Innovation Cooling's Diamond 7 TIM test results

[123bob]

It's kinda like ever diminishing concentric circles of Chaos theory. I may be signing up for therapy myself soon. The minutia is getting to me

123bob if nothing else you are the most consistent guy I have ever seen. Looking over the farm 9 results you have 9 tests. At load averaging all cores with three different pastes except for the one AS5 which on a remount along with 7 other pastes tested within about 1/2 a degree from each other with the IHS lapped and unlapped. I know you noted the compression issue but first time I had a chance to look close at he numbers. pretty remarkable

LOL, well, my name is Bob. We Bobs have a reputation for being nothing if not consistent.....or is that stubborn maybe?

I WILL get this CPU to do something before I'm done. Of course, if I keep at it, I might get it to ambient at load because I broke the darn thing....

I figure it's probably time to move on to a few other machines to give you some more results...I'd like to see what my other six fuzion block machines show. I have better control of the mount with those.

Bob

Bob

[SiGfever]

It's kinda like ever diminishing concentric circles of Chaos theory. I may be signing up for therapy myself soon. The minutia is getting to me

123bob if nothing else you are the most consistent guy I have ever seen. Looking over the farm 9 results you have 9 tests. At load averaging all cores with three different pastes except for the one AS5 which on a remount along with 7 other pastes tested within about 1/2 a degree from each other with the IHS lapped and unlapped. I know you noted the compression issue but first time I had a chance to look close at he numbers. pretty remarkable

He always does very well in therapy also. When he leaves the office he can be heard muttering..."Need more Quads, need MORE Quads".

[littleowl]

LOL, well, my name is Bob. We Bobs have a reputation for being nothing if not consistent.....or is that stubborn maybe?

I WILL get this CPU to do something before I'm done. Of course, if I keep at it, I might get it to ambient at load because I broke the darn thing....

I figure it's probably time to move on to a few other machines to give you some more results...I'd like to see what my other six fuzion block machines show. I have better control of the mount with those.

Bob

Bob

yes please move on I want to see what happens with the fuzion setups also.

[littleowl]

Ok so here is where I am at with the E6600. I told in my other post that I was on stock air and changed it over to the 6002 mig shift water loop. Well I did some playing today and I will be doing some more playing here as soon as tasty gets in gear.

3 Day cure time with D7
Load
Core 0: 29c
Core 1: 29c

Room temp was about 22c

I put Ceramique on the chip for and ran full power for 15 min.
Load
Core 0: 30
Core 1: 29

Room temp had went up some to 23c

now I changed it again to AS5
After about 5 min the temps were
Core 0: 30
Core 1: 29

It will stay on AS5 till I get a package from Tasty


Check out these pics from the 3 day cure of D7!!! I could win this pressure test looking at this.

[tastymannatees]

I sent out the first round of film for digital mapping - and I will email your results monday or tuesday only 9 went out this week.

I did talk to the rep about the part of the film that shows voids.- The white or unexposed portions "voids" are a go-no go type of situation on the pressure side and no indicator of defects. Voids could be .0001 or 10 miles deep as far as the film reading is concerned. Just something to be aware of.

[tastymannatees]

Check out these pics from the 3 day cure of D7!!! I could win this pressure test looking at this.

From 123bobs pics the small "beaded" paste pattern looks best for good contact and you may be running third place though a little thick on the edges.

We will have to see how the prints come out.

[123bob]

Started doing some more research into the linearity of the internal DTS temp diodes in these CPUs. I have not waded through Intel's dissertations on the subject yet but did come across a couple of interesting links. I'll wade through Intel's stuff this weekend.

The other big program out there is Real Temp. I'm starting to run that with Core Temp to see how they compare.

Some 33 pages of stuff on it available in our own forum here. I have not read it fully but there appears to be some calibration options available with it.

Real Temp program available here.

Interesting article on the theory behind CoreTemp here.

I'm going to start an attempt at looking at the linearity of my readings using both of these programs. I'll be reporting both sets of temps for the new machines I'm about to crack into. As soon as I get some new film, I'll be cracking into these machines, all with fuzion blocks.

