Innovation Cooling's Diamond 7 TIM test results

[Das Capitolin]

Let's get this back on topic:
I've used AS5 for years and after what?? 2-3 weeks now I see a solid 3.5C improvement with D7 vs AS5..
That's me, one machine and not a scientific process at all.
Applied in a cross pattern, fairly heavy.
BUT it is fact here and thats what matters to me.
I get better temps on my Q6600 with D7 than with ASS.

I completely understand your situation. You might notice that the results I just published in this thread (using a heavily overclocked P4) are different than those of my first TIM comparison article results which used a E6550 Core 2 Duo. IC7 was a top product in that article and received an award, but in this new article it didn't do nearly as well.

This is why I retested with a new application, and then retested again a third time. I only have one syringe of the material and it's nearly all gone... so perhaps it's a forumla issue? At this point though, it didn't do better than AC5 (with a 200+ hour cure time) on a stock socket 478 CPU cooler and P4 overclocked to 3.6GHz.

[D_A]

i've seen the same with my opteron system too, much better temps than as5.

YES! errrr, sorry. I've been waiting to see a result on an Opty system, coz that's what I'm running. Opty 170 with a stock HS (from a 185) clocked to 2.31 (board limited in voltage options and the chip frapps WCG units like a machine gun at 2.35)

[fart_plume]

Ok, Das Capitolin I have a couple of questions. First Are you using software to read the temps from the internal diodes of the cpu or do you have a thermal probe in the heatsink. The reason I ask is most of us use software to read the temps. And yes we know the thermal probe method is more accurate. It took me testing several different programs to get one that even read my temps even close to correct.
The second question i have is your setup housed in a case or an open test bed?

[David_L6]

I've always used Ceramic or AS5. I'm out of Ceramic so I haven't compared it to the Diamond 7.

One of my computers showed ~1°C improvement with the Diamond 7. I'm watching it to see if the temps stay the same, get higher, or get lower over time. So far, no change.

Another computer showed ~ 4 1/2 °C improvement with the Diamond 7. This one surprised me so much I tested three times! Don't know why the improvement was so drastic. Side note: This was with a Zalman 9700 heatsink. Had it been a TRUE there's no way I would have taken it off and re-mounted it that many times.

I have two more computers to test. I'll try to get to one of them this week. I had to RMA then Ballistix in the other one. Once I get the RAM back I'll test the Diamond 7 in that computer too.

[Movieman]

I completely understand your situation. You might notice that the results I just published in this thread (using a heavily overclocked P4) are different than those of my first TIM comparison article results which used a E6550 Core 2 Duo. IC7 was a top product in that article and received an award, but in this new article it didn't do nearly as well.

This is why I retested with a new application, and then retested again a third time. I only have one syringe of the material and it's nearly all gone... so perhaps it's a forumla issue? At this point though, it didn't do better than AC5 (with a 200+ hour cure time) on a stock socket 478 CPU cooler and P4 overclocked to 3.6GHz.

Maybe, just maybe, since the D7 is "thicker" in texture and with most of the guys here(and myself) useing TRUE HS's and the crappy finish to many of them that with that thicker texture it is helping to make up for the deficiencies of that HS.. I don't know, just a thought.
Also and this is just my feeling, maybe get away from using the Northwood for testing as it's outmoded today.
I'ds suggest 2 chips to use..a 8400 and a Q6600..
Gets you closer into what most are using now.

[Das Capitolin]

Ok, Das Capitolin I have a couple of questions. First Are you using software to read the temps from the internal diodes of the cpu or do you have a thermal probe in the heatsink. The reason I ask is most of us use software to read the temps. And yes we know the thermal probe method is more accurate. It took me testing several different programs to get one that even read my temps even close to correct.
The second question i have is your setup housed in a case or an open test bed?

