New alpha prototype Swiftykiller

[Bender]

This is an alpha prototype called the Swiftykiller. No I don't hang out at places besides O/C I just spy on them.
http://hardocp.com/new_img_01/july/swiftykiller.jpg

[William]

thats not even made out of copper. That doesn't make any sense.

[Colin]

Look at the size, it extends to the four mounting holes on the motherboard. If it's from Alpha, it makes sense.

[Plat]

william

look again. it has a copper base. look right below the cpu. it is based on the same principle as the swiftech, use copper where the hs touches the core and aluminum pins.

[Bender]

Correct me if I am wrong but doesn't aluminum dissipate heat better than copper, but copper conducts heat better. It makes total sense since the copper base conducts the heat, spreads it out into the alu fins to lose the heat faster than copper. I am sure Alpha knows what they are doing. They have all copper heatsinks for other applications so I know they aren't afraid of the stuff.

[klosters64a]

A very knowledgeable chap, I believe he was a retired Navy engineering type, wrote an article for Overclockers.com's Front Page some months(maybe a year?)ago. He said that copper indeed radiates heat better than aluminum does. The catch is that VERY little of a heatsink's heat is radiated. Most of a heatsink's thermal energy is "stripped" off of it by the forced air(don't mean to leave water cooling out of it, but the article only addressed air cooling--if I remember right) that blasts through the fins or pins of the HS.

With this in mind, he wrote that copper has no practical advantage over aluminum, at least as far as radiation goes. Personally, copper has a practical dis-advantage when compared to aluminum. It's too danged heavy! And it oxidizes. Sure, aluminum does too. But aluminum oxide has the magic quality of stopping. Then the rest of the metal is protected from further oxidation by the thin layer of oxidised aluminum.
AFAIK, copper keeps oxidising until only green powder is left.

I'd love to see a new Alpha designed specifically for Socket A. The folks at Alpha really do their R&D meticulously. The Millennium Glaciator weighs about one and three quarter pounds. That weight scares me, no lie.

[YahooDave]

That really looks like copper paint on aluminum. Otherwise that has to be a VERY good job of inserting that piece of copper into a piece of aluminum.

[Colin]

The copper insert looks like the one in my PEP66. It's not paint.

Edit: On second thought, look at the recessed corners. Where is the copper?

Edit again: I blew the pic up in Photoshop, the copper is there in the recessed corners. I stand on my original statement, it's not paint.

[Richard]

Design is king. A well designed aluminum heatsink will outperform a poorly designed copper heatsink. However, a well designed copper heatsink will vanquish the inferior aluminum and copper aluminum hybrid.

A year ago there were very few "high quality" designs. We are now seeing the limitations of design and material. (Thus the all too common, and annoying, Delta 38.)

The fact of the matter is the socket A CPUs are emitting a ridiculous amount of heat in a very concentrated area. It won't be long before even the absolute best designed air-cooled heatsink will be unable to cope.



Bender (Jul 19, 2001 08:38 p.m.):
Correct me if I am wrong but doesn't aluminum dissipate heat better than copper, but copper conducts heat better. It makes total sense since the copper base conducts the heat, spreads it out into the alu fins to lose the heat faster than copper. I am sure Alpha knows what they are doing. They have all copper heatsinks for other applications so I know they aren't afraid of the stuff.

[William]

a copper aluminum hybrid is key. Copper will distrubte the heat better, and aluminum will transfer it to the air better. I don't think a lot of the heatsinks using a delta on the socket are bad designs, the heat output requires either a monster hsf(Glaciator) or a delta to overclock it and keep it in a good temp range.

[Richard]

Baloney.

Aluminum doesn't "release" heat better. (If that was the case aluminum waterblocks would perform better than copper waterblocks.) Aluminum, is cheaper and lighter (that's why it is more common in heatsinks) but it doesn't have magical heat releasing properties. The all copper sink may seem hotter to the touch, but that is because it is conducting MORE heat from the core. Pair up the cupric based sink with an adequate fan or water supply and the aluminum/copper hybird will NOT perform as well.

Plus, I never said that the monster delta fans were in place due to poor design. Merely that they are in place to maximize design. Case in point. A water cooled aluminum block will outperform an air cooled copper sink. However, a water cooled copper sink will outperfom an aluminum sink with the biggest fan you can find.



William (Jul 20, 2001 03:18 a.m.):
a copper aluminum hybrid is key. Copper will distrubte the heat better, and aluminum will transfer it to the air better. I don't think a lot of the heatsinks using a delta on the socket are bad designs, the heat output requires either a monster hsf(Glaciator) or a delta to overclock it and keep it in a good temp range.

[Richard]

I decide to do a search on google.com Just to make sure I wasn't putting both feet in my mouth.

I typed the following into google...

does aluminum radiate heat better than copper

What's the first article that popped up?

Why, one from overclockers.com, of course.
"Cooling - An In-Depth Look" Mike Larsen - 9/6/00

Give it a read.
http://www.overclockers.com/articles223/index02.asp

Just as an FYI. Radiant cooling isn't the idea in relation to the manufacture and use of heatsinks. Convection cooling is. Thus, the fan in heatsink/fan.

[Hoot]

Great Article. I just bookmarked it for future insertion into posts. Thanks Richard, for doing the legwork.