I should report that I DID have one interesting result on Farm-09 right after I did the second pressure test. I booted after re-appling paste after the second lapped pressure test. Coretemp was reporting temps about 3 degrees LOWER than before. That was idle and load..... I thought it would be a eureka moment, but I didn't trust the result. I re-booted the machine and the temps went right back to par with the prior results.....Very suspicious....I could not repeat it, but clearly some mechanism was involved. (And no, I was not eating too much paste at the time....)

(Side note to Dr. SiG. I've actually been muttering "need some octo-cores, need some octo-cores... It might have sounded like "need more quad-cores", but I mutter very well..... Don't know if that makes a difference in your therapy plans, just thought I should be honest with the good Doctor....)

Regards,
Bob

EDIT: Another interesting link. Good for reading if you can't sleep....G'Night all!!

EDIT2: The documentation and theory of Real temp here, suggests that only the high end of the temp range is calibrated. This is interesting since what we are seeing seems to be compression at load. I'll have to give this some thought....

I read further into the reference made in the above doc and thought this extract was of note. The whole article is here.

"More than a few programs have been released over the last few years, each claiming to accurately report these DTS values in real-time. The truth is that none can be fully trusted as the Tjunction values utilized in these transformations may not always be correct. Moreover, Intel representatives have informed us that these as-of-yet unpublished Tjunction values may actually vary from model to model - sometimes even between different steppings - and that the temperature response curves may not be entirely accurate across the whole reporting range. Since all of today's monitoring programs have come to incorrectly assume that Tjunction values are a function of the processor family/stepping only, we have no choice but to call everything we thought we had come to know into question. Until Intel decides to publish these values on a per-model basis, the best these DTS readings can do for us is give a relative indication of each core's remaining thermal margin, whatever that may be."

My bold italics. I may be tilting at windmills to think I can get reasonable, linear data out of the internal diodes.....

This would say that for TIM testing, the lab setup is the way to go and the testing we do is really more about the mount efficiency and characteristics. The pressure film testing is a very good addition for this purpose. (Side note - Tasty, just for completeness, I would be curious to see what a film test of your lab setup would look like. Would it be a relatively solid patch in the contact area of the test rig? )

That said, I'll move on to the Fuzion machines when I get more film.

[123bob]

Here's some interesting shizz. I got to playing around with Real Temps "Test Sensors" button. What this appears to do is to take an idle machine, measure the DTS temp reading then apply full load and measure the DTS temps again. It then reports the difference as an indication of sensor movement. The purpose is to see if you have a stuck diode on a given core.

I think I can also interpret this to show the non-linearity of the diodes across a few machines. I also think the load presented to the CPU cores should be consistent at least to all the cores in that machine. My data is below followed by more thoughts on the potential meaning of it.

First, a screenie of my main rig's results. The pop up explains the author's intent on the results.




Now to tabulated data;

Farm-04: D7 Tested, 3.6 gig, X3350, Mass WC, Fuzion > 4,4,6,3
Farm-07: D7 Tested, 2.8 gig, Q6600 G0, Z-9700 > 9,7,9,10
Farm-09: D7 Tested, 1.6 gig (EIST on), Q6600 G0, TRUE > 6,7,7,8
Farm-11: Reference, 2.97 gig, Q6600 G0, Z-9700 > 10,9,11,8
El-Machino-2: Reference, 3.15 gig, QX6700 B3, Independent WC setup, Fuzion > 11,11,11,11

I have machines labeled as "Reference" these were not paste tested, yet. The rest were tested and reported on.

I realize that overclocks and cooling come into play here. The numbers have some variance within each machine, except for my oldest rig, which is my main rig. This is rock solid consistent. Does this mean that it was calibrated well and the others were not? I should think this directly shows non-linearity on the measurements, both within the machines, and with respect to each other.

Thoughts on the meaning of this are welcome. I'm not totally sure myself, but it looks interesting....Many other thoughts come to me. A few are, could the diodes be calibrated in a linear fashion in respect to a given chip? My main rig would suggest so. Could it be calibrated to an absolute measure of temperature? I don't know......