You are incorrect. A thermal probe is not more accurate than the core temp diode inside the processor (source: Intel Corporation, Arctic Silver Corporation, and my own testing.) Even in theory, this concept is completely flawed. How can you obtain an accurate reading of what's inside by measuring from the outside? The internal core diode is the best way to go. Here's a quote from my article, which further details my answer:

At the start of each test, the ambient room temperature was measured to track any fluctuation throughout the testing period. EVEREST Ultimate Engineer Version 4.20.1170 was then utilized to create core loads and measure each individual CPU core temperature. It's important to note that software-based temperature readings reflect the thermistor output as recorded by the BIOS. For this reason, it is critically important to use the exact same software and BIOS versions throughout the entire test cycle, or the results will be incomparable. All of the units compared in our results were tested on the same motherboard using the same BIOS and software, with only the product itself changing in each test. These readings are neither absolute nor calibrated, since every BIOS is programmed differently. Nevertheless, all results are still comparable and relative to each products in our test bed.

One unfortunate problem is that CPU's report temperatures as a whole number and not in fractions. This in turn causes the motherboard BIOS and subsequent software applications such as EVEREST to also report to the nearest whole number. To compensate for this, our tests were conducted several times after complete power down thermal cycles. Conversely, the ambient room temperature levels were all recorded and accurate to one-tenth of a degree Celsius.

The Q6600 was used in our Best CPU Cooler Performance - Q1 2008 project. There are several reasons NOT to use this processor for testing and comparing thermal pastes, but here are my top three:

1) It doesn't offer a single core reading. Multiple core readings allow for a degree of fluctuation and error.
2) The Q6600 does not get hot enough to create large differences between good/bad TIM products.
3) The stock Intel Thermal Cooling Solution is the best way to get high temps, and it uses a four-corner push-pin mount. This mounting system is very low pressure, and does not compress the elements together with nominal force.

So while you're probably wondering why I would use a socket 478 P4 3.0 OC'ed to 3.6, here are my top three reasons:

1) Single core reading. It is what it is, and there's no confusing the results with an average of four cores. Furthermore, dual and quad core CPU's rarely ever have the same temperature between cores.
2) It gets hot. Add some voltage, and it gets REAL hot.
3) The Intel socket 478 compression-lever thermal cooling solution is probably the best ever made by Intel. The unit does not slide or twist, and it compresses evenly with good force. Additionally, it features a perfect 1" round copper core, which ensures that the same amount of TIM coverage is compared in every test.

[Movieman]

snip.....

I can't argue with your logic at all.
Makes perfect sense.

[[XC] riptide]

Das Capitolin: It would appear that then some pastes are better for different processor/Heatsink combinations?

It seems that some pastes do a reasonable job in counteracting the less than perfect surfaces of some of the heatsinks that we use, and certainly the less than perfect HSF that appears on some of Intels recent chips.

[fart_plume]

YES! errrr, sorry. I've been waiting to see a result on an Opty system, coz that's what I'm running. Opty 170 with a stock HS (from a 185) clocked to 2.31 (board limited in voltage options and the chip frapps WCG units like a machine gun at 2.35)

I'm (the Wife) running a opty 144@2.4ghz runs 24/7 the difference between as5 and IC Diamond was 6*C. IC Diamond works much better on this system than the as5.

[Das Capitolin]

Das Capitolin: It would appear that then some pastes are better for different processor/Heatsink combinations?

It seems that some pastes do a reasonable job in counteracting the less than perfect surfaces of some of the heatsinks that we use, and certainly the less than perfect HSF that appears on some of Intels recent chips.

This is absolutely correct. I don't mean for this to be a shameless plug for my site, but our Best Thermal Paste Application Methods was just published yesterday. There are so many different mounting styles, heatsink shapes, and surface finishes, that no one product will be a universal match. And then there's the mounting pressure; given enough compression between elements and any product will do well because you're ultimately trying to press metal against metal. We go into detail in regards to HDT coolers, which benefit from a compound like IC7.