Hoot

[CalCoolage]

>There is no such physical phenomenon as to how well a material 'gives up heat'. This is an
>internet-overclocking myth that has propagated for far too long and will now be laid to rest!
So why do Swiftech and Alpha use copper bases and aluminum fins?
Geez, those poor idiots at Swiftech and Alpha just don't know how to design a proper heatsink. I feel sorry for them. Yeah they tried to save money on those poorly crafted Swiftechs. Think of how much money they saved by sticking in aluminum pins instead of copper. Maybe as much as a nickel. I sure wouldn't want to pay five cents more for a Swiftech.


>So why don't the current copper heatsinks far outperform (all tests I have seen show that
>copper heatsinks do outperform aluminum, just not by much) their aluminum counterparts?
>In my opinion it is because of a few things; namely poor design, multiple piece heatsinks
> (remember thermal contact resistance!), impure copper, and difficulty of producing a
>copper heatsink in the desired form.

These are feeble excuses to cover the fact that the theory presented doesn't account for real life performance.

>Whenever there are multiple layers through which heat must conduct, there is what is
> called a contact thermal resistance. This is defined as such that where the two
>solids meet (i.e. the CPU slug and the heatsink) there will be a thermal contact
> resistance that results in an immediate temperature drop across the joint.

Assuming this is something besides poor contact:
What is the thermal contact resistance between aluminum and copper? Are there tables of these resistances? Why is there no thermal contact resistance.between air and copper?

[inertia]

CalCoolage (Jul 20, 2001 09:08 a.m.):
So why do Swiftech and Alpha use copper bases and aluminum fins?

Because by the time the heat gets out to the fins instead of being concentrated at a single point over the core, the thermal conductivity becomes less significant. I'm not saying it's not significant, but the weight becomes a more significant factor by then. Solution = use copper where heat is most concentrated and thermal conductivity is crucial, use aluminum where thermal conductivity is important, but weight is more so.

CalCoolage (Jul 20, 2001 09:08 a.m.):
Assuming this is something besides poor contact:
What is the thermal contact resistance between aluminum and copper? Are there tables of these resistances? Why is there no thermal contact resistance.between air and copper?

Yes, there are tables for measurements taken at standard temp and pressure (STP = 25C 750mm Hg). Why do you assume there's no thermal contact resistance between air and copper? This question demonstrates how much knowlege you are lacking. I happen to have a degree in chemistry (entirely not necessary in this application, but it serves as qualification methinks) and can tell you how to determine the the contact resistance: 1) Measure the temp of the exhausted air 2) Measure the temp of the heatsink 3) subtract air temp from the heatsink temp and viola you've got the contact resistance for the specific conditions of your system (pressure is a function of the fan you're using.) This'll work for the aluminum pins and the copper base as well. When doing this, you should get as close to the junction of the two materials as possible. You'll find that while the air/copper or air/aluminum temps are significant, the copper/alum is minimal.

Actually, a thermocouple temp probe works on this same principle only it's with electrical conductance and not thermal. I could go into it, but I won't.

[cjtune]

inertia

CalCoolage (Jul 20, 2001 09:08 a.m.):
Assuming this is something besides poor contact:
What is the thermal contact resistance between aluminum and copper? Are there tables of these resistances? Why is there no thermal contact resistance.between air and copper?

Yes, there are tables for measurements taken at standard temp and pressure (STP = 25C 750mm Hg). Why do you assume there's no thermal contact resistance between air and copper? This question demonstrates how much knowlege you are lacking. I happen to have a degree in chemistry (entirely not necessary in this application, but it serves as qualification methinks) and can tell you how to determine the the contact resistance: 1) Measure the temp of the exhausted air 2) Measure the temp of the heatsink 3) subtract air temp from the heatsink temp and viola you've got the contact resistance for the specific conditions of your system (pressure is a function of the fan you're using.) This'll work for the aluminum pins and the copper base as well. When doing this, you should get as close to the junction of the two materials as possible. You'll find that while the air/copper or air/aluminum temps are significant, the copper/alum is minimal.


Once you measure the temperature of air, you are also including the CONVECTIVE thermal resistance of the heatsink -alongside the thermal interface resistance and the conductive thermal resistance between the mating interface to the surface exposed to airflow. The only way would be to measure temperature between two standard sized pieces of (insulated at sides but not at the ends of the heat source and heat sink) copper and aluminium so you can calc. and substract their conductive thermal resistance as well as being able to *assume* a uniform heat flux (Q/m2). Then you can get a more accurate value of contact resistance per area. For even more accurate results, you gotta include the contact resistance of the thermocouples at the ends.

A can see this thread has been influenced by the thread 'Copper and Aluminium' by JigPu.

[Plat]

this copper and aluminum thing.

how can a copper base w/ aluminum pins/fins ever be better than an all copper hs other than the weight issue?

it seems to me that the zalmans have the best idea, because the base and fins are one continuous strip of copper. there is no way that a marginal advantage (if such an advantage exists) that aluminum has of releasing heat can make up for the loss of efficiency due to the joining of copper to aluminum.

[CalCoolage]

inertia (Jul 20, 2001 10:03 a.m.):
Why do you assume there's no thermal contact resistance between air and copper? .

