Industry heatsink+TIM testing@frostytech

[Nugit]

Frostytech has put up a small article about testing of heatsinks and TIMs. It's pretty basic, but there are a couple of pretty pictures to look at.

http://www.frostytech.com/permalink.cfm?NewsID=47649

[BillA]

wth ?
pretty basic ?
this is the actual article
http://www.frostytech.com/articlevie...articleID=1915

totally beyond the comprehension of most here (no die temp !)
this is the Intel doc http://download.intel.com/design/Pen...s/30255304.pdf

it is an excellent article

[jaydee]

Not beyond the comprehension. Just many here are enthusiasts trying to squeeze every last mhz out of their system and the lower the DIE temp the lower the mhz. IHS temp is irrelevant, to them, and not what thay want to see tested.

Is that not simple enough for YOU to comprehend?


My view is not the same. I don't like the groove milled on top of the area being tested but whatever. If IHS are going to be better attached then they are not coming off and this may well be the only option we have unless the on-die diode becomes reliable and consistent.

[BillA]

please read 4.2.7 of that Intel doc for comprehension

the IHS temp has no 'fixed' relationship with the 2 different die temps from the diodes in the CPU due to the variations from its operation

now if Intel cannot / will not define either of those temps as THE DIE TEMP (which would have a very real relationship to the IHS temp), who are you ? (or pH, or Stew, etc.)

-> in the pursuit of every last mhz, did Stew use a die sim ? what die temp was used ?

you guys need your heads examined

no die temp is needed to design a cooling solution, the IHS is fine
the real problem remains that of the applied power when using a CPU as the heat source

[jaydee]

Originally Posted by BillA

please read 4.2.7 of that Intel doc for comprehension

the IHS temp has no 'fixed' relationship with the 2 different die temps from the diodes in the CPU due to the variations from its operation

now if Intel cannot / will not define either of those temps as THE DIE TEMP (which would have a very real relationship to the IHS temp), who are you ? (or pH, or Stew, etc.)

-> in the pursuit of every last mhz, did Stew use a die sim ? what die temp was used ?

you guys need your heads examined

no die temp is needed to design a cooling solution, the IHS is fine
the real problem remains that of the applied power when using a CPU as the heat source

We are not trying to design a cooling solution.

[BillA]

?
well jd, my jaw dropped with that answer

"Just many here are enthusiasts trying to squeeze every last mhz out of their system and the lower the DIE temp the lower the mhz."

so you are squeezing mhz how ?

[Les]

Originally Posted by BillA

so you are squeezing mhz how ?

Possibly a side-squeeze
There are indications from Waterloo that improved edge-cooling nay be 1000fold more effective,for control of hotspots, than better attached heatsinks .

[jaydee]

Originally Posted by BillA

?
well jd, my jaw dropped with that answer

"Just many here are enthusiasts trying to squeeze every last mhz out of their system and the lower the DIE temp the lower the mhz."

so you are squeezing mhz how ?

Well we are not? Is pH, Joe, or Brian designing a cooling solution or the hundreds of readers? I know I am not anymore. We as in the majority of the readers here. I don't even watercool or overclock myself. Completely happy with my stock IHS attached Sempron 2500+ with stock AMD all aluminum cooler. I don't even have a temp reading on this comp to look at.

Worded the quote there wrong. The lower the die temp the more the mhz to gain. To a point anyway. There is no other reason to water cool. Not even for quiet anymore.

If you guys want to debate words other than the tech issues then whatever. This site has become useless if so.

[BillA]

if you are back to air, why the hassle ?
read your post jd, there is no reason for your pitching into something in which you have no interest or knowledge

be so kind as to be quiet as you are unable/unwilling to contribute
note the thread title

[jaydee]

Originally Posted by BillA

if you are back to air, why the hassle ?
read your post jd, there is no reason for your pitching into something in which you have no interest or knowledge

be so kind as to be quiet as you are unable/unwilling to contribute
note the thread title

Bill what are YOU contributing to? What is your goal? It surly has nothing to do with the majority of the readers of this site. That is why I come here to help them out. Not the handful of commercial interested people such as yourself.

