Heat sink grease

I see a whole bunch of charts on the net, which show the freezing temperature of various glycol-and-water mixtures.

For ethylene glycol, there's a very clear eutectic effect taking place. The freezing point decreases from 0C to -50C, as the concentration of EG in in the mix increases from 0% to about 65%. At ratios above that (more than 65% EG, less than 35% water), the mixture freezes at increasingly _higher_ temperatures. The freezing point of

100% ethylene glycol is shown as being around -20C, or not much "below zero" Farenheit.

The curve for propylene glycol is very different. A mixture of PG and water freezes at a few degress higher than an equivalent EG/water mix, up to that magic 65% ratio... but the freezing point continues to drop (slowly) all the way up to 100% PG. There's no visible eutectic effect that I can see. However, the rate-of-decrease is quite low above about 70% - going all the way up to 100% PG gains you only a couple of degrees of increased freeze protection.

So, the question of "what's the right mix of antifreeze and water, for the best freeze protection" depends very much on *WHICH* glycol is in the antifreeze. I suspect a mixture of the two may have a more complex behavior.

If you're starting with a 100% (or close to it) ethylene glycol antifreeze, then the advice to stick to 50:50 looks very good, as far as freeze protection is concerned. If you go up to straight 100% glycol, you're good down to around -10F or -20C, but below that, the glycol *will* freeze. You'd save money and have significantly better freeze protection with 60:40 or 70:30.

If you're using propylene glycol, loading up to 80% or more won't hurt your freeze protection... but it's still not as good protection as you'd get with a 60:40 or 70:30 EG mixture.

Keeping it absolutely dry is going to be difficult in practice. It's going to absorb at least some moisture via the reserve tank, I imagine.

Well, it's called thermal grease, silicon grease, high viscosity semi-solid, or maybe magic grease. While not a lubricating petrochemical compound, it is a type of grease.

It also has to have a grain size small enough to fit in the crevices.

Actually, it's worse than that. When heated, the viscosity of commodity silicon grease is lowered, making it flow easier. It also expands when heated. If I built a sandwich with a spring loaded heat sink, some thermal goo, and a CPU, the expansion and easier flow will cause the thermal goo to sloooowly ooooooze out of the sandwich. The spring tension (as found in some CPU coolers) will help prevent air gaps and thermal goo losses. Setting the spring pressure correctly is tricky. If too little, the assembly will eventually rattle. If too much pressure, and if too much thermal goo was added, it could potentially make a mess. Fortunately, as the thermal goo slowly ooooozes out from the sandwich with every thermal cycle, the gap between the heatsink and the CPU slowly decreases, causing an improvement in heat sink performance. That's why Arctic Silver and others mention that you should see lower temperatures after the thermal goo has had time to "break in" [1].

Yep. Actually, energy distribution is by the inverse square of the distance, but at tiny distances, it might as well be linear. Having too large a grain size will certainly ruin the thermal conductivity but not because they don't fit in the cracks. It's because large particles offer fewer points of contact between adjacent particles than smaller (nano) particles: Particle Radius Thermal Conductivity 50 micro meters 0.8 W/mK 1 micro meter 1.1 W/mK 0.003 micro meters 2.4 W/mK I don't have numbers handy for the grain size used in commodity thermal goo.

[1] Due to the unique shape and sizes of the particles in Arctic Silver 5's conductive matrix, it will take a up to 200 hours and several thermal cycles to achieve maximum particle to particle thermal conduction and for the heatsink to CPU interface to reach maximum conductivity. (This period will be longer in a system without a fan on the heatsink or with a low speed fan on the heatsink.) On systems measuring actual internal core temperatures via the CPU's internal diode, the measured temperature will often drop 2C to 5C over this "break-in" period. This break-in will occur during the normal use of the computer as long as the computer is turned off from time to time and the interface is allowed to cool to room temperature. Once the break-in is complete, the computer can be left on if desired.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

or 39 micro inches (one too many zeros).

Arctic Silver 5 uses: Average Particle Size:

But they sure would be spendy. I don't even wanna remember what I paid for my ceramic gage blocks. Eric

In my opinion you should never 'squeeze' it tighter than the clip / fastener is going to hold it. Otherwise you squeeze too much out and are left with voids.

No such luck. They're usually where there was little contact. Metal-to-metal tends to stay that way duriung thermal cycling as long as pressure is constant.

