Mounting Noctua heatsinks to bare die?

[Hi-Fi Man]

I have a Noctua NH-C14S that I want to mount bare die on my i7 5775C. Does anybody have any experience with using Noctua's mounting system to mount on a bare die?

I saw this picture from a thread on [H] from a user who mounted his D15 bare die. How would I go about this?

 

[mikeymikec]

Was it haswell or skylake that Intel started using thinner PCB for their CPUs? I remember reading an article showing a bendy CPU (not intentionally bendy).

 

[Hail The Brain Slug]

In looking at reviews, it would seem the 5775C has a heatspreader. Are you prepared to delid it in order to mount a bare-die heatsink?

Even if you are prepared to delid it, consider taking the conservative approach and replacing the under-heatspreader TIM with Liquid Ultra. That way you get improved thermal transfer while still keeping the original heatspreader and cpu hold down mechanism, allowing you to simplify heatsink mounting. This also reduces your chances of damaging the die with the heatsink.

 

[BonzaiDuck]

Was it haswell or skylake that Intel started using thinner PCB for their CPUs? I remember reading an article showing a bendy CPU (not intentionally bendy).

Since you brought it up, Mikey, I can point him to my thread about a "rumored advisory," which only revealed itself to me through the ThermalRight website. I think there were a couple pictures. Tech-support people, perhaps from ASUS, joined the discussion in some user-forum queries.

People outside TR said it was a "shipping, transportation or careless handling situation" to make the weight of a heatsink and fans change from a static torque-loading to a dynamic one. Those who first raised the issue with the photos spoke to the 400 gram spec which goes back many generations to early last decade in Intel spec sheets. There is a spec for pressure with the fittings of a heatsink on the CPU and socket: 50 lbs.

Now, that being said, I had a discussion with the tech-support (maybe proprietor?) of Silicon Lottery. He fell in with the rest here on whether putting a heatsink and fan on a bare die was a good idea. And if it were a good idea, I might have done it years ago.

But it begs the question. This isn't just about weight; it's about torque. And torque is the product of weight at a center-of-gravity or center-of-mass multiplied by the distance from a pivot-point, or in this case -- the distance from either the IHS or processor die after the delidding.

So there definitely ARE HSF's -- possibly to include yours -- which have a low profile and reduce the torque considerably or at least significantly.

And this is why it might make sense. Myself and the few oddballs like me may not care about those aspects of computer design which may cause you to compromise from picking the heatsink with the lowest thermal resistance or the lowest spread between idle and load temperatures under some test-bed status-quo. And if the cooler is good enough, and if you could get a 10 to 20C improvement in temperatures and heat removal, it may also offer value for a compact case -- a small case.

Blah blah blah. Sorry, my fingers ran away with my thoughts.

 

[DrMrLordX]

I talked to a guy who mounted a D15 to a bare-die AMD CPU. He complained that he had to modify the mounting mechanism to accomplish the feat since the default configuration prevented the heatsink from being able to make contact with the die. I suspect you would encounter similar trouble with an Intel CPU.

So I will second the notion of using the IHS + superior TIM (CLU/Phobya Liquid Metal/Grizzly Conductonaut) as a way to improve temps. You will do better with bare-die, but you will have to modify some stuff, and that may make your D15 incompatible with a "normal" CPU.

There were no obvious reports of thin PCB problems for Broadwell-C that I can recall.

From my own experiences with using an NH-D14 on a delidded/relidded A10-7700k, I will offer you a warning: CLU basically welded the IHS to the lapped copper base of my HSF. I had lapped both the NH-D14 (actually I did that back in . . . 2009?) and the IHS, and I used CLU on both sides of the IHS for maximum thermal transfer. There was no particularly interesting bond between the die and the IHS, but the copper <-> CLU <-> copper bond became very powerful, causing the IHS to come off with the HSF when I finally retired that CPU. I had to heat up the stuck IHS with a hair dryer and twist it off with a wrench.

 

[ehume]

What a Noctua guy told me was that Noctua's spring loading system was to apply exactly what Intel specified -- no more, no less. Applying greater pressure can improve temps (I showed that when I reviewed the Fuma heatsink) but leaves the user vulnerable when Intel uses thinner PCB's. So when Intel put out thinner PCB's starting with Skylake, it was "virtue is rewarded." Of course, since you use a Noctua heatsink, distorting the PCB shouldn't be a problem.

