Testing the NVIDIA GTX 970 with IC DIAMOND

[Dicehunter]

I ended up just buying a tube of IC-D, Haven't put it on my GPU yet but CPU is running nicely

 

[Scottland]

I picked some up earlier in the week as I needed some for a new CPU. Haven't done any scientific tests, but seems to have lowered GPU temps by a couple of degrees.

 

[frostdragon]

Nice one Scotland.

Might get a bit more out of it once its gone through its 4 hours continuous burn-in (cure time) on that GPU.

 

[Bradmax57]

Good man as I have the same card good to know as the standard paste they put in it is quite worrying, as you will soon find out.

PM trust me your addy and ill get some sent out.

regards

frosty

Hey mate I totally forgot about this, did any ever get sent out as I never received anything and I sent you my addy via pm?

 

[frostdragon]

still awaiting more to arrive from hq fella

 

[Bradmax57]

still awaiting more to arrive from hq fella

Cheers!

 

[tipes]

Would love to give this a go on my 980GTX sc. Would be happy to let you know results.

Mark

 

[Dicehunter]

After taking the H100i off the 4790K I noticed a sort of brown circle caused by IC-D, It discoloured the CPU's IHS a little bit too i.e darkened it a little bit but that's to be expected and didn't leave any marks perse as it did on the H100i's base.

Is there any way to get rid of the brown circular mark left behind on the copper base of the H100i from IC-D ?

*EDIT*

This isn't a picture of mine but it's the mirror image of what IC-D has done to the copper on the base of the H100i, Although the brown ring like mark on my H100i's copper base is a tad bit worse than this -

After seeing this I'm a little sceptical of putting it on the GPU die to be honest.

 

[Josh]

Although not a 970 I changed the TIM on my GTX 780 SC, have had it a year and the temps were around 78-80 is Fifa and 73C in BF4.

Took the cooler off and the sheer amount of TIM that was applied was disgusting, it was overflowing, clumpy and some of the poorest application I have ever seen, however at EVGA that is responsible for the application needs a slap.

I used FIFA for the temp mainly as this game I can run maxxed out.

FIFA 15/Max settings/16xAA/Std TIM/ 78-80C
FIFA 15/Max settings/16xAA/IC Diamond/ 68-70C

Couldn't believe how much cooler it was running so tried it on BF4

BF4/Custom settings/Std TIM/ 73-75C
BF4/Custom settings/IC Dimaond/ 66-68C

Very impressed, loaded up Heaven just to make sure.

Heavenbench/Ultra settings/4xMSAA/Std TIM 78C
Heavenbench/Ultra settings/4xMSAA/IC Diamond 72C

100% worth the 5mins it takes to take apart, clean, apply new TIM and refit the cooler for me.

 

[pastymuncher]

It's most likley that you would have seen a similar drop from a good paste such as Gelid Extreme or MX-4. All this praise of IC-D is pointless without comparisons to other tim's. In my pc it made no difference on the gpu and less than 1 degree on the cpu.

 

[Josh]

It's most likley that you would have seen a similar drop from a good paste such as Gelid Extreme or MX-4. All this praise of IC-D is pointless without comparisons to other tim's. In my pc it made no difference on the gpu and less than 1 degree on the cpu.

It's most likely I would have but Frost sent me a tube to test and I have.

If I had the others to hand I would probably have tested them also.

 

[IC Diamond]

It's most likley that you would have seen a similar drop from a good paste such as Gelid Extreme or MX-4. All this praise of IC-D is pointless without comparisons to other tim's. In my pc it made no difference on the gpu and less than 1 degree on the cpu.

Just to note

As I recall you were you were one of the outliers in our 1st giveaway.

http://forums.overclockers.co.uk/showthread.php?t=18215803&page=18

With a waterblock issue which we noted was investigated by frostdragon and resolved to be a contact and pressure issue.

Just to note

Our program of giveaway's are probably the most consolidated group of independent test comparisons available.


Most recently here

http://forums.overclockers.co.uk/showthread.php?t=18598613&page=3

Generalizations about single test comparison without consideration of test variables like heat density or contact and pressure do not mean much.

Basically in order to qualify your result you have to quantify it, which being a complex process is accomplished by incorporating multiple tests.

An example below of the effects of heat density

Note: Comparisons are an average of results on each test platform and reflect that as higher thermal densities (area/watts) increase thermal performance differences increases between compounds. High thermal densities pose a unique challenge for thermal pastes because there is a large amount of heat passing through a restricted area. This challenge, however, is where IC Diamond shines in its ability to dissipate heat even under such intense and extraordinary conditions. The dismissive claim that "all thermal compounds perform about the same". Not true. A high performance racing car might idle as well as a CPU. but the true test is to take that car on the road and see how it handles under stress at high speeds.