I've done a lot of work on laptops, a lot of them nasty to get into so it's best to do it right first time. I've also lapped more than my share of CPU 'spreaders' (IHSs) and (desktop machine) heatsink bases.

--
Shaun. 

"Humans will have advanced a long, long way when religious belief has a cozy  
little classification in the DSM*." 
David Melville (in r.a.s.f1) 
(*Diagnostic and Statistical Manual of Mental Disorders) 

>> "Thermal expansion in an extrusion isn't uniform and 
> you will see bending in the heatsink. " 
> 
> Mass production. I am surprised things work as well as they do. but 
> all the mid fi audio amps are junk. Of course they want things to 
> last two days past the warranty. One day is pushing it a bit too 
> much.

ner is going to hold it. Otherwise you squeeze too much out and are left wi th voids. "

Disagree. What seem like voids are either the points of actual metal to met al contact or so close that you can't see the compound. Remember I said a b ead or daub, if it is spread it is almost for sure not even and that will c ause actual voids where it traps the air at the low spots of the compound. Some of those clips are too weak to overcome the viscosity of the compound, and some of that aluminum stuff they make the heatsinks out of is very sof t and the screws might strip. The main thing is not to let it come up or mo ve. If it is done right they are usually stuck pretty good though.

's best to do it right first time."

Agreed, with anything. It's been said "How come there is never enough time to do it right but always enough time to to it again". I have even told peo ple now and then "Hell no I don't want to do it right, I just don't want to do it again". I also don't want the liability, if the part fails guess wha t ?

'spreaders' (IHSs) and (desktop machine) heatsink bases. "

I haven't done that many processors in PCs, I just follow instructions. I d id much work in power supplies and amps, unfortunately lapping is simply im possible on most. I also agree with Jeff Leiberman's caveat about making th e surface concave. If you do that it could be worse than not lapping at all . I only lapped in some output ICs once actually. I was extremely low on co mpound so I thinned out what was leftover and lapped with it. The ICs requi red no insulator being a totally plastic package. At the end of the lapping I had them in position and moved no more so that if there were any larger particles in there they would stay put, embedded and not keep the surfaces away. The only reason I was even able was because I was retrofitting a pair of LM3886 in place of a big STK and made a clip that went from one mountin g screw to the other. I had plenty of room. (I would have put i pre outs an d had the guy use an external power amp but the damn tone controls were in the global feedback loop)

Luckily I don't have to do much of that anymore or I would have a caulking gun full of compound like this one place I worked. Makes it easier to apply and I would never run out. They did but it took years even with a bunch of techs working.

I've been hearing of laptops with mo fan, and of course smartphones don't h ave them. Will it eventually get to the point where they don't need heatsin ks ? Get an efficient enough switch and it could happen. I have already see n power switching transistors with no heat sink except the tab itself.

At 0 Vcesat, 0 Vceo, 0 Vbe, 0 tOFF and 0 tON there is no heat to sink.

Argh. It's Jeff Liebermann. Spell my name correctly or I shall call upon the gods for divine retaliation. Heed this first and last warning lest the wrath of the gods fall upon thy house, when darkness shall envelope thy LCD monitor, and a plague of bugs infest the software by which you have taken my name in vain.

Most of my Chromebooks do not have a fan. Heat is conducted from the CPU to either a metal case or the metal shield on the bottom of the keyboard.

Methinks not. For every decrease in CPU power consumption, there is an equal and opposite increase in clock speeds. The result is that the maximum power remains about the same or slightly lower. However, when the CPU is at idle and not active, the power consumption is much lower.

There's also no signal. Might be an IGBT FET. Very low Vce(sat) at high currents. However, it's not zero, so there's still some heat that needs to removed.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

n the gods for divine retaliation. "

So you're the pagan sorcerer responsible for my internet outage of a few ho urs today.

to either a metal case or the metal..."

Something that is not plastic ? Isn't there some law against that ?

an equal and opposite increase in clock speeds."