What does the title of this thread have to do with anything? Shows a method industries might use. Who cares? What does it have to do with any of US? Who here is going to build such a bench and why? And if they build it what good will the results be for the audience here?

What is the goal of ProCooling? I do not think it is catering to commercial interests. Always thought of it as an enthusiast site which doesn't seem to be YOUR goal.

Also find it humorous that when anyone disagrees with you they all the sudden become ignorant. Whatever Bill. Good luck designing a usless test bench.

[BillA]

“Bill what are YOU contributing to? What is your goal?”
I’m a testing guy, and I use that info. My goal is better testing which can be defined variously.

“It surly has nothing to do with the majority of the readers of this site.”
Guilty as charged, if one has no interest in test results then my posts should be skipped. But I suspect most here are interested in test results.

“That is why I come here to help them out. Not the handful of commercial interested people such as yourself.”
Contributions take different forms; I accept that you do not wish to “help me out”. (Is this the new angle ?) I do suggest you drop your anti-business bias, look around you. How many of the things you see did you make ? Who made the tools ? Who pays your salary ?

“What does the title of this thread have to do with anything? Shows a method industries might use.”
Might ? both Intel and AMD were described.

“Who cares? What does it have to do with any of US? Who here is going to build such a bench and why? And if they build it what good will the results be for the audience here?”
Part of the article was about grooving the IHS of a CPU, did you see that ? I suggested this for review sites to improve the accuracy of the temp used for “C”. While you may not find value in such data, others will (not all have persistent ignorance, they can learn new things).

“What is the goal of ProCooling? I do not think it is catering to commercial interests. Always thought of it as an enthusiast site which doesn't seem to be YOUR goal.”
Joe can speak to the goals of procooling. My commercial interests ? As expressed how ?
Again you leave me speechless jd, I am NOT enthusiastic ? (goddamn I think I’m obsessed)
Or must all be just like you ?

“Also find it humorous that when anyone disagrees with you they all the sudden become ignorant. Whatever Bill. Good luck designing a usless test bench.”
Yes, I have a bit of knowledge of some things. Such is relative.
Look up Luddite.
Be a bit humble jd, you are ignorant wrt a lot of this technical stuff.

[Brians256]

Here's my goal:
the design of a cooling system that will give me the most performance (how fast it does my tasks) while being reliable, affordable, and quiet.

Any of those goals that is missed will give me a failed system. I use air in one system because I had not the time or money (at that moment) to put a better watercooling system into it.

The idea that we are (homogenously) not interested in designing a cooling system is not accurate. There is a limit to which we can go on optimizing single components for performance, and I believe we are at that limit right now.

Tests are easiest when done for isolated components for easily measured criteria, such as AMD64 die temp given same system and only changing the waterblock. That does not guarantee it is generating data that is desired, as pH will readily admit. That is one of the reasons, I believe he is interested in doing system tests. That is in addition to the fact that doing the same test over and over again gets boring.

[Brians256]

Originally Posted by Les

Possibly a side-squeeze
There are indications from Waterloo that improved edge-cooling nay be 1000fold more effective,for control of hotspots, than better attached heatsinks .

Edge of die or IHS?

[maxSaleen]

....Let me go get some gasoline....there's a flame here I want to pour some fuel on...

I looked over the article about 3-4 days ago when it was first put up. Good read. A lot there to comprehend, and I agree that it is beyond most (even here including me). I believe in continuing education; reading something like that (the article from frostytech) falls in line with my belief.

[BillA]

slippery slope there Brian
you are suggesting that a single component may not yield system excellence

I agree; as a practical matter GOOD methodologies are required to distinguish between top wbs, but what is this saying ?
use is little different, the users can't tell either (uncontrolled)
hence the pissing and moaning over the Apogee
there are lots of lessons for the experts here

yea Les, give us a bit more to chew on

Thanks mS, I believe most are here to learn

[Les]

Originally Posted by Brians256

Edge of die or IHS?