Regards,
Bob

[SiGfever]

This would say that for TIM testing, the lab setup is the way to go and the testing we do is really more about the mount efficiency and characteristics. The pressure film testing is a very good addition for this purpose. (Side note - Tasty, just for completeness, I would be curious to see what a film test of your lab setup would look like. Would it be a relatively solid patch in the contact area of the test rig? )

That said, I'll move on to the Fuzion machines when I get more film.

I totally agree, we are just testing our mounts and not the TIM. Without a controlled and constant pressure on the IHS we are just "taking a wack at it". And then because each chip can be so different in the IHS and the soldering there of that also invalidated the testing to a degree. just my thoughts.

[tastymannatees]

I totally agree, we are just testing our mounts and not the TIM. Without a controlled and constant pressure on the IHS we are just "taking a wack at it". And then because each chip can be so different in the IHS and the soldering there of that also invalidated the testing to a degree. just my thoughts.

Pressure and contact are both inextricably linked to TIM performance as well as being able to measure any result. So for me, the manufacturer to be able to provide guidance to those that can not observe a result is pretty valuable. People are constantly trying to add refinement to their equipment, so it's a benefit if you know what directions to focus energies rather than wasting energy on those things that have minimal impact.

Any one test is anecdotal with the current state of the art, reviews of paste (and cooling hardware) are for lack of a better word "squishy". Most buyers of products know this intuitively and read several reviews and consult the forums to see other user results and mentally average on their own to get a picture of performance. The whacks at it are pretty effective in the broader scope of averaging and trouble shooting.

[123bob]

Pressure and contact are both inextricably linked to TIM performance as well as being able to measure any result. So for me, the manufacturer to be able to provide guidance to those that can not observe a result is pretty valuable. People are constantly trying to add refinement to their equipment, so it's a benefit if you know what directions to focus energies rather than wasting energy on those things that have minimal impact.

Any one test is anecdotal with the current state of the art, reviews of paste (and cooling hardware) are for lack of a better word "squishy". Most buyers of products know this intuitively and read several reviews and consult the forums to see other user results and mentally average on their own to get a picture of performance. The whacks at it are pretty effective in the broader scope of averaging and trouble shooting.

Amen Tasty. I have bought CPU coolers on wonderful reviews only to find them not worth the powder to blow them you-know-where.

I think this entire experience has been very interesting. I know I've learned a lot. I just wish we could get Intel to properly calibrate parts and give us the missing specs. They would only help themselves out by doing so. I can't imagine it would be some kind of competitive disadvantage to do so. AMD already reverse engineers Intel stuff, and vice-versa.

I hope you have found this to be useful too. I've come to trust D7 after working with it. The mounts have all been very good, and consistent. I can't say that about other pastes I've used.

I'm now curious to see how my main rig performs on the D7 test. It has the consistent "sensor test" figures. It is also a dead flat lapped machine. I'm wondering how it will do pressure testing and with D7 temps.

Regards,
Bob

[tastymannatees]

I

'm now curious to see how my main rig performs on the D7 test. It has the consistent "sensor test" figures. It is also a dead flat lapped machine. I'm wondering how it will do pressure testing and with D7 temps.

I have been wading through all the documentation you supplied and then some.

The real temp stuff was interesting to me. One thing was room temp calibration. If you calibrate for room temp with a closed case seems to me you would 10-15C off from the start as the local "room" is now the case? and would be 30C + even on an open case recirculated air from a sink can raise the local ambient a few degrees unless you an extra fan removing the the exhaust air. So a calibration at ambient room with closed case is probably as good as no calibration at all.

http://www.techpowerup.com/realtemp/docs.php

That's great but I've found that when the DTS moves more than 35°C away from TjMax, the DTS data no longer changes at the exact same rate that the core temperature is changing at.

The chart he has would explain a lot as regards to a compression of results.

http://spreadsheets.google.com/ccc?k...Zw&hl=en&pli=1

This is the OCF. forum data.

What is interesting to me is that a wide delta spread at idle to load can converge to a zero result as well as a zero at idle can be zero at load. A 20 C increase in temps would normally be attributed to a 2C ICD7 and and a 4C to most the compounds tested with a 1-2 C reduction at idle.

