does NT-H1 TIM suffer pump-out on IHS application too, or just bare-die mounting?

[BonzaiDuck]

6600gt so no but killed a Galaxy 8800GT and then a XFX 9800GT. That was when I swore off AS5 on a bare-die mount, and am afraid to do anything like it ever again. I swear swear I only put maybe 2 grains of rice of AS5 and spread it with a credit card and installed it and that was enough to kill both, FFS, FML.

Then, ICD. It should be neither conductive nor capacitive.

I knew this was always a very slight risk with AS5. If I ever replaced the TIM on a GPU, I used ICD. I may have killed a P4P800 Springdale with either AS5 or an Arctic epoxy (Alumina?) with some conductance/capacitance which I used to put heatsinks on certain motherboard components.

I'd also suggest that ICD won't spread all over the place like the picture. If you can only identify the application area of the HSF where it will mate with the IHS or die, you can spread it thin on the HSF and then apply it to the die or IHS. Any excess will form a slight rubbery bead around the edge.

Few of the good TIMs are non-conductive or electrically neutral. I think the nano-diamond paste fills the bill.

 

[crashtech]

Suppose you had 4 grams of micronized natural diamond? ICD is made from synthetic. I'd be more worried about the uniformity of the natural variety, but the micron specs were consistent with ICD's. [ . . [And we should have a forum about "Hardware Addiction and Spending Excesses." $95 for 5 grams. M***-erF***-er!!] . . ]:twisted:

Some . . . industrial abrasives outfit . . . in Pennsylvania. I should have a record of the receipt and the 2007 Quicken file.

2 months before -- possibly even a year earlier -- some enthusiast posted his own project. He acquired one of those cheap drugstore vinyl-plastic contact-lense cases. I've even seen something similar for a pair of ear plugs I bought somewhere.

You drill a hole dead-center of the circular cap. You then obtain a standard paper-clip -- plain metal -- and fashion a propeller-shaped paddle on one end, so the remainder of the clip extends from dead-center of the paddle.

Then you fill your lense-case with as much "base" lubricant and diamond powder you care to use, mindful of the published loading for ICD.

So. We have a pump-out problem with NT-H1, and maybe even some pump-out with HSF-base/IHS applications -- but maybe not noticeable or worth mention. From the web-forums for the topic, people fret over IX re-flow, thermal damage to components, heat-guns and blow-dryers. Better not use it on a direct-touch, but maybe still feasible. CLU will corrode an aluminum base for a direct-touch of this or that manufacture.

So a few questions.

Noctua HSF designs use flat, solid-copper HSF bases -- nickel-plated. The Nickel doesn't matter -- better to get rid of it. But does NT-H1: (a) corrode aluminum? (b) Is it ill-advised to use with direct-touch? And . . .

(C) OTHERWISE, WHY NOT? The only thing against it would be neck-and-neck performance between the two ingredients. But if they're approximately the same, then you might possibly resolve the pump-out problem in the other applications.

Tell ya right now! You DO NOT want to spend the hundred bucks! WHAT was I THINKING?! "Enthusiast enthusiasm overrides budget?"!!:$

ADDENDUM for PM response: I can WEIGH grams! I'll offer a gram of that sh . . stuff at 20%, or maybe $9 over the price of an ICD 24-karat tube. And! Free Forever-stamp shipping!

ANOTHER ADDENDUM: Here's an old 2008 review of NT-H1 compared to AS5 -- most prominently, with no mention of ICD. The test bed was a Q6600 OC'd to 3.6 Ghz and volted to something like 1.55V. I've forgotten the spec TDP; it might have been 95W. Maybe even a little more, but not much. Since earlier than 45nm Yorkfield, maybe easier to cool. But only 1C improvement over AS5.

Probably not worth it. No gains -- maybe the slightest loss. Plus side: I could use up my micronized natural-diamond powder. . . .

A__A__: Forgot the link!

http://www.overclockersclub.com/reviews/noctua_nt_h1/5.htm

NT-H1 isn't too terribly thick, I wonder what would happen if some of the diamond powder was mixed into it.