I didn't assume it. I was asking why to the guy that wrote the article I quoted from, who made that claim, IMO, when he said ALL things transfer their heat EQUALLY to air. Later he said there was such a thing as thermal "contact resistance" for differing substances. However if it exists for air-aluminum, it is the same thing as "how well aluminum transfers its heat to air". If there are tables, as you say there are, it would give the values for copper-air and aluminum-air. But if heat transfer depends ONLY on the thermal conductivity, why would there even be a "contact resistance table", since there would be no such thing as contact resistance in this case.

>This question demonstrates how much knowlege you are lacking.
No doubt. Thankfully you are not so lacking. So what are the numbers in this table for aluminum-air and copper-air?

>Solution = use copper where heat is most concentrated and thermal conductivity is
>crucial, use aluminum where thermal conductivity is important, but weight is more so.

There are sinks that use copper for both, so why do the premier heat sink designers suddenly get concerned by the weight? Could it possibly be that they really designed their sinks for maximum thermal performance? NO! IMPOSSIBLE! And there is no need for concern about weight in the Swiftech and the new Alpha, because they use bolts to attach the sink, instead of the little plastic nubs.

A completely copper sink would be well under 3 pounds.(The all-copper Thermalright SK6 is 330 grams which is 12 ounces.) 3 pounds is not much to carry for the bolts. The bolts could carry 10 or 20, or even 200. I'm not sure 3 pounds would even be excessive for the plastic nubs considering that the clip pressure is something like 25 pounds, dwarfing the weight of the sink.

Quote from Swiftech:
"We conducted crash tests by dropping repeatedly a computer equipped with the MC462 from the roof of a one story building, with no damage to the processor."

[asmodean]

I think people should use a weight spreader behind their mobos, if using the 4 boltholes. Just to minimize the possibility of damage...

I'm not getting in to the debate over pure copper vs copper/aluminum hybrids, better to wait till they release that HSF and judge it after few reviews.

Daniel~ might move this post soon, lol.

[most deaf]

bout time they made another heatsink

[SleepyKat]

I may be mistaken but isn't silver a better thermal conductor than copper or aluminum?

[Bender]

Yes a silver base would be a little better than copper. Silver would be very soft and probably be a much better choice for delicate amd cores. Silver would also conform a tiny bit to the core creating a better contact area.

[asmodean]

Only problem with silver is the price... For the end-user, I don't know what kind of problems the manufacturers would run in to.

I'd rather have a diamond waterblock, tho. Well, if I could afford a diamond waterblock, I could afford a diamond mercuryblock too.

[Bender]

Silver is expensive but not out of the question for a heatsink. Silver runs about $4-5 oz so if you are making the very best heatsink it would be worth it.

[Big Mike]

I'd suspect that silver's ROI factor <return on investment> is way too low for heatsink building. Pulling this out of my arse but say a swifty costs 80 bucks, would you be willing to pay 110 for a silver swifty that probably would be lucky to get you another 2 degrees celcius <id say that was optimistic at best...>. I think we're all but destined to end up with watercooling or perhaps an inventive heat pipe design as our future cooling means...however with the sudden stagnance in the computer market die shrinks and isotropically pure silicones and similar inventive chip design may keep John Q. Public below the critical heat levels...us in the O/C world may find ourselves with fewer and fewer alternatives...I'm contimplating dropping my FOP38 for water...

But when it all comes down to it, I think it all has something to do with polar bear fur...the answer's in there somewhere...<I may regret bringing that subject back up...hehe>

[cjtune]

CalCoolage (Jul 22, 2001 09:33)

I didn't assume it. I was asking why to the guy that wrote the article I quoted from, who made that claim, IMO, when he said ALL things transfer their heat EQUALLY to air. Later he said there was such a thing as thermal "contact resistance" for differing substances. However if it exists for air-aluminum, it is the same thing as "how well aluminum transfers its heat to air". If there are tables, as you say there are, it would give the values for copper-air and aluminum-air.

All 'things' at steady (not varying with time) temperatures, really WILL transfer the same amount of heat, given a CONSTANT heat source (constant power). The only difference, with thermal resistances coming into play, it the deltaT that is set up. Higher conductivity materials permit such a heat transfer....at a lower deltaT. Wouldn't you want a CPU temp to be as close to ambient as possible?

Contact thermal resistance is only used for the interfaces between two SOLIDS. Copper-to-air and aluminium-to-air and whatever-to-fluid thermal resistance depends only on exposed surface geometry (eg. for area and impedence to airflow), the flowrate and properties of the fluid. The thermal properties of the metal or whatever solid is not needed as it will reach a surface temperature (common to what all other materials will reach) in order to transfer a constant amount of heat (rate of transfer of thermal energy, ie. power). The only proviso is that the phase transition point (eg. melting) of the solid is above the the current surface temperature. There in no contact resistance defined for between metal and fluids -convective thermal resistance takes its place.

Let's say, between metal and liquids, there might be micro-bubbles trapped in the surface irregularities of the metal but this should be due to manufacturing tolerances and not *just* because it is aluminium or copper or whatever. There will be no tables of contact resistances between any metal to air. A table of convective thermal resistance due to fin geometry and airflow, yes, but also not due to metal type. A table of contact resistances exist not just because solids are in mechanical contact, but also because they may be bonded chemically to each other -and solid metal particles obviously do not move about due to density differences.

[asmodean]

Just remembered reading about a synthetic diamond somewhere, might have been this forum... It's not cheap enough yet for mass-mfg, but hopefully it will be soon, because diamond is currently the best material (price aside) for heat-transfer (or heat-capacity). I'm not going to vouch for this, as usual, due to lack of more evidence.