This raises another question, Calibration @ idle there is the assumption that little or no power is being used. An adjustment of 6 or 7C implies even underclocked 30 or 40W is being used.

so with a good water system vs stock cooling that adjustment range that spread might be 2-3C water, 7-8 stock sink or more at 40 watts?. Just a guess on my part but in a box at idle I would guess the number is more like 40C? is that what is normally calibrated?

Some of the more exaggerated improvements are usually linked to a long term breakdown of the pre installed grease or sometimes a misfire on the original mount still some probably are neither so Ineed to put toghether a more complete sample

At the higher temps where the convergence on accuracy should be occurring at -35C of DTS I do not see it happening ?

[DDTUNG]

I'll ask again the question that I've put forward before.

Has anyone tested to see if the improvement in load temps translates into a noticeably higher overclock?

DDTUNG

[brot]

If someone says yes here, some shop will get an order for at last 10kg of icd7

[Bail_w]

I'll ask again the question that I've put forward before.

Has anyone tested to see if the improvement in load temps translates into a noticeably higher overclock?

DDTUNG

2-3 C less will not get your any higher overclock. So i must certain say not possible.

[Jaco]

I've one machine where I have a temp drop of 10-15°C .
But that's mainly because of a bad HS mount and crappy TIM spread method
maybe 3-4°C gain thanks to the D7 compound , not enough for a higher overclock.

[DDTUNG]

To be fair to the ID7 I can see how in some situations those 2-3 degrees would help achieve stability at a certain clock and voltage.

DDTUNG

[SiGfever]

As stated the Ambient, meaning "surrounding air" temp can be anything from the room air temp to the temp surrounding the CPU. When my free standing a/c was running the supply air temps effected the perceived ambient causing me to have to time the a/c units cycles before taking a temp reading. I was not saying this testing is all wasted just that since we can not control the variables precisely there can be a large room for error. A torture rack would make for a better test bed since there are no enclosures to trap air.

I know that diamonds transfer heat better than other compounds so the testing should have shown an improvement in every case but as with mine it did not so the other factors were preventing the TIM from excelling. The main advantage to this testing was to teach us about TIM application verses perceived techniques and to further validate how difficult it is to get a perfect mount each time. just my $.015

I would like to thank tasty for giving us this opportunity to explore these variables so we could try to fine tune our computer systems. I know the knowledge gained here will help all of us to achieve a better TIM/mount for each CPU.
John

[tastymannatees]

I'll ask again the question that I've put forward before.

Has anyone tested to see if the improvement in load temps translates into a noticeably higher overclock?

DDTUNG

The original review from AMDZone and some number tests from the giveaways had some marginal higher overclocks and better stability. The Overclocked test tends to muddy the water on the temperature results and it has not been a focus of our effort as is a little more subjective than raw temps alone.

[Martijn]

To be fair to the ID7 I can see how in some situations those 2-3 degrees would help achieve stability at a certain clock and voltage.

DDTUNG

Victor, I believe how you can get a higher overclock, simply because you can supply a little extra voltage if needed. However, for us crunchers, I don't think it'll make much of a difference, since we're never on the bleeding edge of stability.

[tastymannatees]

As stated the Ambient, meaning "surrounding air" temp can be anything from the room air temp to the temp surrounding the CPU. When my free standing a/c was running the supply air temps effected the perceived ambient causing me to have to time the a/c units cycles before taking a temp reading. I was not saying this testing is all wasted just that since we can not control the variables precisely there can be a large room for error. A torture rack would make for a better test bed since there are no enclosures to trap air.

well I am a little confused about how to do an accurate calibration. If it's an air sink case closed you would calibrate to the case ambient not the room Ambient temp. 10 C difference in the calibration number?

Water would seem more complicated, I have not had a chance to noodle it through completely but. Open case cpu would be some few degrees higher than the ambient room, higher than the motherboard temp so with a small delta you probably have some heat flow to the MB? marginal at best. Now case closed Ambient would be higher in the case but the MB Temp is now higher than the CPU because you have an exterior radiator to ambient room. Deltas have reversed so now you are drawing heat from the motherboard. So how large a delta? The larger the more influence on the calibration.