 

[ehume]

I've been doing a bunch of heatsink runs lately. Same heatsinks. Same cpu, same mb. The NT-H1 is a bit easier to work with, but the GC Extreme is almost as easy and cools about a degree more. Then there is this. The Kryonaut I have yet to explore.

 

[BonzaiDuck]

NT-H1 isn't too terribly thick, I wonder what would happen if some of the diamond powder was mixed into it.

It would make it thicker -- like ICD.

IDC showed a 3C performance edge of NT-H1 over ICD, if I recall the earlier posted excerpt from his de-lidding thread. The difference of that with ICD versus AS5 and NT-H1 versus AS5 must be explainable somehow. If the review link I gave earlier is accurate, than NT-H1 is between ICD and AS5 in performance for HSF-to-IHS applications.

Following that information, there's very little difference, so mixing micronized diamond with NT-H1 would give you similar performance for that sort of normal application, but might (I say MIGHT) inhibit pumping out for the direct-die usage with better results than ICD.

 

[MongGrel]

I think NT-H1 is very idiot proof, tolerant of over-application, poorly fitting joints, and less than spotless surfaces. That makes it a no-brainer for me.

Is what I use all the time personally.

I stopped using AS5 a long ago myself, and have tried many others.

NT-H1 is just easy to use, holds up over time, and has almost no cure time more or less.

 

[MongGrel]

NT-H1 isn't too terribly thick, I wonder what would happen if some of the diamond powder was mixed into it.

I actually have some diamond paste for polishing I tried just putting on by itself once and it didn't work so well.

Was a bit too thick on its own to begin with I think.

Would be interesting to see what mixing a little bit with NT-H1 might do, but not in the mood to rip anything apart atm.

 

[BonzaiDuck]

Would be interesting to seem what mixing a little bit with NT-H1 might do, but not in the mood to rip anything apart atm.

Exactly! You don't want to be taking your computer apart every day!

Even if you ever do -- you don't!

That's probably the second-best way to do it at less cost. With the powder, you avoid the oil base of the ICD, so you only use the base of the NT-H1. Except that ICD is specific to contain "certain oils" which slowly dry out, I couldn't anticipate a problem of mixing both greases, but I couldn't rule it out entirely, either.

 

[DrMrLordX]

Seems to me that as long as you aren't using AS5 these days, odds are you'll be using a better paste. NT-H1, GC Extreme, and ICD7 are all better products. I used to use Shin Etsu x23-7783D before I switched to CLU, and it seemed to work pretty well too (on lapped surfaces anyway).

 

[BonzaiDuck]

Seems to me that as long as you aren't using AS5 these days, odds are you'll be using a better paste. NT-H1, GC Extreme, and ICD7 are all better products. I used to use Shin Etsu x23-7783D before I switched to CLU, and it seemed to work pretty well too (on lapped surfaces anyway).

No doubt.

I added a new thread in "cases & cooling" asking for insight about applying IX. CLU corrodes aluminum; IX does not. I have a "direct-touch" cooler with an aluminum base, into which are pressed the heatpipes separated by aluminum ribs. It does not appear there are any gaps between the pipes and aluminum.

Even with IX, I'd expect only a 2C to 3C performance improvement over ICD, given a review comparison I'd read a couple years ago. It also has a tedious installation process.

With NT-H1, I'm stymied somewhat by IDC's table of results in his delidding investigations. It outperforms IC Diamond by as much as 3C -- for a bare die usage. Yet the review I linked shows a 1C difference between AS5 and NT-H1 for the standard HSF-to-IHS application -- favoring NT-H1. And I know that ICD outperforms AS5 by more than NT-H1 -- in the usual way on top the IHS.

Something else is going on there -- most suspicion goes to bare-die versus IHS usage. It may also have something to do with pressure, and I think IDC was using a Corsair water-cooler and its block. Then, there were also issues about a gap between the IHS and the die.

Note also that bare-die application of CLU resulted in temperature improvements of 10 to 20C over the Intel putty, yet the review comparison of CLU and ICD in conventional usage shows at most a 3C edge for CLU. The bare-die improvement with ICD was closer in magnitude to that of CLU.