Just did a search on google with "diamond heat transfer", 36,700 hits :) The material is CVD-diamond. Popped out a lot of links to many universities too and there seems to few books written about it too. Let's hope it'll become available soon, then we won't have to worry about heat so much and can stop worrying about shorting our chips alltogether :)

[William]

yes, that would be cool too say when someone asks what hsf you use(i get asked this ALL the time ;-) ) you can say i cool with Diamond, lol.

[joey_rjm5]

So if they can make a synthetic diamond, do you think they can make something better than diamond? That would be cool, but I guess that synthetic diamond would be adequate enough. I wonder how they do that. Doesn't it take like millions of years for the earth to make one? This is a cool subject.

[asmodean]

I quote from P1 Diamond, Inc (http://www.p1diamond.com/engine.html):


Thermal: CVD diamond has the highest known isotropic thermal conductivity, ranging from 4 W/cm/°K to over 20 W/cm/°K, depending on the manufacturing process. Diamond has an expansion coefficient of approximately 1.0 x 10-6/°C, and has excellent thermal shock resistance due to its low expansion coefficient and extreme thermal conductivity.


Sadly they are not in the business of making/selling heatsinks made out of it.

If someone here has a lot of excess money, they could contact the business and have a heatsink made out of CVD diamond.

I doubt they'll be able to make anything better than diamond at the moment. It takes huge pressure to compress carbon in to a diamond. I can only imagine what kind of pressures it would take to form a diamond out of metal and if it even would be better.

[cjtune]

asmodean (Jul 26, 2001 06:26 a.m.):
I quote from P1 Diamond, Inc (http://www.p1diamond.com/engine.html):


Thermal: CVD diamond has the highest known isotropic thermal conductivity, ranging from 4 W/cm/°K to over 20 W/cm/°K, depending on the manufacturing process. Diamond has an expansion coefficient of approximately 1.0 x 10-6/°C, and has excellent thermal shock resistance due to its low expansion coefficient and extreme thermal conductivity.


Sadly they are not in the business of making/selling heatsinks made out of it.

If someone here has a lot of excess money, they could contact the business and have a heatsink made out of CVD diamond.

I doubt they'll be able to make anything better than diamond at the moment. It takes huge pressure to compress carbon in to a diamond. I can only imagine what kind of pressures it would take to form a diamond out of metal and if it even would be better.

Coldplate? Hotplates? Bases for new diamond-copper hybrid HSFs? Will spread the heat like wildfire the all the HS fins! Looks like there's still potential practical applications for it yet. I want! I want!

[asmodean]

I doubt you can attach copper fins to CVD well enough. Better to have a full CVD hsf. Or rather, waterblock. I think it's quite expensive stuff (something like 10 Swiftechs or so, just a guess).

[Mictlan]

I might add the following:

There are 3 forms for the heat to be exchanged: by conduction, by convection and by radiation. Conduction is direct contact almost always between same phase materials (solid-solid) or with different phase materials (solid-gas) if the gas is stationary, convection is when you "carry away" the heat using different phase materials and one of it is moving and radiation only occurs at very high temperature difference. Some examples here:

The die of the processor exchande heat to the heatsink by conduction as they are side by side.

You remove the heat from the heatsink with air. (convection)

If you have some expereince with soldering, if you put you finger near the soldering tip you will be feeling the heat from it. (radiation)

When you design a heat exchanger (I do this for a living) the main heat transfer method you are using is convection as the heat transfered by conduction needs more time to be of some significance.

For the convective heat eschanger design you'll need the heat transfer coefiecient. This physical property is the resistance of a material to the heat flow. It will be easier to explain this with another example. If you have a water pipe, the heat transfer coeficient will be akin to a valve. The smaller the valve is, the harder to get a lot of water flowing. As we are talking about moving the heat from one material to another the smaller is the limiting factor. In this case, the air or water. Metals have aproximatly the same heat transfer coeficient.

For a good heatsink design you will need then:

1) A good material that is good at conductig heat from the CPU die to the heatsink body.

2) A good material thats light and cheap to be the heatsink body.

Heat Transfer is claculated using the following formula:
Q=U*A*DT where Q=heat, U= is a relationship between the two material's heat transfer coefecient, DT is the difference in tempeartures between the two materials and A is the surface area that be in contact. This area is one of the most important design challenge when constructing a HS. Given a high area you can compensate with bad heat transfer coeffients.

Now you'll be moving this heat with a fluid (water or air). the quantity you need to do this work will be given by the following formula:
Q=m*C*DT where Q is the same heat that we transfered, C is the heat capacity of the material (how much heat you need to increase the temperature of the material, here water is better than air), m is the flow of fluid you'll be needing and DT is the allowable temperature will be having across the heat exchanger. This temperature is very critical, as a rule applies here "you can cool the hot material more than the cooling media". For example, if the highest temperature you have in the surface of the HS is 45°C, the air or water can't reach more than 45°C. So, the only variable we have here is the flow, so if you increase the flow (someone mentions Delta fans) you increase the heat removed from the HS.

After all this, that's why mixed copper and alluminum HSF are becoming popular. You use a copper contact plate to channel the heat to the body of the heatsink, and you use an aluminum body to convect that heat to the air. Also aluminum is cheaper, lighter and more workable than copper.

