• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Why You Can't Measure Beneficial Effect OF Silver Greases

J

johncar

Senior member
Here's a very logical and proven explanation why you may not "show" a lower temp with better/silver greases.

First of all, nothing changes the heat flow/power dissipation of the cpu except the cpu itself and app/s running. And all heat generated is removed, but higher thermal resistance cooling systems force temp in cpu to rise in order to force the heat out, as "Temp drop" is the motive force in heat transfer. Cpu temp rises until heat out equals
heat generated...that's the physics.

But the cpu/hs heat transfer path has several components with their individual thermal resistances, and yes, the silver grease lowers the "overall" resistance, so it therefore lowers the cpu internal temp and the cpu case top
temp on the "cpu side" of the lower resistance silver grease interface.

But it "does not" lower the hs resistance. So for the same heat transfer, any sensor measuring hs surface or related temp will "not be lower". In fact, it will be slightly
WARMER, because the lower resistance of the hs heat flow branch causes a bit more, (of total heat), to flow thru the hs rather than thru the constant resistance branch of the pins...a minor "secondary/negative feedback effect".

Hope it's clear now that you can't measure the effectiveness of an interface material UNLESS you can measure the temp on the cpu side of the interface...which is a practical impossibility for most cases. Like real estate..."location" is everything.

You could try reading internal thermal diodes on Intel chips, as they are on the cpu side of interface. Diodes are not "accurate", but they are "repeatable"...ie, tell same lies every day.
John C.
 
"You could try reading internal thermal diodes on Intel chips, as they are on the cpu side of interface. Diodes are not "accurate", but they are "repeatable"...ie, tell same lies every day."

Repeatable is good. It is irrelevant if it's actually giving the correct actual temperature. I'm not sure if you agree with that, but it wasn't clear from your explanantion. (Not flaming, just trying to clarify the statement.)

If a diode consistently reports 42 degrees +/- 2 degree with a certain setup, and then after changing the setup consistently reports 35 degrees +/- 2 degree, then the results are likely valid. It's not as important that the actual true temperature may be 43.8 and 37.1 respectively.

 
i always thought that it was a rule of thumb that the hotter the heatsink was, the better cooling it's doing? 😕
 
Nick,
Thank you for your support.

Eug,
Yes, agree...repeatabilty is more important than absolute accuracy for this type of comparative measurement..."same lies" just cute way to say repeatable.;-)

yo2tup,
No, as Nick and Eug would agree... best heatsinks run cooler for given heat load and case temperature...cause lower thermal resistance means lower temp/temp drop to case temp required to drive same heat flow thru the hs. And lower hs thermal res means lower overall thermal res...leads to lower chip temp to drive the heat generated.
John C.
 
John,
With a little trimming of the plastic end, a thermal probe CAN be located against the edge of the die. A second probe can be placed on the heatsink. The lower the temp delta, the better the thermal interface, (interface compound and heatsink capability). I'm a true believer in Phase Change Thermal Compound, which most people just remove. A slight increase in temp is offset by the greatly improved capability to handle mutiple thermal cycles without drying out like most silicone based greases.
 
yo2tup,
Here's any easy to visualize hydraulic analogy to heat flow and driving forces....why better hs runs cooler.

Pour water at constant flow rate into a vented cylinder with partially closed drain valve at the bottom. Water level in cylinder will rise until drain flow = inflow...right??...level is indication of "pressure" required at this equilibrium.

Now open the valve a wee bit, lowering its resistance to flow, and the level will drop to a new equilibrium, as less pressure is neeeded to force inflow thru lower resistance.

Water flow equivalent to heat flow.
Valve positions equiv to 2 different heatsinks.
Hydraulic res equiv to thermal res.
Water level/pressure equiv to temperature at hs base.
Atmospheric pressure valve exhausts to equiv to case temp.
Pressure drop equiv to temp drop.

Can substitute chip temp and "overall" res....same principle...
John C.
 
DaddyG,
Can't and wouldn't challenge claims re PCTC, just don't have the experience.