Which brings me to another thought... since the purpose of testing TIM products is to determine which has the best thermal conductance, it almost makes me wonder if I'm wasting months of work only to learn that in the end there are several excellent products which perform roughly the same. Actually, I already know this to be the case, but what I mean is that after testing 55+ thermal pastes all I will know is which ones are bad, which ones work okay, and which ones work best. As my testing has already prooved, the "best" product may work great on one system with one specific cooler surface, but that same product may not work nearly as well on a different system.

I suppose that with products reaching nearly the same performance level, this is what it all boils down to.

[littleowl]

so what your saying is your tests work great for an old 82w heat load but for a newer more popular 100w heat load your site is very wrong.
I don't see that as a good reason to jump me because I said something doesn't seem right with what you posted here on the forums.

[DDTUNG]

You are incorrect. A thermal probe is not more accurate than the core temp diode inside the processor (source: Intel Corporation, Arctic Silver Corporation, and my own testing.) Even in theory, this concept is completely flawed. How can you obtain an accurate reading of what's inside by measuring from the outside? The internal core diode is the best way to go. Here's a quote from my article, which further details my answer:



The Q6600 was used in our Best CPU Cooler Performance - Q1 2008 project. There are several reasons NOT to use this processor for testing and comparing thermal pastes, but here are my top three:

1) It doesn't offer a single core reading. Multiple core readings allow for a degree of fluctuation and error.
2) The Q6600 does not get hot enough to create large differences between good/bad TIM products.
3) The stock Intel Thermal Cooling Solution is the best way to get high temps, and it uses a four-corner push-pin mount. This mounting system is very low pressure, and does not compress the elements together with nominal force.

So while you're probably wondering why I would use a socket 478 P4 3.0 OC'ed to 3.6, here are my top three reasons:

1) Single core reading. It is what it is, and there's no confusing the results with an average of four cores. Furthermore, dual and quad core CPU's rarely ever have the same temperature between cores.
2) It gets hot. Add some voltage, and it gets REAL hot.
3) The Intel socket 478 compression-lever thermal cooling solution is probably the best ever made by Intel. The unit does not slide or twist, and it compresses evenly with good force. Additionally, it features a perfect 1" round copper core, which ensures that the same amount of TIM coverage is compared in every test.

It seems to me that by selecting a very hot heat source on one end and employing a not so efficient heat sink on the other, you are limiting your ability to measure the heat conducting ability of the TIMs being tested.

Just like you can't race cars in congested traffic, and expect the fastest car to win.

DDTUNG

[Das Capitolin]

so what your saying is your tests work great for an old 82w heat load but for a newer more popular 100w heat load your site is very wrong.
I don't see that as a good reason to jump me because I said something doesn't seem right with what you posted here on the forums.

I've tried to explain the theory and principal to my tests, but it appears that my explaination has made no impact on you. Would someone else care to explain how a single 81.9W core can test thermal conductance of TIM between a 1" round contact point as well or better than four cores of 23.75W each over a 1.5" square base can offer? Aside from already explaining the major downsides, I'm not sure what else you're looking for.

It really just seems to me like you're looking to start problems again. It also seems like what you've just said has nothing at all to do with what you initially said which provoked me.

[Das Capitolin]

It seems to me that by selecting a very hot heat source on one end and employing a not so efficient heat sink on the other, you are limiting your ability to measure the heat conducting ability of the TIMs being tested.

DDTUNG

Please explain your view, because from my position you would want readings to be as high as possible to encourage a disparity between products. It's impossible to claim that one item is better if they are all within .01 degrees of each other. So please tell me how I could do this all better.

[DDTUNG]

Please explain your view, because from my position you would want readings to be as high as possible to encourage a disparity between products. It's impossible to claim that one item is better if they are all within .01 degrees of each other. So please tell me how I could do this all better.

If the heatsink you use is not capable of moving enough heat, then the best TIM in the world will not have a chance to perform. I hope that's clear enough for you.

DDTUNG

[D_A]

I've tried to explain the theory and principal to my tests, but it appears that my explaination has made no impact on you. Would someone else care to explain how a single 81.9W core can test thermal conductance of TIM between a 1" round contact point as well or better than four cores of 23.75W each over a 1.5" square base can offer? Aside from already explaining the major downsides, I'm not sure what else you're looking for.