[Bender]

Mictlan since you seem to know what your talking about how much efficiency is lost when going from a copper base to aluminum fins? Machining would have a big role in how well the one material conducts its energy to another.

[Mictlan]

I don't expect a big difference. I personally don't own a Swiftech MCX370 or a Glaciator to do some test but if you see all the test that has been done you don't loose heat transfer efficiency when you use an all copper or a mixed HS.

I'm going to write about this with a little more data (1 or 2 weeks) and send it to ED to post it in the main page.

[asmodean]

Mictlan (Jul 26, 2001 05:16 p.m.):
I'm going to write about this with a little more data (1 or 2 weeks) and send it to ED to post it in the main page.

Yay! That'll be an article NO-ONE can bypass!

[CalCoolage]

asmodean (Jul 26, 2001 10:22 a.m.):
I doubt you can attach copper fins to CVD well enough. Better to have a full CVD hsf. Or rather, waterblock. I think it's quite expensive stuff (something like 10 Swiftechs or so, just a guess).

I'll give my guestimate. In a quantity like that used in the base of current heatsinks, it
would be more like $1,000,000 or $10,000,000. I don't think anybody is claiming that
CVD is cheaper than natural diamond at this point.

Industrial use of diamond for cutting tools is made practical by using extremely low quantities. This works just fine because diamond is at least 10 times harder than what it is used to cut. (unless they use it to facet cut diamonds.) First, they use diamond that is unusable for anything else. Then they coat the surface with a fine dust one particle deep and spread it sparsely, like one particle per 1/100 th inch.

For instance, a "diamond stone" used for honing tools costs about $60 (or it did 10 years ago). A similar size aluminum oxide "stone" might be $10, but it probably contains more than a million times the amount of abrasive.

Using diamond for conducting heat requires that it be thick in cross-section.

[Greebe]

Richard, CalCoolage (et al.)
Hard anodized black (is best) Aluminum has a greater emissivity coeffecient by a few percent (that's the term/factor you didn't know about) than copper and thus will radiate heat better. Aluminum cladded copper would make a better HS than an all copper heatsink due to this.

Bare aluminum (silver finish) flat out sux tho.

Silver would be a waste because it's not that much better than copper thermally and much more reactive than copper to the elements... which is why Zalman has gold washed their 3100G's (it's not "plated") finish. That way it won't corrode, which will reduce it's emissivity over time, depending on enviromental conditions. ( and especially if you touch it!)

With silver you would still have to lap it! It will not settled in and will still cause chipping of the cpu substrate "corners".

Some specs (no hype here:)

Thermal Conductivity
Silver 4.08
Copper 3.94
Aluminum 2.18

[Greebe]

BTW if you use a flat black silicone spaypaint on a bare aluminum (silver finish) the emissivity will increase to ~upto 0.89 of a blackbody, ie 89% which is better than that of copper and as good as hard anodized black aluminum.

Untreated bare aluminum is ~0.02 ie 2% of a blackbody.

[cjtune]

Originally posted by Greebe
BTW if you use a flat black silicone spaypaint on a bare aluminum (silver finish) the emissivity will increase to ~upto 0.89 of a blackbody, ie 89% which is better than that of copper and as good as hard anodized black aluminum.

Untreated bare aluminum is ~0.02 ie 2% of a blackbody.

...or copper. The main thing here is the colour of the surface as radiation, although is a voulumetric phenomena, is for our purposes a surface phenomena. This has very little to do with the metal underneath (if @ same temperatures).

Good thing Mictlan is here! I hope he'll make it clear that heat transfer by convection is a not a function of the material used as the surfaces.

[Sohryu Asuka Langley]

the alpha goes like an original mc462 :) i just wish that alpha didnt make the ally dark and left it silver

[Mictlan]

Originally posted by cjtune


...or copper. The main thing here is the colour of the surface as radiation, although is a voulumetric phenomena, is for our purposes a surface phenomena. This has very little to do with the metal underneath (if @ same temperatures).

Good thing Mictlan is here! I hope he'll make it clear that heat transfer by convection is a not a function of the material used as the surfaces.

In Radiation, the colour of the material will change the amount of hest dissipated to the external media (in this case the air). One of the main contribution to heat radiation in the last century (20th century) was the black body law. I don't quite rember this one.....let me check some of my books and later I'll return to this one.

Opposed to this, convection is more like a surface phenomena. Lets try to explain this with an example. Imagine two bunches of balls of different size and colour. Now lets arrange these balls in two layers 3 balls depth each one. At the end will be having something like this:

.....................
.....................
.....................
OOOOOOOOO
OOOOOOOOO
OOOOOOOOO

Neat, right. Now if we began to push the topmost layer. What would happen? Some of the balls from the second layer will began to move also, and if we push a little harder some of the bottom layer of the smaller balls might move also. Still with me.....OK, now imagine that the larger balls are painted red and still wet. What will happend? The bottom layer of the smaller balls will paint red also, and as some of this balls are also moving, some paint eventually will be getting to the top most layer of balls. Right?

Now, let me change this once more, this ball are really molecules, and the paint is heat. This example can give you some idea of how do the heat is carried away from the heatsource.

As you see, if I change the bigger balls' surface there won't be a change in how the paint will reach the topmost layer of smaller balls. In the smae way, if I add another layer of molecules (paint), I'll only add another layer thru which I need to conduct the heat.