But we've had Circuit Works Conductive Silver Grease as interface for
over 2 years...about 1 year since last install...and every time we re-installed saw no drying...though we operate a very lightly loaded system...might never get hot enough to degrade the oil.
John C.
 
Just a note to the newbies that the CircuitWorks stuff is electrically conductive. If you use this stuff, be VERY careful that you don't get it on any exposed contacts of anything in your case. I use Arctic Silver, which is excellent as well (rated better than the CircuitWorks stuff actually) but not electrically conductive. A tube of the stuff costs less than the CircuitWorks stuff (same sized tube).

All that being said, I can't measure anything because my mobo doesn't see the diode of my Celeron, and I can't be bothered to stick in a thermal probe (partially for the reasons mentioned and partially because I frankly don't care all that much anyway).

With a PEP66 I've been running at 920 MHz 1.9 V since late July or early August or something.
 
I can see your point, but what he might mean is that when the hs runs warmer, you know its working and absorbing heat better than if your same setup was running a cold HSF.

explanation:

If you had a warped HS that wasn't touching the processor, it would run cooler.

If you had a HS that is touching the processor using a thermal paste, it would run warmer because now the heat can transfer better.

I can see what you are saying though.
 
JohnCar I will have to disagree with you. It is not nearly impossible to mount a thermistor on the underside of the cpu. A while back I've seen specialy made thermistors that were specificly designed to be mounted on the underside of the cpu. They were wired with flat wires that went around the cpu pins. I imagine it would not be too hard to find super fine coil winding wires and make one yourself and just use a bit of dense foam to hold the thermistor snugly against the underside of the cpu. On a second point... "greases does not lower the resistance of the heatsink" you are correct only in the absolute sense. Applied to the real world silver(impregnated) greases does help because the resisance of the heatsink is much lower than the resistance by the interface betwen the heatsink and the cpu. The uneven surface, air gaps, and difference in materials increases thermal resistance. Like a chain being only as strong as it's weakest link so is moving heat. If the highest resistance is found not at the heatsink itself but rather at the juncture between the heatsink and cpu. Then lowering the resistance there will increase the capacity to move heat. I've in the past seen quite a few tests done on this already on various websites, most tests have measured cpu heat from the underside of the cpu. Most of the tests have found differences in performance between different brands of thermal greases. Then again, why take my word for it? why don't you go to radioshack buy a thermistor, find very high gauge (thin) winding wires and mount the thermistor on the underside of a cpu. Run some measurements yourself with different greases and see if your statement that silver greases do not "help" On a third and most important point(and perhaps most disturbing) you refer to the tempreature sensor as a "thermal diode" It is patently clear that you have no idea which way is up or down and are just pulling this out of your rectal orifice. Let me explain why I say this. If you had even the most basic knowledge of electronic principles you would have known that thermal mesurements are not done through diodes. Diodes regulate the flow of electricity in one direction. Just how are you going measure tempreature with that? Thermistors are what is used in temp measurement and they are resistors. Resistance of a thermistor changes with tempreature so you get a linear constant between resistance and temp within a certain range. As for your wild claim that they are inaccurate and only give you a constant "lie" and not actual measurements. That too is out of your rectal orifice and is in itself a lie. The boiling point of water at sea level does not change. You could be in timbuktu, or in the united states. At a given barometric pressure (sea level) the boiling point of pure water will always be the same. The relation between resisance and temp is a constant everywhere and like the boiling point of water (at a given barometric pressure) cannot change. Therefore the only limiting factor in accuracy is manufacturing tolerance of the thermistor itself relative to the temp chart. Therefore being a constant, if the thermistor tells you that you had a 5 degree drop in temp it can be relied upon as very accurate. The measurement itself can be relied upon as accurate, but not it's reference to the real world. That can be calibrated by another constant (boiling water at sea level) Once you calibrate your resistance chart to your temp chart using the boiling point constant you can have extremely good accuracy. The only real variable at that point will be the true barometric pressure for that instant in time. Considering the limited range of thatit will affect your accuracy to less than one tenth of one degree F or C. Oh yeah one last thing... if you have located a diode that will change it's poles in relation to temp (I've never heard of it) what do you plan on doing? have an array of a a thousand diodes? And how will you get all those diodes in the measurement area?
 
eia430,

From the Intel web site: (Emphasis added)



<< The Pentium® III processor incorporates an on-die diode that must be used to monitor the die temperature (junction temperature). A thermal sensor located on the motherboard, or a stand-alone measurement kit, may monitor the die temperature of the Pentium III processor for thermal management or instrumentation purposes. >>



Before you insult someone as experienced and knowledgeable as johncar, you should know what you are talking about.

