Shimm for delidded haswell to get safe direct die cooling?

[stuff_me_good]

Remember these?






Delidding haswell has been on my mind for a while and if you remove IHS, what's the point use it anymore other than that to protect your die? Besides it adds more one extra layer making heat conductivity worse than just direct die contact with cpu cooler.

I'm really surprised that no company has made one of those shimms available for ivy or haswell. Or no one has even done their own DIY version of it.

 

[DigDog]

..because solder?

Ivy is soldered, you can't remove ths IHS unless you go with thermal gun and a sack full of man-balls, and so every die since forever has been covered by heat spreaders. Ergo no need for shimmies.

But, it would be nice to have one right now, thoug h i doubt anyone would get around manufacturing some. How many people do you think have delidded their CPU ...

 

[Tristor]

..because solder?

Ivy is soldered, you can't remove ths IHS unless you go with thermal gun and a sack full of man-balls, and so every die since forever has been covered by heat spreaders. Ergo no need for shimmies.

But, it would be nice to have one right now, thoug h i doubt anyone would get around manufacturing some. How many people do you think have delidded their CPU ...

You are incorrect. Ivy Bridge is not soldered. Sandy Bridge was the last to be soldered. Both Ivy Bridge and Haswell are sealed using black silicon and can be delidded safely. Delidding Ivy Bridge is actually easier and safer than doing Haswell because of the lack of surface mount components to worry about.

To the OP, why not make your own? It should be relatively simple to get some very thin non-conductive padding (either silicon or get some Fujipoly Sarcon) and then take a 0.032" copper sheet and cut it to fit using a Dremel. Lap both sides, apply padding, put on top of the CPU and press gently towards the PCB, and then attach your waterblock on top of the die and call it a day. As far as I'm concerned though, it's easier to just put the IHS back on top since we're talking maybe 2-3C difference at most if you lap the IHS well.

 

[Headfoot]

..because solder?

Ivy is soldered, you can't remove ths IHS unless you go with thermal gun and a sack full of man-balls, and so every die since forever has been covered by heat spreaders. Ergo no need for shimmies.

But, it would be nice to have one right now, thoug h i doubt anyone would get around manufacturing some. How many people do you think have delidded their CPU ...

Ivy isn't soldered. Sandy is

 

[BonzaiDuck]

IDontCare had addressed this problem in his own way.

I can think of a way to make a gasket from . . . here it is again . . . . foam art-board. I'm thinking if you had to compress the art-board, it would reduce the chance of damaging the die.

 

[Tristor]

IDontCare had addressed this problem in his own way.

I can think of a way to make a gasket from . . . here it is again . . . . foam art-board. I'm thinking if you had to compress the art-board, it would reduce the chance of damaging the die.

That might work pretty well. I was thinking something cut to size like that attached semi-permanently with liquid electrical tape over the components to give them some cushion against the mounting pressure as well as help prevent possible shorts. I was thinking kneading eraser might work pretty good as well and could be custom fit and mold around the die and IHS when you mount it as well, but I was worried it might cause heat buildup due to its insulative properties. In the end I ended up just putting my IHS back on after lapping it. I can't see direct-to-die making enough of a difference to allow either a voltage drop or a multiplier's worth of speed gain.

 

[BonzaiDuck]

That might work pretty well. I was thinking something cut to size like that attached semi-permanently with liquid electrical tape over the components to give them some cushion against the mounting pressure as well as help prevent possible shorts. I was thinking kneading eraser might work pretty good as well and could be custom fit and mold around the die and IHS when you mount it as well, but I was worried it might cause heat buildup due to its insulative properties. In the end I ended up just putting my IHS back on after lapping it. I can't see direct-to-die making enough of a difference to allow either a voltage drop or a multiplier's worth of speed gain.

You sure about that? You'd eliminate one copper barrier on the way to either heatpipe tips or a radiator.

Way I see it, if you're going to de-lid and you use a heatpipe cooler, you might want to keep the IHS unless you studiously avoid putting fans on the cooler. And in fact, those coolers use more pressure than a waterblock.

So it dawned on me that switching to either Ivy or Haswell would require de-lidding, and some sort of water-cooling (even AiO -- Corsair, CM or Kraken) without the IHS. And I figure using either core without de-lidding is not currently worth it -- unless it's a machine for my fam-damn-ily with no mandate to overclock.

On the foam artboard: It might be ideal, but you'd want to cut it precisely with an Xacto knife. That stuff is useful in ways I never thought of before!

 

[thilanliyan]

Forgive the ignorance of a non-delidder, but is the pressure from the mounting mechanisms of most coolers enough to crush/destroy the die? I figured going direct to die would be the best way, without another interface to reduce conductivity.

 

[Yuriman]

Forgive the ignorance of a non-delidder, but is the pressure from the mounting mechanisms of most coolers enough to crush/destroy the die? I figured going direct to die would be the best way, without another interface to reduce conductivity.