It really just seems to me like you're looking to start problems again. It also seems like what you've just said has nothing at all to do with what you initially said which provoked me.

It seems that rather than listen to people's honest concerns over your methods you are more than happy to accuse then of provoking you. If asking you an honest question about your methods is all it takes you need to get a much thicker skin.

[fart_plume]

Simple get a hs that won't suffer from heatsoak. A block of aluminum with fins will retain heat for a longer period than a heatpipe system with cooling fins. The greater the hs cooling ability the greater the deltas between the TIMs

[Bail_w]

If the heatsink you use is not capable of moving enough heat, then the best TIM in the world will not have a chance to perform. I hope that's clear enough for you.

DDTUNG

indeed true. for example, the stock intel cooler (copper base) comes with the intel quad core. No matter if you use the stock paste or high end TIM, the differences in temperatures is very small (assume same pressure was applied).

[Das Capitolin]

So in your mind it would be better if I used a super-efficient cooler? You do realize that the core temps would have very little difference between TIM products at that point, right?

Right now I can see a difference in products by almost 1~2 degrees using the overclocked P4 at 3.61 GHz and a stock cooler. If I used your suggestion, I would see .1~.2 difference, and reduce the consistancy in TIM amounts because of enlarged surface. It wouldn't be feasable to use an after-market cooler with this kind of test, because the mounting method would come into play.

[Bail_w]

@Das, why don't you try a AMD single core S939 and a stock heatpipe heatsink instead of the P4. Report back with the results.

[Das Capitolin]

@Das, why don't you try a AMD single core S939 and a stock heatpipe heatsink instead of the P4. Report back with the results.

Because I don't have one on hand. Feel free to send the equipment my way and I might consider it.

[D_A]

Please explain your view, because from my position you would want readings to be as high as possible to encourage a disparity between products. It's impossible to claim that one item is better if they are all within .01 degrees of each other. So please tell me how I could do this all better.

Let's consider an analogy that may help. Say you have a pump. This pump can push 1000 gallons of water an hour through a given pipe and spray head combination. You want to know if different pipes will give you less back pressure. Through changing out your section of pipe you determine that one pipe give less back pressure than another.

All sounds reasonable so far, doesn't it.

Now let's look at that a little closer. You're using a spray head that will flow 1000 gallons per hour. All you can do with your pipe is restrict it BELOW that, so if you put a section of pipe in that's rated at 2000gph at a given pressure you're not going to get an oz over 1000 because of your spray head.

Applying that to our situation here, if your HS is only good for 100BTU but you're giving it 200, your CPU will heat up. If your TIM can move 300, 250, 200, 150, 125 or 102 but your HS can only move 100 it doesn't matter a squirt. You can only gauge the effectiveness of the TIM if it is the limiting factor.

[DDTUNG]

So in your mind it would be better if I used a super-efficient cooler? You do realize that the core temps would have very little difference between TIM products at that point, right?

Right now I can see a difference in products by almost 1~2 degrees using the overclocked P4 at 3.61 GHz and a stock cooler. If I used your suggestion, I would see .1~.2 difference, and reduce the consistancy in TIM amounts because of enlarged surface. It wouldn't be feasable to use an after-market cooler with this kind of test, because the mounting method would come into play.

At least you will then be focusing on the TIMs as the limiting factor. Isn't that the purpose of your test?

DDTUNG

[Vapor]

Use the best cooler you can and push the temps up to the levels you're at now--best of both worlds.

Prescott at high volts (and moderate clocks) should be able to make and stably sustain ridiculous temperatures on even the best coolers.

[Das Capitolin]

How can we presume that the supplied Intel cooling solution is a limiting factor? It cools down to 67C with AS5, but only to 77C with another paste. So it seems that it's not limited at all based on output.

If you use an aftermarket cooler, you lose the ability to have a wider delta. You also lose the 1" contact area which will remain consistant in every test, and you'll lose the level mounting pressure.