Nevin
 
Nevin,
Thanks for your support and verification of the existance of thermal diodes...appreciate it.

eia30,
Will ignore childish language/insults and respond to the issues you brought up, in the interests of readers, as that has always been our motivation. They deserve a response, however questionable your points
and presentation.

You wrote...quote..
Applied to the real world silver(impregnated) greases does help
because the resisance of the heatsink is much lower than the resistance by the interface betwen the heatsink and the cpu. The uneven surface, air gaps, and difference in materials increases thermal resistance. Like a chain being only as strong as it's weakest link so is moving heat. If the highest resistance is found not at the heatsink itself but rather at the juncture between the heatsink and cpu. Then lowering the resistance there will increase the capacity to move heat. end quote....

But the subject &quot;does not say&quot; silver greases are &quot;not beneficial&quot;, it says you can't &quot;measure&quot; their beneficial effect. So it's apparent that you don't read too well, too fast a speed reader??

The above statement is true, esp with temp sensors on the hs side of the interface, and we're sorry if our explanation why was not clear enough...hope most other readers understood.

And while a thermistor under the chip may read a lower temp with silver grease, it may not read the &quot;full effect&quot; of the silver grease,
because of thermal resistance between the die and thermistor, and the
shift in heat flow from that branch to the hs branch. Again, &quot;measurement of the full beneficial effect&quot; is the issue.

And we made &quot;no reference&quot; to the accuracy of &quot;thermistors&quot;, which took up most of your response. Again, poor reading on your part.

But as Nevin pointed out, you were obviously unaware of the existance of Intel's on-die thermal diodes, and &quot;their&quot; inherent inaccuracy due to something called an &quot;ideality factor&quot;, which is quite variable. We don't say so, Intel does. Go to intel.com, search &quot;thermal diodes&quot; and read some of the many app notes that display. To get accurate measurements, Intel says to rip out any generic mobo readout interface and install special kits from Analog Devices or Maxim which compensate for the variable ideality factors by applying current in both directions, something the generic mobo interfaces can not do. But for &quot;comparative&quot; purposes, their repeatability is adequate as we inferred.

Finally, only Intel chips have thermal diodes, and not all mobos have the generic interfaces. So there was cause to post this issue of the &quot;measurement&quot; of the beneficial effects of silver grease.
John C.
 
Nevin, I still stand in the same place as I stand before. Please explain to me the principles behind how a diode measures tempreature. I can show you plenty of faulty statements on various websites and brochures. I have explained to you why I don't think a diode can be used effectively or at all in tempreature measurement. It is now your turn to explain to me beyond &quot;because they say so&quot; how a diode is used in tempreature measurement. I compare this to a person telling me &quot;bandsaws are used in heart surgery&quot; and as proof the person points to a website. I look at this and think... hmm why use a bandsaw when a scalpel will do a much better job? I'm also curious why you didn't think the same thing when you saw &quot;diode&quot; instead of thermistor on their webpage. I see your &quot;proof&quot; and I see nothing more than a typo made by a marketing person. Again, please tell me how a diode is used in tempreature measurement, you see &quot;proof&quot; I see a typo, show me proof. I will try on my side to get the tech papers on the PIII and see if I can find information about this beyond what would be a typo.
 
Where are the &quot;typo's&quot; in intel's statement?

Not to mention your completely rude and obnoxious behavior in all of this....



Mike
 
eia430,
It's not Nevin's job, nor our's, nor anybobody else's to prove to you here on this BBS the existance and operating princples of &quot;thermal diodes&quot;...and they are not thermistors.