I haven't ever damaged a die from direct mounting, and I have used a fair number of bare AMD chips with rather large coolers going back to the late 90's. I would be reluctant to transport a bare-die chip with a tower heatsink with the system in an upright position, but if it's just sitting under your desk, you have very little to be concerned about.

 

[BonzaiDuck]

I haven't ever damaged a die from direct mounting, and I have used a fair number of bare AMD chips with rather large coolers going back to the late 90's. I would be reluctant to transport a bare-die chip with a tower heatsink with the system in an upright position, but if it's just sitting under your desk, you have very little to be concerned about.

We're speculating about damage, only knowing that a waterblock seems safer for such pressure. It's probably much less a risk over a longer period of time with less weight (and torque) from heavy fans on the cooler. But like you say, transport comes with its own risk of vibrations and jolts.

If you want to "go after that last grain of rice" -- it's not much of a problem. Various mod-shops have sold brackets for mounting fans from the PCI-E backplate screws, and I'd purchased some cheap Sunon fans that came with little brass brackets of a similar type for such mountings. So you could tap a hole in some available mobo real-estate, or in the vertical post of the HDD cage to secure the bracket, then mount the fan so it is maybe an eighth-inch removed from the cooler fins. You could even use Akasa noise-deadening foam rubber wrapped around the fan to block leakage from the gap.

 

[serpretetsky]

how different is this from the amd xp days? I know i never bothered with shims (along with many others) for my amd palamino 1800's, amd palamino 2100, and amd barton 2500. I don't recall having any problems with these processors.

Is there a pressure difference with modern heatsinks/blocks?

 

[stuff_me_good]

How hard is it to release processor with shimm already in place? Apparently too hard, but AMD has managed to do this in gpu's for years now without any problem.




So the question is, why not cpu's?

 

[serpretetsky]

How hard is it to release processor with shimm already in place? Apparently too hard, but AMD has managed to do this in gpu's for years now without any problem.




So the question is, why not cpu's?

you mean realease identical processors with the heatspreader removed and a shim installed instead? Not being sarcastic, just wanted to clarify.

 

[stuff_me_good]

you mean realease identical processors with the heatspreader removed and a shim installed instead? Not being sarcastic, just wanted to clarify.

Yes.

 

[Tristor]

Because there's not a large enough market for them. The vast majority of people would rather have the IHS installed. Heck, even I would rather just have the IHS installed, although I'd prefer to receive them in two separate pieces if Intel can't be bothered to continue soldering them on to save me the trouble of delidding. Doing direct-to-die cooling as commonplace fell to the wayside for a reason. It's risky, because changes in pressure across the die can cause damage. For instance someone with a large tower cooler inside of a tower chassis. You attach everything with the motherboard laying horizontally, only to then have it be vertical for a long period of time.

Why risk it for very little market?

 

[serpretetsky]

I have never damaged my 3 shimless bare amd processors (and my brother and cousin never damaged their shimless bare amd processor). I don't know how on earth you damage a cpu with a shim on it. Jam it with a screwdriver?

I would definitly buy the shimmed version. Heatsink manufacturers would have to keep in mind that people can have two different versions of the processor.

 

[DigDog]

Sorry for my blunder, it was really late here at work. Yes sandy is soldered, Ivy isn't.
Still, the rest holds true - few people delid (or at least, before this haswell TIM crisis), and there isn't a market for them.

 

[know of fence]

Lids were a good idea when Indium solder was cheap, they hide the die and it's sometimes shrinking size as well as for a few other reasons, but as demand increased fueled by Solar Panels and LCDs, Intel tried to replace it, and surreptitiously pass blame on the increased transistor density.

Consumers were stuck with worse thermals and a sizeable increase in peak power consumption. It's alarming how to this day the usual suspects, calling themselves tech review press, kept completely mum on this subject. In fact when Ian brought delidding up on the podcast he was met by Anand with pretty awkward silence.

It's not the first time companies saved cost by increasing the consumers power bill. In fact before the UuP Directive, that's what makers of electronics did for a living.

Selling shims for delidded processors is obviosly a pretty crazy thing to do, though one could just punch them out of aluminium sheets. The long term solution to the problem is to politely ask Intel to stop enraging the custom builders as well as delivering worse than necessary product to every single desktop user and sell a (shimmed?) bare die version. Maybe we should also nag the press to relay this message to Intel as often and as emphatically as possible.

From Wikipedia article on Indium:
Demand increased as the metal is used in LCDs and televisions, and supply decreased when a number of Chinese mining concerns stopped extracting indium from their zinc tailings. In 2002, the price was US$94 per kilogram. The recent changes in demand and supply have resulted in high and fluctuating prices of indium, which from 2006 to 2009 ranged from US$382/kg to US$918/kg. It has been estimated that there are fewer than 14 years left of indium supplies, based on current rates of extraction, demonstrating the need for additional recycling.