I'm still working in the article. Might finish it this weekend.

[shadowdr]

we are still missing somthing here.the hybrid will outperform the solid of either provided it is constructed correctly.
throw out cfm.it doesnt always apply because we are talking heat and dissapation,and also cooling affect.
the copper spreads heat more efficiently.we know that to be true.
dissapatin of this heat is not really what we are doing with aluminum,we are cooling the aluminun itself with forced air.
look at your refridgerator the outside coils are copper but the inside coils are aluminum.it stays colder than copper would because of the lower boiling point of the metal.
in resteraunts the cold plates are allways made of aluminum as well as pitchers for water or tea.
the best radiator you can buy still has a copper core and aluminum finning.ice trays are still made of aluminum while cooking pans have copper bases.

[Mictlan]

Originally posted by shadowdr
we are still missing somthing here.the hybrid will outperform the solid of either provided it is constructed correctly.
throw out cfm.it doesnt always apply because we are talking heat and dissapation,and also cooling affect.

The cooling effect you are mentioning and dissipation are in essence the same phenomena: heat transfer

the copper spreads heat more efficiently.we know that to be true.
dissapatin of this heat is not really what we are doing with aluminum,we are cooling the aluminun itself with forced air.
look at your refridgerator the outside coils are copper but the inside coils are aluminum.it stays colder than copper would because of the lower boiling point of the metal.
in resteraunts the cold plates are allways made of aluminum as well as pitchers for water or tea.
the best radiator you can buy still has a copper core and aluminum finning.ice trays are still made of aluminum while cooking pans have copper bases.

If you go to bigger radiator (I've done a few air cooling devices for some petrochemical plants) are made with carbon steel. Lets try the following. Get two bars, one made of aluminium and the other made of copper. The same size and shape. Heat a tip with the same heat source (the heat from your stove, for example). Tell me which one will get the other tip hot first. *****Hint*****, the cooper will get hot first. Why? Because heat travels (by conduction) quicker on copper than aluminium.

The motor by which heat travel is temperature difference. The higher, the better. That's why when you whant to use a cold plate, a thick one is better than a thin one. Imagine a hose. When you turn on the water valve, the water will pour in a cone shaped stream, with the hose in the center. Exactly that happens with the heat when it enters the cold plate. That's why you ofen see in heat exchangers a core of copper (heat will be conducted over all the core, so they get a high temperature difference) and fins of aluminium (cheap, resistant and light).

[shadowdr]

i agree,now put those pieces of metal in a fridge and see which gets cold first,aluminum.

[Richard]

Are you choosing to ignore that copper is much more dense than aluminum? If you're dealing with size alone of course the aluminum will cool faster. *It doesn't have as much heat to radiate* The difference in radiation of heat in metals is insignificant when talking about heatsinks. Thermal Conductivity plus convection cooling whether by air or water is what we're after.

An extruded thin fin copper sink will function better than a copper/aluminum hybrid. I stand by that, and the results of the umpteen million reviews have born that out.

[Mictlan]

Originally posted by shadowdr
i agree,now put those pieces of metal in a fridge and see which gets cold first,aluminum.

Yep I agree, but we are talking about a different thing here. When you use a refrigeration unit you are using convection as a heat transfer, not conduction. Even if you feel like the inside of the refrigerator is static, inside is a liquid (freon) moving.

[Mictlan]

Originally posted by Richard
An extruded thin fin copper sink will function better than a copper/aluminum hybrid. I stand by that, and the results of the umpteen million reviews have born that out.

If you check a review using an hybrid (copper-aluminium) heatsink and in the same site using the same procedure to test a pure copper heatsink, you'll se that they are not that far. The difference in they're C/W factor is around 0.01.....if you can show me that they did the same test ten times, I would agree that the copper HS is better than the hybrid. But as they do the test at the most twice, it's within statistical error this difference.

[Greebe]

Currently I'm writing up a review on a Zalman CNPS3100G... which is not setup in it's typical means with the supplied fan/bracket or fin pattern... and it's sitting here happily doing 0.12 C/W idle, 0.14C/W peak while dissapating 80.4 watts

Mictlan, yes you are correct, but when discussing a typical HS design, they aren't taking into consideration forced air cooling and thus with their flat fins are born out of yesterdays natural convection cooling type HS designs. Thus don't begin to think about surface to air contact being a factor, and not a subject I've bothered to mention.... til now.

Flat finned HS are a poor choice for they do not have the surface contact with air (typically refered to as laminar airflow) that a dimpled or waved HS has. I'm sure I could tweak another 0.01-.02 C/W out of the Zalman, but at this point I could care less.

So many times I have seen crappy commentary of the subject of HS by those who can only run their mouth and have no experience in the actual design and analysis of them. Most times it's just regular Joes on forums, but often times it's reviewers too and that's a shame. :\

I like this read... :)
http://www.electronics-cooling.com/html/2000_jan_a3.html

BTW if your wondering... I've been OCing computer for 17+years now... not a newby by any extent of the word! :)

[CalCoolage]

The thermal condutivity of copper is nearly twice that of aluminum. (397/238 = 1.67 ) So why are all-copper sinks marginally better, if at all, than similar all-aluminum sinks, and why are copper sinks with aluminum fins virtually identical?

What factor of aluminum is compensating for its inferiority in thermal conductivity?

Is thermal conductivity irrelevant?