As we wrote in post reply, intel.com, search &quot;thermal diodes&quot;, and you'll get a handful of app notes, some of which go into great detail, esp the the ones that deal with the special Analog Devices and Maxim kits that are used to calibrate those thermal diodes.

Have a good read and learn, just like we did.
John C.
 
Johncar, I have looked over my previous post and unfortunately I do agree with you. There was some rather colorful language/insults injected in there. I apologize for those and in my defense an explanation. I must have had (and did) quite a nasty day and you were the unwitting and undeserving recipient of my built up frustrations for that day. What else can I say? it happens to the best of us. Onwards to the points...But the subject &quot;does not say&quot; silver greases are &quot;not beneficial&quot;, it says you can't &quot;measure&quot; their beneficial effect. So it's apparent that you don't read too well, too fast a speed reader??

The above statement is true, esp with temp sensors on the hs side of the interface, and we're sorry if our explanation why was not clear enough...hope most other readers understood.

And while a thermistor under the chip may read a lower temp with silver grease, it may not read the &quot;full effect&quot; of the silver grease,
because of thermal resistance between the die and thermistor, and the
shift in heat flow from that branch to the hs branch. Again, &quot;measurement of the full beneficial effect&quot; is the issue.

Ok colorful language and insults ignored on my side as well so to address your &quot;cannot be measured&quot; statement refering to benefits of silver grease. Yes it can be measured. When you have 4 variables and you want to measure 1 then keep the other 3 variables unchanged and any difference in the result MUST be from the change made on the single variable. If the pathways of heat through the pins, radiation, thermal convection, etc remain unchanged and the results have changed then the single changed variable (interface grease) must account for 100% of that change. I do agree that everything is inter-related and changing the delta T in one variable will change the delta T on the rest. It does not however change the fact that those changes (benefits) can be 100% attributed to a single change and therefore benefits of that change can be measured.

And we made &quot;no reference&quot; to the accuracy of &quot;thermistors&quot;, which took up most of your response. Again, poor reading on your part.

But you did, you said &quot;Diodes are not &quot;accurate&quot;, &quot; in the third to last line of your post. Despite our differnce in terminology I'm sure you were refering to a temp sensing device.

Sigh, I just read the tech papers on the PIII and a paper on RMS thermal diodes here (to those interested) http://transducers.stanford.edu/stl/Projects/rmsconv.html and yes it indeed is a diode and not a thermistor. On the debate of thermistor/diode I humbly concede defeat. This however changes nothing on the original debate on weather it is an imposibility in determining measured benefits of thermal greases. Your stance that benefits cannot determined or measured is patently wrong. A drop in cpu temp measured from below the die can be attributed solely on the performance of a thermal grease if that thermal grease is the only changed variable.


 
Mikewarrior you said

&quot;where are the typo's in intel's statement?&quot;

I never said there was a typo I infered the possibility of one. I also stated that I would look at it myself to determine weather I should change my stance or not. I have addressed this issue in my post to Johncar, read it there.

you said &quot;Not to mention your completely rude and obnoxious behavior in all of this....&quot;

There are all sorts of different personalities out there, get used to it. I am consistantly one of the nicest people around, but I too can have a bad day and my attitude can change to reflect that. I could also say... do you really think I care what you think? Actualy I do agree with you, my post to Johncar was obnoxious and rude. I have apologied to him for my behavior and hope that it didn't affect him too much.



 
eia430,
Communication is man's biggest problem....
There are 3 basic locations where relevant temps can be measured.

1: In the die with a thermal diode. That measures a temp drop, and the &quot;full&quot; temp change in question.

2: Touching the chip somewhere on the side or under. That may show a change in temp but may not be the same as the full temp change in 1 because of a shift in heat flows that occurs between the various heat flow paths when you change the res of any one path.

3: Touching the hs anywhere. Will show no change, or even be a bit higher because the thermal resistance of the hs does not change...and temp drops along the &quot;chain&quot; of resistances are proportional to the thermal res of &quot;each individual&quot; resistance. If you have 3 resistances in series with some current, and change only one resistance, the voltage drop across the other 2 does not change for the same current.
It works the same way for heat flow, thermal res, and temp drop.

