidden and System flags set, but it shouldn't be super-ultra-hyper-hidden as you describe. Are you sure Windows hadn't decided that you didn't want to see hidden files in that directory?
Hot hot hot
In a past Atomic, I saw a review of a CPU cooler that was tested at an ambient temperature of 19°C. My current summer ambient is a mere 29°C.
In mid 2001, Atomic did a review of 10 Athlon motherboards. Regrettably, I chose the ASUS A7A266, one of four with no fan on the Northbridge chip. The heat from the CPU cooler blew over it as well.
After about three months of running the Great Internet Mersenne Prime Search, either the CPU or the Northbridge expired. The 1400MHz Athlon didn't have a built in temperature diode, and the motherboard guessed it was running at 55-65°C.
I ordered an EPOX 8KHA and Athlon XP1600+. This motherboard reports a max CPU temp of 45°C, and worked fine for 20 months of its two year warranty. Then COM1 went bad, then one USB port, then COM2. Must be the Southbridge that's going flaky (no cooler on it); the case has a total of four ventilation fans (two inlet/HD, one exhaust, one PSU).
What is the point of your testing hardware under conditions that only apply in Tasmania and the South island of New Zealand?
John
My CPU simulator. You can tell I'm a professional, can't you?
Answer:
The ambient temperature for cooler tests doesn't matter, as long as you
know what it is. Air and CPU temperatures change by exactly the same amount
for a given thermal load. A cooler that keeps a given load down to 50°C
(say) when the input air temperature is 20°C will score 70°C if the input
air's 40°C.
I'm saying "input air temperature" rather than "ambient temperature", because the air inside the case of a running computer won't stay at ambient temperature for long after you turn the PC on. The better ventilated the PC is, and the less heat its components emit, the closer its internal temperature will be to the outside ambient. A real wind-tunnel PC won't exhaust air that's more than a few degrees hotter than it was when it went into the case, and most tweaky machines can keep the internal air temperature less than 10 degrees hotter than ambient, without making too much fan racket.
Atomic magazine's "Chernobyl" test rig, like the far uglier one I use for my Dan's Data CPU cooler tests, just sits there in the open air. There's nothing wrong with this, as long as you correct for ambient temperature. You could sit the Chernobyl rig outdoors at Mawson Station in July and still make perfectly good test numbers (well, as long as you put a windbreak around it).
When an 80 watt simulated P4 load only gives a heater temperature rise of 16°C above ambient on the Chernobyl rig (the results for the 3D Cooler Pro), you're looking at a seriously good air cooler. Such a cooler will definitely deal with the heat output of quite ferociously overclocked CPUs, even if the case it's in has highly questionable ventilation and the ambient temperature's well into the computing-in-your-underpants range.
To simulate the conditions your computer has been running under, a magazine or Web site wouldn't just have to test hardware at high temperatures; they'd have to do it for months on end. They'd also need to test several units, to get a reasonable idea of the real probability of failure at different times. In other words, you'd like them to set up a full MTBF reference lab. This could mean another couple of zeroes on the cover price of a magazine, and it wouldn't allow for lots of other variables - possible static damage when you worked on the computer, mains power irregularities, et cetera.
PCs in hot climates are more likely to die early, though the difference isn't generally large unless the ambient temperature's really high. You seem to have had an unusually bad run. Generally speaking, well-ventilated PCs breathing air that's cooler than around 35°C do fine.