Thermal conductivity evidently is important for steel. Steel heatsinks seemed to be very poor in the few cases I've seen one measured. So again, why are copper and aluminum sinks so close?

What factor makes water a better cooling fluid than air. One would think it is heat capacity. If so, why is heat capacity irrelevant for the sink itself?

Aluminum has a far higher specific heat than copper. But that is per unit weight, and aluminum is a third the density of copper.

specific heat = Joules/ (grams x degees)
density = grams/ cubic centimeter
specific heat x density = Joules/ (degrees x cc)

Aluminum 0.9 x 2.7 = 2.43
Copper 0.385 x 8.96 = 3.55

Per unit volume, aluminum holds 70 % of the heat versus copper, per degree, so if equal amounts of heat are removed from equal volumes, the aluminum gets cooler. If the copper goes down 10 degrees, the aluminum goes down 10/.7 = 14 degrees.

Is this relevant? Who knows. But something is undoing most of the theoretical advantage of copper.

[Mictlan]

Originally posted by Greebe
Currently I'm writing up a review on a Zalman CNPS3100G... which is not setup in it's typical means with the supplied fan/bracket or fin pattern... and it's sitting here happily doing 0.12 C/W idle, 0.14C/W peak while dissapating 80.4 watts

Mictlan, yes you are correct, but when discussing a typical HS design, they aren't taking into consideration forced air cooling and thus with their flat fins are born out of yesterdays natural convection cooling type HS designs. Thus don't begin to think about surface to air contact being a factor, and not a subject I've bothered to mention.... til now.

Flat finned HS are a poor choice for they do not have the surface contact with air (typically refered to as laminar airflow) that a dimpled or waved HS has. I'm sure I could tweak another 0.01-.02 C/W out of the Zalman, but at this point I could care less.

So many times I have seen crappy commentary of the subject of HS by those who can only run their mouth and have no experience in the actual design and analysis of them. Most times it's just regular Joes on forums, but often times it's reviewers too and that's a shame. :\

I like this read... :)
http://www.electronics-cooling.com/html/2000_jan_a3.html

BTW if your wondering... I've been OCing computer for 17+years now... not a newby by any extent of the word! :)

I ought to agree with the reading you post. We all are working with delicate components with nothing but lots of iniciative, hearsays and a welding torch :) .

Take for example the C/W factor. I you check this factor for the same HSF from 10 different hardware sites and reviews made by whoknowwho, you can find extreme difference. Take for example the Zalman you are reviewing. I don't question your capabilities, or the training you have or your experience but that very same HS is reviewes here (http://www6.tomshardware.com/cpu/01q2/010521/cooler-32.html) and the C/W is 0.45. The point is that the method of calculating C/W might be different. I have had the same experience at work when we calculate the eficiency of some steam-power systems. We take some data from our estimations and get an eficiency, but when we go an compare it with some other engineers we get very different numbers, as they take into account other factors.

The design of a heat exchanger is something akin to an art form. But there is one principle above everything else: the KISS (Keep it Simple). Specially with air cooling. Air cooling is difficult specially because air is a poor heat transfer media (that's why when you want to insulate something, its recomended that you use a thin layer of air trapped between two other materials). You need lots of cooling surface, and lots of air. That's why you force (blow) or induce (suck) air to/from the heatsink. Also, the fins (extended cooling surface) have to be well designed to help cooling.

But, you might ask, what can you design from a fin? A fin is just a sheet of metal or a pin.....So you just add some texture to the normally flat fin or pin, just like the design in the MCX370 or the pressed sheets inside some refrigaration units.

So, as you see there is no simple explanation of many of the very interesting phenomenas that happens inside your PC, which is a very complex system of electronics.

[Greebe]

Mictlan, maybe that's because I've modified the HS, built a shroud (duct incompassing/funnel... whatever:) that wont stifle the 120mm 115cfm Rotron Muffin XL AC fan I'm using (connected to a variac for speed control).

It's wired so I can exeed the fans rated spec by outputting upto 136VAC which results in ~130cfm (unrestricted... and this setup is restricted) airflow. Even with the fan set to a 58.3% speed (as many don't like the full blown SPL's big fans produce) which equates to a 12vdc fan of similar spec at 7VDC... it still has a C/W of 0.19 at idle. :)

Actually using this monster fan is "part duex" of my review... the first part uses nothing but the supplied parts with a shroud which doesn't exceed 0.23 C/W. The fan is running at 12vdc and no recirculated pre/reheated air returning to the HS.

Total system cooling is important for these results to be achieved... it's not a "slap a shoud on it and get these results" type of thing.

I do know there are theoretical problems with my testing, but it's good enough for goverment work. At least til I have what I really need to produce the best test results:)


BTW I helped ALBPM design his COP HS testbed seen on www.trainwrecker.com

Looks like Kyle Bennet has a new HS testing round up also... http://hardocp.com/reviews/cooling/roundup0601/


No comment on THG reviews

[Greebe]

Check this awesome HS's out! (shame it's not for CPU's!)... and look at them C/W ratings! :D

http://www.d6industries.com/heatsinks/hspic1.jpg


D6 Industries (http://www.d6industries.com/heatsinks/)

[Mictlan]

Originally posted by Greebe
Total system cooling is important for these results to be achieved... it's not a "slap a shoud on it and get these results" type of thing.