Edge of die.
Only,as yet, exploratory musings.
As mentioned previously(link) the h values considered are similar to those expected from forced air convection in a 0.5mm gap(heat capacity would rule).

[BillA]

do we have access to the die edge ?
(did I not understand ?)

[Les]

I dunno
but if bunged-up with TIM1 then yes
otherwise only consider the edge contact resistance
the numbers will be different
the point was to indicate that a small edge h (be it convection or contact) may have a major effect

[BillA]

are you assuming the bulk resistivity of the TIM ?
what thickness ?

does not seem likely (but more area always helps)

[jaydee]

Originally Posted by BillA

“Bill what are YOU contributing to? What is your goal?”
I’m a testing guy, and I use that info. My goal is better testing which can be defined variously.

“It surly has nothing to do with the majority of the readers of this site.”
Guilty as charged, if one has no interest in test results then my posts should be skipped. But I suspect most here are interested in test results.

“That is why I come here to help them out. Not the handful of commercial interested people such as yourself.”
Contributions take different forms; I accept that you do not wish to “help me out”. (Is this the new angle ?) I do suggest you drop your anti-business bias, look around you. How many of the things you see did you make ? Who made the tools ? Who pays your salary ?

“What does the title of this thread have to do with anything? Shows a method industries might use.”
Might ? both Intel and AMD were described.

“Who cares? What does it have to do with any of US? Who here is going to build such a bench and why? And if they build it what good will the results be for the audience here?”
Part of the article was about grooving the IHS of a CPU, did you see that ? I suggested this for review sites to improve the accuracy of the temp used for “C”. While you may not find value in such data, others will (not all have persistent ignorance, they can learn new things).

“What is the goal of ProCooling? I do not think it is catering to commercial interests. Always thought of it as an enthusiast site which doesn't seem to be YOUR goal.”
Joe can speak to the goals of procooling. My commercial interests ? As expressed how ?
Again you leave me speechless jd, I am NOT enthusiastic ? (goddamn I think I’m obsessed)
Or must all be just like you ?

“Also find it humorous that when anyone disagrees with you they all the sudden become ignorant. Whatever Bill. Good luck designing a usless test bench.”
Yes, I have a bit of knowledge of some things. Such is relative.
Look up Luddite.
Be a bit humble jd, you are ignorant wrt a lot of this technical stuff.

Those quote tags are available for a reason. I am not going to try and reply to that.

[jaydee]

Originally Posted by Brians256

Here's my goal:
the design of a cooling system that will give me the most performance (how fast it does my tasks) while being reliable, affordable, and quiet.

Any of those goals that is missed will give me a failed system. I use air in one system because I had not the time or money (at that moment) to put a better watercooling system into it.

The idea that we are (homogenously) not interested in designing a cooling system is not accurate. There is a limit to which we can go on optimizing single components for performance, and I believe we are at that limit right now.

Tests are easiest when done for isolated components for easily measured criteria, such as AMD64 die temp given same system and only changing the waterblock. That does not guarantee it is generating data that is desired, as pH will readily admit. That is one of the reasons, I believe he is interested in doing system tests. That is in addition to the fact that doing the same test over and over again gets boring.

You are not designing a cooling solution though. Just slapping together some parts so and so says works good together. Designing a cooling solution is from paper to production IMO. There is no one here that I know of that is doing such. Not sure what BillA is doing now though if anything it is commercial. That is the only reason for such testing.

As for the article reviewers are simply NOT going to do that Bill. They are not going to spend what they have in their pocket. And the very few that do will simply be wasting their time. By the time they get the bench working a new CPU will come out and they will have to do it all over again.

None of the testing done today or this year will be relevant to Presler style CPU's. That is why I say design another useless test bench because it will be just that.

If you got the cash to continue to change your bench and always be a year or 2 behind current CPU design then have at it. pH is already behind the times with his 939 system.