This is demonstrated by comparing notebook and desktop CPU tests, notebooks while having perhaps a CPU that runs @ 35 W vs. a desktop @ 95W will run much hotter with a 2X delta increase in compound performance due to the increased thermal density @ one third to one half the power of a desktop CPU

.

Results

Results show that careful attention to contact and pressure can yield significant benefits and that performance requires having both good contact and good pressure.

The total span of temperatures in the sample group ranged from -5ºC to +3.9ºC, a non trivial 9ºC spread due to either contact or pressure.

So as an example which test above best matches your result, Data point #2 or #10? To qualify you result you have to quantify it, if you can not do that the data is just just static noise on it's own.

Generally you need about 60% contact or all pastes run into a "Heat Dam" effect similar to applying a match to a piece of aluminium foil where the the end where your fingers are holding it is cool. There is just not enough material to conduct the heat and so homogenizes the result making the performance look the same.


This is why with some thermal compound reviews you only see a degree or so difference, they mostly get the hierarchical order correct the absolute difference can not be quantified with lack of test controls.

 

[DarkBahamut]

Was reminded of this thread as I went to re-paste my GPU earlier with my ICD I had from awhile back. It's a replacement GTX 780 I got after I had to RMA mine. I noticed it was running a bit hotter than my last one, then I found out why

No paste on part of the core at all! ICD on and temps down 9C. Had a similar sort of drop when I had it applied on my 6970 a couple of years back too. Good stuff

 

[Dicehunter]

Wonder how effective this would be on a 295X2

 

[theaterix]

Any chance of getting some to test on my Reference Sapphire R9 290? Im sure if it tames that heat beast, it will tame anything
I've got MX-4 on there at the moment, and im sure ive got the stock temps saved somewhere on my PC.
Hopefully your still able to set someone up

 

[cjgardens]

I had some ICD in the draw for around a year. When I got a new cpu I got it out to put some on and it's rock hard. Is this normal? No other paste I've used has ever done that in a sealed tube.

 

[IC Diamond]

I had some ICD in the draw for around a year. When I got a new cpu I got it out to put some on and it's rock hard. Is this normal? No other paste I've used has ever done that in a sealed tube.

I have samples from 2007 that I take out periodically and check and none have dried out or separated as a matter of course as manufacturers need a shelf life number we quote 2 years.

leaving a cap off with the tip in contact with something porous like a paper towel can cause the liquid to wick away over time.

If can return the tube to frostdragon I can arrange to have him replace it as I would like to take a look at it.

 

[cjgardens]

Ive got 2 somewhere. It's from when you gave testers away and asked us to compare it with other TIM.

The cap has been on the whole time and left in my bits and bobs draw.

if you want me to send it back that's no problem no need for a replacement it's a few years old now and I didn't pay anything for it

 

[Telecaster]

Surely the overriding factor here is the removal of poorly applied stock TIM?

I bet if you used the same stuff but cleaned the die and applied it properly there'd be next to nothing in it.

I have some IC-D coming in the post, I also have some TIM that is supplied with my yet-to-fit CPU cooler, I think I'll do a little test.

 

[IC Diamond]

Surely the overriding factor here is the removal of poorly applied stock TIM?

I bet if you used the same stuff but cleaned the die and applied it properly there'd be next to nothing in it.

I have some IC-D coming in the post, I also have some TIM that is supplied with my yet-to-fit CPU cooler, I think I'll do a little test.

That's why we also mostly compare groups of IC Diamond tests as variables in a single test are usually unaccounted for.

Some may measure ambient high or low, contact and pressure will vary between systems, some may apply to little compound as well as varying heat densities are unaccounted for which is why we break it out.

Do your test and results may be accurate but are only true for your system and is not reflective of what an average user would experience.

You can remove 10% the outliers on either of the extremes on the graph above and the overall average will only change a few 10th's of a degree.

As an example of how it works Below is a comparison of IC Diamond vs Arctic Silver 5 from independent user tests on 24 different forum giveaways.

After the first 30 or so results you had a 95% confidence interval on the results and has only varied by a couple of 10th's of a degree on the the next 200 tests.

Consequently you can can take our overall averages from our forum tests and assume it is a good indicator hierarchically of what an average user is likely to experience.

You could assume IC diamond was a Placebo in the results and the real test result is how all the other compounds perform against the stock compounds in which case the comparison might be Arctic silver 5 and The MX compounds are only a couple of degrees better than stock compound with Shin Etsu the only one having a clear advantage. Just how you look at it.

Although I have been noticing an uptick in stock compounds performance in the last year or so so the margins there may be evaporating.