I stopped all attempts at overclocking in the Pentium 2 days, but I grasp t he concept. However, from what I have gleaned the processor speed is not th e end of the world usually. In a PC for example the RAM is slower, the HD s lower than that, and so on until we get to the speed of access to data sour ces, i.e. the internet. So conceivably couldn't they just not crank the clo ck to the max and save power and generate less heat ? Or do they already do that ? My laptops have a setting, performance, balanced and battery life. Is that possibly an indirect "underclocking" control ? If not, what is it ?

high currents. However, it's not zero, so there's still some heat that needs to removed. "

The current should not be that high, all that current is cumulative right ? Seems like the main issue is charging and discharging the input capacitanc e. As such,lower clock speeds should be quite effective. If they could get it down to the point where no heatsink is needed at all, wouldn't there be enough advantage in cost to justify a slightly lower clock speed ? A phone accessing the internet for example, how much does that clock speed really m ean then ? I am not being sarcastic here, that is a valid question, (I thin k) how processor intensive is all this ?

Maybe there is advertising value. I could see the yuppies in the cellphone store looking at specs and saying "LOOK, this one has a higher clock speed !".

Or are these the issues that keep the engineers up at night ?

I think the screw down components can warp under torqueing, but flat seems good to try and achieve. I noticed some power modules were not perfectly flat, and I started to sand them on a flat precision table. They were pretty far off from flat.

Greg

the concept. However, from what I have gleaned the processor speed is not the end of the world usually. In a PC for example the RAM is slower, the HD slower than that, and so on until we get to the speed of access to data so urces, i.e. the internet.

OTOH the CPU does a lot more than the others. Sometimes cpu is the limiting factor, sometimes not. IMLE with clocking it did make a significant differ ence. I don't bother clocking now, but once it was a deal maker.

power and generate less heat ? Or do they already do that ?

that's done nearly all the time. It's one reason why most CPUs are clockabl e.

t possibly an indirect "underclocking" control ? If not, what is it ?

probably. It also affects HDD power down settings on HDD equipped laptops.

? Seems like the main issue is charging and discharging the input capacita nce. As such,lower clock speeds should be quite effective.

There's capacitance everywhere that gets charged & discharged each time it changes between 0 & 1. There's also that conduction overlap time when somet hing changes state.

, wouldn't there be enough advantage in cost to justify a slightly lower cl ock speed ?

you need a huge speed reduction to go heatsinkless.

d really mean then ? I am not being sarcastic here, that is a valid questio n, (I think) how processor intensive is all this ?

Modern net browsers are real cpu hogs

e store looking at specs and saying "LOOK, this one has a higher clock spee d !".

As more & more gets squeezed into a cpu, power use per flop has to fall dra matically. There is no other way. The future lies with very low power gates relative to today.

NT

Yes, almost certainly it is just that (and perhaps more as well).

Most modern CPUs/motherboards/chipsets support multiple CPU clock rates - the CPU's own internal clock signal is created via a PLL/multiplier system, based on a slower fixed-rate crystal clock. The multiplier system is under processor control, so it's possible for the CPU to switch speeds "on the fly" (with a momentary pause while the PLL re-locks). I believe this is usually managed via the ACPI layer in the BIOS/UEFI.

At the higher (user-visible) layer, this is usually set up via a performance setting of the sort you mention. The setting then controls a set of policies managed by the operating system, which specify when to change CPU clock speeds (and sometimes voltages as well) based on your usage patterns.

"Battery life" would probably lock the speed at the lowest supported value, or at least to range of the slower values. "Performance" might lock it to the highest speed at all times. "Balanced" would either be a fixed speed in the middle of the range, or a dynamic system which increases the CPU speed in increments when the CPU is mostly busy, and reduces it when the CPU is idle more than a certain fraction of the time.

On the Linux laptop I use, I have a choice of several such dynamic policies... some are more aggressive about increasing CPU speed, some are more conservative. The CPU speed can be varied over a range of about 2:1, on a per-core basis (and a core which is currently sitting idle isn't using much current at any clock speed, although consumption is less at the lower clock speeds even when idle).

Depends what you're doing. Just downloading a file is probably not CPU-expensive. Rendering a web page full of fancy animated graphics and video, considerably more so. Doing full-screen video may require little from the CPU, but may push the GPU quite hard (MPEG-4 or similar video decoding).

Phones _tend_ to be designed to optimize battery life, as this is a key selling point... and so they'll be somewhat more conservative about speeding up their CPUs.

And, phones do have heatsinks. They're called "hands" :-)

The CPU chip's Front-Side Bus has been the main choke point for well over a decade. The CPUs already execute 40 instructions in the time it takes to fetch one, so faster internal clock speeds have limited impact.

This is why the RISC/CISC wars of the 90s died out. It's all about bandwidth, not MIPS.