I do know there are theoretical problems with my testing, but it's good enough for goverment work. At least til I have what I really need to produce the best test results:)

Here we reach the main point of all this disccusion. We are not just talking about the thermal conductivity of copper vs aluminium, which is measurable. We are talking about heat exchanger design.

As in my country say, "There are lots of way to skin a cat", there are several designs that would remove a constant load of heat from one point and dump it into the atmosphere. Of all this design there are some that are plain impossible to be carried out. Others are uneconomical. Lots of them will be possible, but not so efficient. Finally, there is one or two that are the cheapest and easiest way to do the job.

Take this practical example. There are lots of HSF that perform OK, for stock CPU's. But used in our oc systems they are no match for others that have been designed to handle a lot of heat. But if you slap a high output fan to this otherwise worthless HS you get a competitive combo that can cool your CPU.

This tell us that if you increase the air availability and the speed you can increase the amount of heat you remove from the hot place to be dumped into the atmosphere. But to some this is not a feasible solution. as it produces lots of noise. So, what they will do is to move to an HSF designed to handle lots of heat and hopefully with less noise.

As for the testing of HS, I'm not saying that some hardware site are not doing a good job when they calculate they C/W factor. But, unless we all agree of how to test the thermal efficience of a given heatsink, its worthless to compare these factors. They are meaningless if compared to the others obtained using a different procedure. Take for example the C/W factor that are shown in the post above. This HS manufacturer is claiming that they're product gets to 0.15 and less. If you have checked lots of hardware sites you'll see that to reach less than 0.1 you'll need another system (like watercooling, chillers) and not air cooling. This reminds me of the propaganda that Intel is doing "I'm 500 MHz faster than AMD". We all know that CPU speed is not all. So please take all the C/W with a grain of salt, and compare the data from lots of different people to get a more accurate idea of the performance of a heatsink.

[Mictlan]

Well, seems that a fellow has posted some good stuff in AMDMB, check the front page to read this good piece of information.....

grumble, beaten again, grumble

[shadowdr]

excellent article!
ok i learned some stuff i didn't know.it seems that copper and aluminun are fairly close in alot of areas.but what makes alot of what is disscussed here pointless is the fact that were not dealing with pure metals,both copper and aluminum in haetsinks are alloys of different mixtures.copper with a content of 5 percent arsenic reduces conductivity by 15 percent.copper used for wire on the other hand is 99 % pure (at least)but to pliable to use for a heat sink.
combining economics with efficiency my idea of the most useful heatsink will be: imagine a table turned upside down consisting of pure copper(thin)with a leg added to center and radiating outward with more legs to edges.add a piece of aluminum to the base for strength with holes for the legs or rising pins with solder.add very thin aluminum fins to the pins in a side to side alignment.add another piece of aluminum to the sides and top.side mount the fan ducting outside air in and duct close to outside of case on the other side taking some of the inner air with it.
i wish i could build it but i will have to wait for someone eldse to do it.

[Bender]

Originally posted by Greebe
Check this awesome HS's out! (shame it's not for CPU's!)... and look at them C/W ratings! :D

http://www.d6industries.com/heatsinks/hspic1.jpg


D6 Industries (http://www.d6industries.com/heatsinks/)

I would love to stick that to a T-Bird. The only problem with going so big is spreading the heat out from the tiny processor core. A 1/4" copper base would be great on that monster. How much do those things cost? Maybe we will have a few air cooled pelt setups on athlons before we know it. Can you immagin mounting that thing. I would probably cut out spaces in the bottom of it to accomodate capacitors and whatever else got in its way.

[Greebe]

Mictlan, I hear what you are saying about heat, total system cooling and cost... when I invented the first commercial GFD (YES that was me! and reviewed on [H]ocp) there were no aircooled HS's on the market at the time back in '99 to solve my cooling issues/needs, not commercially that is. I looked high and low for anything.. nadda.. oh of course I could have addapted my water cooler (which at the time was already several years old) but since that wasn't practical, nor would 99.9% of users out there build one, let alone risk water being in their system so that was totally out of the question.

In this current testing of the Zalman 3100G I did go the extreme... in several ways... first one was to see how well I could improve upon Zalman's basic setup without producing excessive noise (Delta fans a notorious for this) and in "Part Duex" with the 120mm fan was to show how much better it could be with a litle modification that can be readily done by those who are willing to mod their systems... like we do. Now in addition to have the 120mm fan cranked to objectional SPL's (which I also dislike) have it running 24/7 at a 58.3% rate... which makes it not much louder than a typical 80mm case fan. Thus checking to see what if any improvemnet in total cooling effeciency is over the first mod'd setup and it did :)

I have and do take major considerations to all things... to which many think I'm being overly "testy" hehe

But then we're in an OCers specific forum and those here, like me don't do "normal" things with their computers :)

btw we have that same quote here too :)

[Greebe]

This reminds me of the propaganda that Intel is doing "I'm 500 MHz faster than AMD". We all know that CPU speed is not all. So please take all the C/W with a grain of salt, and compare the data from lots of different people to get a more accurate idea of the performance of a heatsink.

Uhhh... yes I do and T totally agree. Being one of the very first around in OCing cpu's (and all things about computer hardware) always have been.

Also why I'm a moderator of a forum myself... not because I'm always correct mind you, but on average have much more soundful knowledge than the avg peep... and I'm not limited to just computers in that regard.