We said &quot;...can't measure the beneficial effect of silver grease unless you can measure the temp on the cpu side of the interface, which is a practical impossibility for most cases&quot;...later amended to &quot;full effect&quot;...(which occurs exactly at top dead center of chip exterior or in the die itself).

So if you review the 3 cases above, you see that &quot;only&quot; the thermal diode can measure the full effect. But &quot;only&quot; Intel chips have thermal diodes. and &quot;not all&quot; mobos are equipped to interface to those thermal diodes. And it is a practical impossibility to not only locate a sensor top dead center on the chip but to also not contaminate it with grease, res of which which would affect the measurement. So combining that with 2 and 3, our statement &quot;...practical impossibilty for MOST CASES&quot; stands.

Then you wrote.......quote......
But you did, you said &quot;Diodes are not &quot;accurate&quot;, &quot; in the third to last line of your post. Despite our differnce in terminology I'm sure you were refering to a temp sensing device. End quote.....

Sorry, we &quot;were&quot; referring specifically to thermal diodes, not to thermistors, nor temp sensing devices in general. The difference in terminology/interpretation is &quot;yours&quot;. We try to be very careful about our terminology since we recognize that communication &quot;is&quot; man's biggest problem.

In conclusion, intent of original post was to encourage people to use silver grease despite &quot;many&quot; other posts which indicate small or no gains towards lower temps...measured at inappropriate locations described above. But there are a few sites that test with chips with thermal diodes and their stats are therefore believable. And we've been using silver grease for over 2 years, last person to badmouth the stuff.
John C.
 
Gotta add my 2 cents here:

In simple terms, cooling the chip is a 2 step process.

1. Move the heat from the cpu core to the heatsink.

2. Move the heat from the heatsink to the surrounding air.

The better the thermal interface between the core and the heatsink, the more heat will be transferred and a possible higher HSF temp.

Possible because the heat may be dissipated with a highly efficient heatsink/fan
 
Johncar I am not understanding your term for &quot;full effect&quot; I'll elaborate. If I change one variable and one variable only (thermal grease.) Measure it at the ideal location(the very center of the Processor IC itself I will get a percentage gain/loss from the original readings before the single variable change. If I measure it from another less than ideal location say from behind the slug or even from one of the pins I will still have the SAME percentage value. Yes the absolute value will change if compared to another location, but with all other variables unchanged (including measurement location)the percentage gain/loss from the grease change will be exactly the same. Meaning this if your measurement from the middle of the IC is 100F before and after it is 90F thereby having a 10% drop, if you measure it from one of the pins, before the change it was 50f and after the grease change it will be 45F. Realative to the original location of the measurement the absolute value will change but not the ratio of loss/gain. Again, if only one variable was changed and all the rest remain the same any and all change in the results MUST be attributed to that single variable change. This will be irrespective of measurement location (within reason) the ratio gain/drop will be the same. This is going by the logic that heat dispersion in one location affects dispersion of others in direct contact with it. A rate of change in one location will affect rates of change in other locations (in direct contact) in a parallel manner. Of course as you travel further away from the ideal measurement point more variables(akin to noise) such as air flow that constantly and uncontrolably change are introduced and therefore will blur accuracy.

So, in no uncertain terms what I'm saying is that you can show the performance of a thermal grease as a percentage value in relation to the last measurement taken before the change no matter where you take the measurement. As long as you are able to keep all other variables identical to the same values before the single variable was changed.
 
eia430, you wrote.....
So, in no uncertain terms what I'm saying is that you can show the performance of a thermal grease as a percentage value in relation to the last measurement taken before the change no matter where you take the measurement. As long as you are able to keep all other variables identical to the same values before the single variable was changed.
end quote...

But we are &quot;not&quot; able to keep all other variables identical to the same values before the single variable was changed.