Nope. They might be warped on arrival, but are impossible to straighten without sanding or milling.

Back in about 1976, we were having flatness problems with the transistors used in a linear power supply. There was a large extruded aluminum heat sink, and either 4 or 8 2N3055 transistors in TO3 packages. The heat sink was milled and quite flat, but the TO3 packages were warped from what I would guess was a worn stamping die. Some clever person in production decided that the cases could be straightened most easily by simply tightening the 6-32 screws holding the devices to the heat sink. After crushing the nylon shoulder washer, he torqued the hell out of the screw until the head broke off, and then gave up. The uneven squashing of the silpad insulator showed that the case was still warped.

At that point, someone in production decided that I needed some unpaid overtime. It was now my project, errr... headache. I put together a dial indicator and verified that no amount of pressure from 6-32 mounting screws is going to bend the TO3 base when the spot welded lid was acting as a stiffener. Using a strain gauge, a bar of steel, and brute force, I determined that the best I can do with trying to straighten the TO3 packages was to bend ears near the mounting holes, making the flatness problem even worse. I fixed a few packages by milling the bottom of the package, but the cost was too high. Cheaper just to buy new transistors from a different vendor that weren't warped.

The only good thing to come out of this waste of time was a fixture that I threw together to measure the flatness of the TO3 and later RF power transistor packages. It saved having the same problem repeat itself in later years.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558

I've pressed very hard on some components, watched the compound come out all around then, on relasing the pressure to 'fastener pressure' have seen some of the ooze suck back in and sometmes air suck back in. Granted these weren't on ideal surfaces but since then, on those types of interfaces I let the fastener do the squeezing.

I have variously sized bits of glass that I use as backing for the wet'n'dry sandpaper that I use to ensure relative flatness. The glass isn't as flat as gauge blocks but it's sufficient.

For laptops, desktops and critcal ICs I use Arctic Silver. I bought a 'PC builders' size syringe of it. For less critical stuff I use either Electrolube HTC or, for 'cheap jobs' that need a lot of it and have large contact areas / raditaors (mostly aluminium LED PCBs to heatsinks), stuff I get from AliExpress.

--
Shaun. 

"Humans will have advanced a long, long way when religious belief has a cozy  
little classification in the DSM*." 
David Melville (in r.a.s.f1) 
(*Diagnostic and Statistical Manual of Mental Disorders) 

> I've been hearing of laptops with mo fan, and of course smartphones 
> don't have them. Will it eventually get to the point where they don't 
> need heatsinks ? Get an efficient enough switch and it could happen. 
> I have already seen power switching transistors with no heat sink 
> except the tab itself. 
> 
> At 0 Vcesat, 0 Vceo, 0 Vbe, 0 tOFF and 0 tON there is no heat to sink.

TMI.

--
Shaun. 

"Humans will have advanced a long, long way when religious belief has a cozy  
little classification in the DSM*." 
David Melville (in r.a.s.f1) 
(*Diagnostic and Statistical Manual of Mental Disorders)

The reason the glass is so flat is because it is floated on molten tin in order to flatten it. So it should match the curvature of the earth. Though not flat enough to use as a surface plate for much of the inspection work I do it is still very flat and plenty good enough to use with wetordry paper to flatten stuff like the sealing surfaces for air compressor reed valves. Eric

Interesting. I remember reading that ocean water is higher near underwater mountain tops due to increased gravitational attraction and measurable from satellites, I wonder if tanks could be designed with non-flat bottoms to counteract what you mentioned.

underwater mountain tops due to increased gravitational attraction and measurable from satellites,"

The would be one hell of a dense mountain to do that, I would think it more likely to be because of a decreased gravitational pull being forced away f rom the center of mass of the planet, due to its inverse square relationshi p. But that is not my field of expertise. (is anything ? the more I learn t he less I know, if it weren't for learning from mistakes I would be a babbl ing idiot - NO COMMENTS FROM THE PEANUT GALLERY HERE !)

hat you mentioned"

Pretty sure that would not work because gravity is the leveling force. As s uch the shape of the bottom should not matter.

Perhaps at a high altitude with a very large mass (dense, not voluminous) p laced under the center of the tank it could be compensated. But then that m akes splitting hairs look like hitting the broad side of a barn with a plan et.

Actually, maybe a half spherical or hyperbolic shaped tank bottom would work. However that would take a hell of alot of tin.