|
Subject: say what?
Name:
 ludicrous
Date: 3/28/2003 2:39:44 AM (GMT-7)
IP Address: 138.67.74.229
In Reply to: Re: Warning put the drill down & do not try this @ home posted by
Dan Druff
Message:
Okay, I think we got off onto a bad start on this thread, but before you go lecturing people on physics, you may wish to look some of it up yourself.
Here’s the up-front deal: I am a recent engineering graduate with at least twenty-seven credit hours of course material (not including laboratory work) directly relevant to the material in this post. However I am not an Acme Certified Expert, nor do I play one on TV. If you think you can demonstrate with evidence – by which I mean actual evidence, and no more of this dreamy-mouthed verbal diarrhea about dictatorships and oppression – an area where I screwed up on what I was talking about, call me on it. If you can make a convincing scientific case, I will readily admit the error. If you wish to end this early by whipping out your doctoral thesis in applied fluid mechanics from an accredited university, and maybe wave around some material on heatsink design theory that you recently published in a peer-reviewed journal, I will accept that as an appropriate talisman and promptly disappear to update my clearly inadequate research.
If you’re trying to paraphrase your high-school physics textbook from distant memory, in spite of the fact that you spent most of the class staring at that cute Shiela Rittenhall and wondering if she had a prom date yet, consider the following essay to be MY talisman, waved at you:
In greatest simplification (i.e. lots of idealizations being made), forced-air cooling can be modeled as a function of just two things: Air volume, and surface area of the device being cooled. Theoretically, for a given amount of heat production, as surface area approaches zero, an infinite forced-air volume is required; conversely, as the surface area approaches infinity, required forced-air volume approaches zero – the device will cool by radiation and normal air convection.
To be blunt, this is the basic principle of what is going on here – and this much, at least, is not up for you to debate. If a device with relatively low surface area is producing a lot of heat, the best way to cool it with air is to install a heatsink, plain and simple. The reason the heatsink has a modest fan and all those fins is so that even a top-end Athlon (60-70W+/-) need not be cooled by a fan borrowed from the aerospace industry, or by a heatsink containing the same material mass as a 454 V8 short-block.
The reason heatsinks have gotten larger while fans have stayed loosely the same (not really – take a look at a 486 cooler, and a P4 cooler, and tell me the fan is the same. But I digress) is that processors today produce a LOT of heat, yet there are practical limits on fan sizes and speeds – in part because Intel sets standardized noise limits which OEMS tend to follow. Enthusiast types may not mind a little extra noise in exchange for a cooler system, but many users do, especially in a quiet office environment.
It gets even more complex. Heat travels through metals a lot like high-current electricity travels through wires – somewhat slower of course, but the principles are similar. Very thin metal sheets or strips do not transfer heat as well as thicker ones (within limits – overkill is also possible), and relatively good electrical conductors like aluminum, copper, and silver conduct heat better than poorer ones, like iron. And, just like two pieces of wire pressed tightly together will short electricity with relatively little resistance, two pieces of metal pressed tightly together will transfer heat with relatively little resistance.
Now, to be blunt a second time: The above reads like a laundry list of what is flawed in your reasoning throughout this violently surreal discussion thread. Allow me to justify that claim:
1. The cartridge heatplate or CPU heatspreader, whichever, constitutes a fairly thin piece of metal for our purposes here, hence is not very efficient when acting alone. Yes, it will get hot, but if you try cooling your overclocked cartridge PII by blowing air directly on the heatplate from a couple 60mm fans instead of using your HSF, after a couple minutes you will noticed that the outer edges of the plate are relatively cool while the area at the core is unnaturally hot. Welcome to flavor country. A heatplate, or heatspreader, was not designed to be operated alone. It protects the (still-hidden) CPU core from damage, and ensures a good contact between the die and the cooling solution, but for the actual cooling part it was designed to be mated against the thicker base of a heatsink, which brings us to –
2. A heatsink base is not inefficient if it is properly mounted to the CPU – i.e., tightly, and with only enough paste to fill very tiny scratches and dings that naturally occur in the interface between the heatsink and the CPU. The combination of the heatplate and heatsink together makes for a very efficient heat spreading device. Of course, if that’s all you bother to do, the assembly gets very hot and will not cool properly even if you blow air on it, which brings us to –
3. Surface area – this is why heatsinks have thin fins that will cool very rapidly when you blow air over them. You use a thick base to spread the heat out away from the core, then you use thin fins to wick the air off of that base. The surface area is the important factor; a PII cartridge’s heatplate, even if we ignore its thinness problem, still has just 2.5 x 6 = 15in2 surface area or thereabouts. A simplified 2.5 x 3in heatsink with eight vertical fins traversing the shorter dimension has 2.5 x 1in per side of each fin, so even if we ignore the small cooling effects from air blowing down to the base, that’s still 40in2. Advantage: Heatsink.
Now, regarding your heatsink modifications. First, by drilling the hole, you have created a section of the heatsink where there is ONLY the CPU’s heatplate to spread heat away from the die, rather than that nice, big, more efficient heat sink base. Strike one. Next, you have created a relatively narrow-diameter air pocket with no outlet. You will not “blow air down to the die” unless you are using a very high-velocity source, like a compressor. Instead, you will probably create a fairly calm hot air pocket in that hole. Strike two. You perform this modification on the heatsink for a faster PIII, P4, Athlon, or Duron…well, you know what they say after the third strike. Hope you saved that letter to Grandma fairly often, because you’re going to be locking up more often than a night watchman.
You claim your mod on the cartridge PII “worked;” I believe you. After all, it was locking up before and now it’s not, right? Unfortunately, you have tried to ascribe this success to an improvement in the heatsink’s efficiency from removing that excess heatsink material and blowing air down to the die. Everything I know of the science of heatsink operation says your explanation is poppycock. This leads me to ask, why did this guy's project work when he effectively has screwed up his heatsink? There are several reasonable possibilities:
A. Either the heatsink, or the CPU heatplate, had a small imperfection that happened to be centered in the spot where you drilled your hole, and this was preventing the heatsink from firmly mating to the heatplate. By drilling the hole you inadvertently fixed the problem, and while the heatsink’s efficiency is now marginally lower, it’s still a lot better than having a micro-thin air pocket between part of the plate and the heatsink.
B. You had a small piece of dirt between the sink and the heatplate, or the heatsink was not mounted very tightly. In this case, you inadvertently corrected the problem when you reinstalled the heatsink after the Black & Decker Bloodbath, with the same results as case (a).
C. Your heatsink compound had dried out with age, and dry heatsink compound is not very efficient. Same basic result as (b).
D. Some combination of (a) through (c).
Yeek, bro...this is like banging my head against the wall, but without the benefit of finally blacking out.
|