 

[Hi-Fi Man]

How good is Noctua's included paste for applying underneath the IHS? I still would rather go bare die I just need to know how I would go about modifying the mounting system on this heatsink.

 

[Hail The Brain Slug]

How good is Noctua's included paste for applying underneath the IHS? I still would rather go bare die I just need to know how I would go about modifying the mounting system on this heatsink.

If you're convinced you want to take the risk of the heatsink pressure being applied directly to the die, understanding that if you have it in a case with the heatsink hanging over it that it WILL apply uneven pressure and potentially damage your core...

It looks like he removed the C clips that hold in the mounting screws. He added washers to increase the spring tension to make up for the gap where the heatspreader was, and mounted the heatsink as the hold-down mechanism for the CPU itself after removing the original hold-down.

The thickness of those washers you would have to guess or measure yourself.

This seems like a very, VERY bad idea.

If you had a case that mounted the motherboard flat and the heatsink stood straight up, you would minimize the risk from weight shifting.

If you went the delid-relid route, make sure to go the extra mile for CLU/Liquid Ultra under the lid. You don't want to skimp out after all that work.

For everyone's reference, the thread is here.

 

[MongGrel]

How good is Noctua's included paste for applying underneath the IHS? I still would rather go bare die I just need to know how I would go about modifying the mounting system on this heatsink.

I've never tried what you are doing, but I imagine it is NT-H1 paste and have been using that for most things myself for a long time.

Have bought tubes of it on the side over several others the last decade, it's a very good safe general purpose paste.

But your sticking your neck out a bit just going bareback to begin with.

Seems really not worth it these days, but to each his own.

 

[Hi-Fi Man]

Temps are pretty up there (case doesn't have great air flow) so at the very least I'm going to delid. At this point I'm just going to stick some noctua paste under the IHS and call it a day.

 

[PliotronX]

That is actually my picture of my D15

Yes, I had removed the clips to be able to insert the washers. Later I needed more pressure so placed thicker washers there:

It is always worth it to me to remove the noob shields because all those do is reduce heat conductivity. I went from brick walling 4.6GHz at 1.32v to solid 4.7GHz at 1.285v and could suicide at 4.9GHz. A similar story with my Q6600 which I did spend hours lapping the IHS to a mirror as well as the TRUE heatsink which only dropped a couple of degrees and no speed or voltage improvement. Desoldering that IHS and mounting baredie netted nearly 100MHz at lower voltage as well. Just be deliberate when mounting and minimize rocking, this means placing one hand on top of the sink to counteract pressing the first side into mounting position. I have had to remove the metal retention bracket on my LGA boards to run baredie. After that, the D15 mated just fine. If you are uncomfortable or end up using so much pressure that it doesn't make sense, don't do it. I only do it because I have been delidding since the Tualatin and have never been disappointed nor lost a processor due to it.

 

[BonzaiDuck]

A cautionary note: I just remembered this after reviewing the thread today.

Tech-support guy at Silicon Lottery told me that bare-die installation is not so feasible on socket 1151 Skylake, because the socket itself higher than the die. You would have to trim the square plastic around the socket.

All the rest of it makes sense. And if this factor has no bearing on the socket for the OP's processor, the risk of damage with a low-profile cooler is less than for something like a D15 or a TR LG Macho.

 

[Hi-Fi Man]

That is actually my picture of my D15

Yes, I had removed the clips to be able to insert the washers. Later I needed more pressure so placed thicker washers there:

It is always worth it to me to remove the noob shields because all those do is reduce heat conductivity. I went from brick walling 4.6GHz at 1.32v to solid 4.7GHz at 1.285v and could suicide at 4.9GHz. A similar story with my Q6600 which I did spend hours lapping the IHS to a mirror as well as the TRUE heatsink which only dropped a couple of degrees and no speed or voltage improvement. Desoldering that IHS and mounting baredie netted nearly 100MHz at lower voltage as well. Just be deliberate when mounting and minimize rocking, this means placing one hand on top of the sink to counteract pressing the first side into mounting position. I have had to remove the metal retention bracket on my LGA boards to run baredie. After that, the D15 mated just fine. If you are uncomfortable or end up using so much pressure that it doesn't make sense, don't do it. I only do it because I have been delidding since the Tualatin and have never been disappointed nor lost a processor due to it.

The man himself! How hard is it to remove that clip in the picture and what size washers did you use?