[tastymannatees]

Victor, I believe how you can get a higher overclock, simply because you can supply a little extra voltage if needed. However, for us crunchers, I don't think it'll make much of a difference, since we're never on the bleeding edge of stability.

There are 2 sides to this coin. One is the performance side which is pretty well established with approx 80% of the people seeing an improvement and with 50% seeing 3C or more over their previously installed grease. In my testing I have never seen a compound perform better than ICD7, that's not to say a couple do not come pretty close but we are at the bottom performance wise. For some 2C is a kind of "oh well" event and for others who are taking the holistic approach refining conditions throughout their system trying to get 10C lower it is another.

The other side here and the more critical point is what is the most important feature of any product besides it's performance?

It's quality. What if all CPU's steadily declined in performance a month after you bought them? or the shovel you just bought breaks the second time you use it? OEM compounds are quality type, not real exciting on the performance side but great over the long term.
Most of the 3C improvement are compared to installed pastes of less than 30 days. Most of the 4C to 14C improvements are on pastes that have been installed 30 days to one year. Less obvious because most have their eye on the absolute number, but not one of the competitive performance compounds hold up under stress. Over relatively short periods breakdown starts sometimes in only 4 weeks on highly stressed systems, less stressed maybe 12 weeks. ICD7 is bulk loaded Over 90% per Intel recommendation for stability, ICD7 is an OEM quality paste with high performance.

I realize the niggling focus on contact pressures and calibration is an irritation to some, it even annoys me but it is critical refinement to our trouble shooting for the 20% that do not and being able to do some guidance there long term saves people a lot of time remounting, re lapping and just generally beating their heads against a wall

[123bob]

I'll ask again the question that I've put forward before.

Has anyone tested to see if the improvement in load temps translates into a noticeably higher overclock?

DDTUNG

OK Victor, this one's for you.... I'm done testing TIM on Farm-07 so, let's see what I can do with the clocks.

Original Mount, original Clock: Q6600 G0 - I had to back this machine down, gradually, on the clock over time since it would not stay stable at a higher level. I could not regain operation at a higher clock with any additional CPU voltage, or anything else I tried. I had thought the degradation in clock capability was due to the board or RAM getting worn out, but I now think it may have been a decaying mount. It was running stable at 2.79gig for several months and this is the point where I started TIM tests. This has a 9700 cooler and had Shin G751 paste on it originally. It runs on an eVGA 680i board.

Original Temps and clocks.
2.79 gig - CPUz volts @ load 1.344
Ambient 25.9
Coretemp idle 49,49,45,46
Coretemp load 66,65,61,63

New Clock, New Mount: All hardware the same. I had significant improvement in temps when I went to the D7. This was at the same clock. I re-clocked and now have it running with what so far is Boinc stable. It is now at 3.318 gig!! Remember the hardware is all the same as before. The only changes were paste and mount, FSB increase, and VCORE increase. That adds up to about a 19% increase in clock. Now, is this all atributable to D7? No, I don't think so, but I'm sure it had a lot to do with having more temperature overhead to play with. Part of this could be my confidence that this mount won't fry my CPU, or better OC skills.
What I am confident in is that this mount won't change much over time. I have several weeks on some of these mounts and they are dead equal to the starting temps. What Tasty says is the big feature, that we seem to be missing somewhat, is that this paste is meant to last over time, AND be competitive with the other pastes in initial temp reduction.
For crunchers, time stability is very important. (at least until the RAM dies...) We all know we are not PC racing here. We are out for long haul numbers. If we can have both, that's so much the better....

OK, enough blabbing.

New Temps and clocks.
3.318 gig - CPUz volts @ load 1.368
Ambient 26.4
Coretemp idle 47,47,44,45
Coretemp load 68,68,65,67

At load with the higher volts and clocks, I have an increase of about 3 degrees. So far, it's stable and holding Boinc operations at that clock level. I'll take that.....

So, from this perspective, D7 absolutely rocks.....

There are 2 sides to this coin. One is the performance side which is pretty well established with approx 80% of the people seeing an improvement and with 50% seeing 3C or more over their previously installed grease. In my testing I have never seen a compound perform better than ICD7, that's not to say a couple do not come pretty close but we are at the bottom performance wise. For some 2C is a kind of "oh well" event and for others who are taking the holistic approach refining conditions throughout their system trying to get 10C lower it is another.