Also Bill you claim to be an enthusiast but and enthusiast of what? Testing? No commercial intrests? Then what was Lothar rambling on about in that thread he started?

[BillA]

ask Lothar about Lothar
testing, data, better performance; thats my program
pH is dead in the water, he cannot define what he purports to measure (the die temp)

I guess you resent my starving to be able to do that which I wish,
all about choices

reviewers are not concerned with the future, they test what is available today
all this changes continuously, if that is too much be a librarian
I do not know if a review site will groove an IHS, that would qualify for "pro", no ?

[pauldenton]

Originally Posted by BillA

please read 4.2.7 of that Intel doc for comprehension

the IHS temp has no 'fixed' relationship with the 2 different die temps from the diodes in the CPU due to the variations from its operation

now if Intel cannot / will not define either of those temps as THE DIE TEMP (which would have a very real relationship to the IHS temp), who are you ? (or pH, or Stew, etc.)

-> in the pursuit of every last mhz, did Stew use a die sim ? what die temp was used ?

you guys need your heads examined

no die temp is needed to design a cooling solution, the IHS is fine
the real problem remains that of the applied power when using a CPU as the heat source

hmm - i think you may be misunderstanding the issue people have with using the IHS temp (or maybe i am....)

aiui the issue is not that a better source is (or more accurately isn't) available on an intel CPU - but that:
a) the purpose of cooling is to cool the die, not the IHS...
b) it is possible (probable?) that there is no fixed relationship between the temp of the die (or whatever portion most requires cooling for a given use, since the .pdf confirms that it will vary) and the IHS temp (which is taken at a particular point at the centre of the IHS iiuc)
c) that being so it is possible that the more efficient (at cooling the die) of 2 blocks might give the same or worse IHS temp as a less efficient block

of course as old bill might have put it:

[BillA]

lol
a) can't do one w/o the other
b) a given
c) no; the IHS temp is an average, precisely that desired - and with good equip and procedures can distinguish finely between sinks (and I can say so because I have done so)

I relate to the cartoon, feeling rather singular in this contest

[jaydee]

Originally Posted by BillA

ask Lothar about Lothar
testing, data, better performance; thats my program
pH is dead in the water, he cannot define what he purports to measure (the die temp)

pH is the only one on the track I like and think most people here like. However we see that track is broke at the moment.

I guess you resent my starving to be able to do that which I wish,
all about choices

I don't resent it at all. I just can't comprehend what results you are after and why. It doesn't seem to jive with what most enthusiasts are after. Whatever an enthusiasts is anymore. I followed your lead on the die sims just to figure out it was not worth the effort. And now find the only few approaches are not worth the cost. I know I do not have the cash to build a new Presler or dual core AMD machine specifically for testing and then buy whatever equipment and then do whatever mods. We are talking $1,000's here. Not $100's. When you do get a bench going and take your tests and get your results they will be useless to next gen parts and CPU's. So what is the point?

reviewers are not concerned with the future, they test what is available today
all this changes continuously, if that is too much be a librarian
I do not know if a review site will groove an IHS, that would qualify for "pro", no ?

Reviewers are always two days late and $10,000 short. They hardly do today's stuff. How many dual core AMD machines are used for reviews? I have not seen any. This is the problem. Toms and Anantech (and a few others) are the one's that have the ability to keep up but they choose not to and for a reason. The ends do not justify the means.

Pro? Could be. If that is industry standard then I guess. However I deal with industry standards every work day and can say they are not always Pro. Who will and why if the next CPU will be out in 6 months and you have to buy another $2,000 worth a computer parts and mill another CPU and re-calibrate the bench to keep up? And what about your old results on old platforms? Toss out window...

Anyway keep doing it however you want to do it as you have a reason for it and motivation for it. Just don't get to pissed when others can't figure out how to apply your results to anything. That seems to be the biggest problem right now.

It is pretty sad though when end user reviews seem more usefull than bench testing.