When silver grease is applied to the principal heat transfer path thru the hs, this causes a &quot;redistribution&quot; of the heat flows thru &quot;all&quot; heat flow paths. The major heat flow hs path gets a &quot;little more&quot; of the constant total heat flow because of its new lower thermal resistance. But that &quot;little more&quot; comes from the other &quot;minor heat flows&quot; paths, like thru the bottom and/or the edges of the chip. So those minor heat flow paths see a much higher percentage change in their heat flow than the low percentage change in the main/hs flow path. Besides which, the heat flow changes are in opposite directions, increasing in the main/hs path and decreasing in the other minor heat flow paths.

So you cannot expect constant percentage changes in temperature drops in all the branches which have experienced not only non-proportional changes in heat flows but also opposite polarities.

You've used a simple and sometimes valid statistical experimental test analysis in your thinking, but have ignored or were not aware of all the subtleties that occur in a multi-path heat transfer process, and which conequently invalidate that &quot;one variable&quot; change approach.

Your logic would apply if we had 2 insulated bars with variable interfaces between them...heating at one end and cooling at the other.
testing for the thermal conductivity of various interface materials.
Aavid, (heatsink mfr), ran such tests a few years ago. But that describes a simpler &quot;single&quot; heat flow path process.

Hope this explains this last issue. And &quot;full effect&quot; was meant to describe the &quot;absolute change in degC&quot; of die temp...as compared to any lesser changes that might occur at other locations. After all, the main stat we'd &quot;like to know&quot; is how much cooler is &quot;the cpu&quot;, not the absolute degC change at some other location.
Cheers, John C.

 
You know much of the argument/discussion between johncar and eia430 are well over my head, as an attorney and not an engineer/scientist. For whatever reason though I spent the last 20 minutes reading it all and its very interesting stuff. I love learning new things 😀

 
eia430 wrote:
&quot;Oh yeah one last thing... if you have located a diode that will change it's poles in relation to temp (I've never heard of it) what do you plan on doing? &quot;


The resistance of a diode changes with temperature. The change is essentially linear with temperature. The characteristic has long been used to advantage in circuit design.

I have a circuit here in which a common 1N4007 diode (2 for $0.99 at RS?) is the sensor monitoring exhaust temperature of a forced-air cooled power amplifier.

In another, a diode, mounted across, and in contact with, the top of a power amplifier transistor, senses the transistor's temperature and changes the transistor's control voltage to keep the transistor from overheating.

Now, back to the thermal transfer issues.
 
Johncar, yes I do agree with you smaller heat pathways especialy those that are so close as to change poles will experience greater percentages of change. But these are small phenomenons that are the the exceptions and will be found only at the very fringes where variables can nolonger be kept constant. For the purposes of our debate these are quite irrelevent. A heat measurement reading taken from the underside (below)of the cpu slug is only something like 3 to 6 thousands of an inch away from the &quot;ideal&quot; location (inside the ic itself.) This is quite outside of the sphere of influence of smaller heat pathways and quite inside the sphere of influence of the main heat pathway. That location will yield a resolution accuracy that matches the &quot;ideal&quot; spot that might prove to be immeasurable outside a laboratory.
If I measured a 5 degree drop in cpu temp measured from the underside of the cpu, that will be and is the FULL benefit of the performance difference. It is so because it takes into account all the variables. It is a cpu that is mounted on a motherboard with the same heat sink in the same computer etc, etc, etc. The only difference being changing the grease. I would not be able to say measured &quot;full&quot; benefit if this measurement were taken in a laboratory outside the actual working enviroment of it's daily use. To say that your readings would be skewed because of smaller heat pathways would be stating a falsity because those smaller heat pathways are existant during both measurements. Beyond that they will also be there during normal day to day operations. To say that the 5 degree drop in cpu temp is the FULL benefit of the one single changed variable would be stating a measured FACT. It is very possible even easy to measure the full benefit of silver grease vs regular grease. Your stance would only be correct in the even that you change other variables, perhaps in different computers. In those cases then yes, you would not be able to state that a 5 degree drop is the full benefit. In another computer it might be a 7 degree drop. But within the parameters of measuring a single difference (changing grease and that only) and hence our debate, your stance is incorrect.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top