 

[Idontcare]

With NT-H1, I'm stymied somewhat by IDC's table of results in his delidding investigations. It outperforms IC Diamond by as much as 3C -- for a bare die usage. Yet the review I linked shows a 1C difference between AS5 and NT-H1 for the standard HSF-to-IHS application -- favoring NT-H1. And I know that ICD outperforms AS5 by more than NT-H1 -- in the usual way on top the IHS.

Something else is going on there -- most suspicion goes to bare-die versus IHS usage. It may also have something to do with pressure, and I think IDC was using a Corsair water-cooler and its block. Then, there were also issues about a gap between the IHS and the die.

I'm sure the NT-H1 squeezed down to a thinner gap versus the ICD under pressure from the mount. But it truly does not matter because regardless how well NT-H1 did in the tests it still isn't viable as a long-term (>3 months) replacement for the TIM that interfaces between the silicon and whatever metal (be it IHS or HSF) will be employed to the shunt heat away from the silicon because of the pump-out issues with NT-H1.

 

[Geforce man]

I'm sure the NT-H1 squeezed down to a thinner gap versus the ICD under pressure from the mount. But it truly does not matter because regardless how well NT-H1 did in the tests it still isn't viable as a long-term (>3 months) replacement for the TIM that interfaces between the silicon and whatever metal (be it IHS or HSF) will be employed to the shunt heat away from the silicon because of the pump-out issues with NT-H1.

I do not doubt you (seriously, at all), but can you explain to me how to test this? I have used NT-H1 with my Noctua cooler on my 2700k, installed approximately 1 year ago, after i got sick of the corsair h50 pump noise. I have since not seen ANY increase in temps, at all. Peaks are still the same, idles are still the same, etc.

 

[soccerballtux]

I do not doubt you (seriously, at all), but can you explain to me how to test this? I have used NT-H1 with my Noctua cooler on my 2700k, installed approximately 1 year ago, after i got sick of the corsair h50 pump noise. I have since not seen ANY increase in temps, at all. Peaks are still the same, idles are still the same, etc.

he's referring to bare-die applications where you de-lid / remove the IHS from the CPU die

 

[Geforce man]

he's referring to bare-die applications where you de-lid / remove the IHS from the CPU die

Gotcha. In that case, disregard!

 

[soccerballtux]

Gotcha. In that case, disregard!

you can read more here http://forums.anandtech.com/showthread.php?t=2285595

 

[BonzaiDuck]

I'm sure the NT-H1 squeezed down to a thinner gap versus the ICD under pressure from the mount. But it truly does not matter because regardless how well NT-H1 did in the tests it still isn't viable as a long-term (>3 months) replacement for the TIM that interfaces between the silicon and whatever metal (be it IHS or HSF) will be employed to the shunt heat away from the silicon because of the pump-out issues with NT-H1.

Absolutely! I agree totally! But it's still an issue -- if only for bare-die applications. If NT-H1 performs according to your table of results for BARE-DIE, would adding diamond powder to the mix alleviate the pump-out problem? and would performance be "almost as good" as NT-H1 in the bare-die application?

Like some things, it's probably not worth the trouble. Anyway, there's the abrasion damage brought up during your de-lidding thread, per nano-diamond. Somehow, in some of your recent remarks, I thought you may have changed your view about that.

 

[crashtech]

The pump out issue might be pressure related; since the die has much less area, clamping to the same pressure as an "undelidded" CPU will result in drastically higher PSI at the interface.

 

[BonzaiDuck]

The pump out issue might be pressure related; since the die has much less area, clamping to the same pressure as an "undelidded" CPU will result in drastically higher PSI at the interface.

Nano-diamond would spread out into an unmovable lattice at some point. That very advantage also raises the specter of abrasive damage to the silicon.

If it could be proven that there is no risk of abrasion-damage, then the lattice might inhibit NT-H1 pump-out.

Otherwise, there's some turn-around in the results between NT-H1 HSF-IHS applications, and the bare-die experiments. First thing that comes to mind there: whether we're measuring core temperatures for a processor of traditional indium solder under the IHS or a putty-TIM; second thing, we've inserted a copper shim (or rather -- we removed it) for one of the situations tested.

Now I'll have to go back to that review and look closely at their test-bed. I don't think it was an Ivy Bridge or Haswell using a putty. Maybe. I rather doubt it, though.