The other side here and the more critical point is what is the most important feature of any product besides it's performance?

It's quality. What if all CPU's steadily declined in performance a month after you bought them? or the shovel you just bought breaks the second time you use it? OEM compounds are quality type, not real exciting on the performance side but great over the long term.
Most of the 3C improvement are compared to installed pastes of less than 30 days. Most of the 4C to 14C improvements are on pastes that have been installed 30 days to one year. Less obvious because most have their eye on the absolute number, but not one of the competitive performance compounds hold up under stress. Over relatively short periods breakdown starts sometimes in only 4 weeks on highly stressed systems, less stressed maybe 12 weeks. ICD7 is bulk loaded Over 90% per Intel recommendation for stability, ICD7 is an OEM quality paste with high performance.

I realize the niggling focus on contact pressures and calibration is an irritation to some, it even annoys me but it is critical refinement to our trouble shooting for the 20% that do not and being able to do some guidance there long term saves people a lot of time remounting, re lapping and just generally beating their heads against a wall

You've made me a D7 fan. I do appreciate your approach to introducing this product. I realize that educating the general masses on the advantages is not easy, so you took a direct approach and had us try it. That is taking time , effort, and expense on your part. I was one of those who have to have solid scientific numbers to be convinced. Even though I did not have the most drastic results, I'm convinced you have a good product on your hands.

Now calibration.......I agree with your points above and must give it some more thought too. I may post up the questions in the Xtreme thread on the topic. (When I have intelligent questions to ask...) Maybe there is some mis-understanding of what is meant by ambient. My thought on it is that it's the room temp outside the box. Why I would calibrate to something that may or may not correlate to what's inside the box makes no sense to me either. I would think I would want correlation to the inside of the box.

Your points on WC setups are well taken too. Here I would think the water temp would be the calibration number you would want. I'll have to give it more thought. This is why I brought these links up above. Interesting but I'm not sure what it means to us.

Regards,
Bob

[SiGfever]

well I am a little confused about how to do an accurate calibration. If it's an air sink case closed you would calibrate to the case ambient not the room Ambient temp. 10 C difference in the calibration number?

Water would seem more complicated, I have not had a chance to noodle it through completely but. Open case cpu would be some few degrees higher than the ambient room, higher than the motherboard temp so with a small delta you probably have some heat flow to the MB? marginal at best. Now case closed Ambient would be higher in the case but the MB Temp is now higher than the CPU because you have an exterior radiator to ambient room. Deltas have reversed so now you are drawing heat from the motherboard. So how large a delta? The larger the more influence on the calibration.

A torture rack in a constant temp conditioned space with a method to accurately adjust heat sink pressure would be optimum. The room and the components in the room should be brought down to temperature (what ever the test temp will be) and held for one hour before any testing is done to ensure the components are at true ambient. Let the system run for thirty minutes to reach equilibrium and then record the "Idle" and "Load" numbers. I would also remove any fans blowing around the CPU socket so the CPU sees only heatsink transfer. Here is an example of a fairly nice test rack. I believe that Sierra_Bound has one with an adjustable heat sink pressure adjustment.

http://www.petrastechshop.com/hssttdetest11.html

Since most people do not lap their CPU a stock Intel CPU will be best for testing. If all the variables are covered and a CPU such as a Q6600 is chosen (one that can overload most heatsinks) then an accurate before and after can be achieved. AS DDTUNG stated...If the heat load is not greater than the heatsinks capabilities then the test will be inaccurate.

Without such controlled variables the test will be close but no cigar.

[Deux]

One more thermal result here.

CPU:Athlon 64 X2 5000+ BE
Cooler: Stock socket 940 Opteron HS
Ambient: 21.8
ASC Idle: 23, 23
ASC Load 27, 30

Ambient: 21.5
ICD7 Idle: 23, 23
ICD7 Load: 24,27

Excellent performance again. ICD7 is now definitely my paste of choice for everything but subzero and that's mostly because I have a 35g tube of ASC sitting here now.

I'll have both of my pressure tests out tomorrow. Thanks for this opportunity.
ICD7