Water to silicon cooling.

[Quickcorepro]

Hello all-

Long time reader first time poster.
I have done some looking around and searching of threads here and other places but I am struggling to find the information I am looking for.
I know you folks really know your stuff so I made an account to ask.

So here it is. I bought myself a 6700K processor. I watched a few videos on delidding the cpu to apply a higher quality metal paste. It looks pretty strait forward.
I started thinking, why add the IHS back on at all? Why not make a new IHS that allows the coolant direct access to the chip.
I am a manufacturing engineer. I drew up a 3d model of the new IHS. I added a double o-ring sealing face.
My concerns are not about sealing the unit. It's what the risk factor of shorting out the chip would be. I know that the coolant isn't conductive but I thought I would get your thoughts.

Thanks-

 

[Sergeant_Sneaky]

https://www.youtube.com/watch?v=LvS_qgn7OSs

Delidding a Skylake chip isn't worth it, it doesn't really make a difference.

 

[assassin445522]

https://www.youtube.com/watch?v=LvS_qgn7OSs

Delidding a Skylake chip isn't worth it, it doesn't really make a difference.

took the words out of my mouth

Well i guess keyboard lol

+1

 

[synphul]

Coolant isn't conductive at first, though it definitely can become so. Delidding isn't really recommended and from the sounds of it you want to make a custom waterblock where the coolant flows directly over the exposed cpu cores? I see this ending badly and a rather healthy investment in a cpu turning into a buying a new one. I just don't personally see the advantages, only downsides.

Heat isn't really that hard to handle on skylake or previous haswell/devil's canyon cpu's. They typically run out of vcore headroom when overclocking before hitting a thermal barrier. Keeping vcore within safe limits is just as crucial, you can't just continue adding voltage regardless of how cool the temps remain. It will still cook the cpu.

 

[Quickcorepro]

Let me better explain.

As you folks mentioned the Sky series is a cool running processor. I understand that. I also understand that it's risky and so on. Let me ask you to play along, if you will.
Currently we use only the face of the silicon core as our contact point to transfer heat. That would be ok if it were the only surface that was hot.
As Ohio SCUBA diver I understand all to well how well water leaches heat. I also understand how any surface not in direct contact with the water stays warmer. That's all real basic.
Consider this-
Inside of the IHS on ALL CPU'S a "pocket" is cut 80% of the face of the heat shield. That creates a void. Air transfers heat at .075 - If we were to simply design a new heat shield that had a pocket of "minimum" size to allow the core to fit, we increase our contact surface, albeit against the board, thus increasing our heat transfer.
It's silly not to use the entire available surface as a heat sump. Think about it. We make our water blocks larger than the IHS because we want to sump away the most amount of heat.
Take that idea one step further, if our coolant were allowed to touch the entire chip face our thermal dynamic grows exponentially.
Even if direct cooling weren't possible, doesn't it just make good sense to have the IHS touch as much of an area around the core as possible for efficiencies sake?

Thanks-

 

[Sergeant_Sneaky]

I see what you mean but the thing is, you are going to be hitting the "dangerous voltages" if you are planning to push the CPU that far (I assume this is why you want to do this), and having high voltage isn't good for the CPU no matter how cool it is running. If you're not planning to push the CPU further than you can with standard water cooling, why are you doing this?

 

[synphul]

Maybe some other engineers can weigh in on the matter. My theory is that between intel and amd they have some pretty talented engineers of their own. There must be other factors at play. Part of which aside from cooling is to protect the actual cores underneath and add strength to the cpu package. I've heard of people crushing the physical cores when delidding since there's no longer any structure between the cores and cooler base (assuming they left the ihs off for direct contact).

Even if there may be technically more efficient cooling perks to altering the basic design everything comes with tradeoffs. It reaches a point where better cooling hits diminishing returns when the chip is already running well under its max thermal rating. For example an intel cpu with a throttle limit of 105c, what many consider a 'safe' upper limit of around 70-80c for sustained operation. If that cpu is already staying at 55-60c under full load with the ihs intact as designed, what is the benefit to customizing everything to get it to 50c?

Water blocks tend to benefit from being a bit larger but the cores themselves make up for a very small amount of the surface area under the ihs. Also if you take a look here at delidded skylake cpu's you can see the cores are designed in such a way that they have an incredible amount of surface area under the ihs. They're so thin there really aren't any 'edges' to speak of trapped in dead air space. Trying to encompass more than the surface by including the 'edges' of the core area is almost a moot point.
http://www.overclock.net/t/1568357/skylake-delidded

In fact it does touch as much of the core as possible by default, it's not as though the ihs is only in contact with one face of a cube shape leaving many meaningful surfaces ignored.

 

[Quickcorepro]

I took the lid off. Here are the hard specs.
The core size is .420 wide by .520 long
The inside of the IHS the "recess" is 1.018 wide by 1.110 long with a pocket depth of .012

Sorry Synhul- The numbers don't lie. The heat spreader does NOT touch anything accept the face of the core. So let's be objective and use our minds. Intel and AMD make products for the masses not for the extreme hobbyist. Rather than follow the masses why not think outside the box with me for a bit and see where this could go. If for nothing more than for fun.

Ok, given the dimension above we can see that there is at least a .012 air gap on just over 1/2 of the heat spreader. In my minds eyes it would be simple to machine a heat spreader with a pocket of say .480 x .580. That increases the surface area in contact with the Phenolic board by roughly 2x the previous surface area. Of course Phenolic will not transfer heat of the same rate the silicon will for obvious reasons. However it will transfer heat that previously was un transferred to the IHS. I am betting its enough to be worth investigating.

Thanks-

 

[fabiodrm]

do it! send us pictures and temp result!

 

[synphul]

Sounds like too much effort trying to reinvent the wheel to me. I'm not sure there's 2x the surface area, check the depth of the core itself by the length on all four sides, add them up and compare the face surface area. Not even close. The majority of the heat is transferred through the face of the core area. As simple as it is to machine a custom ihs I'd be curious to see the results you get. No way to know unless you give it a whirl.

Maybe we could make the silicon wafer connect to the socket via wifi and 'float' the entire wafer in liquid for complete 100% immersion with no pressure at all on the delicate cores themselves.

 

[Quickcorepro]

You can do the math but my cad program says the area of the pocket is 1.279 in^2 and the area of the silicon is .230 in^2. Also, I made a mistake on my original width size. I used a scale to get rough dimensions. After using digital calipers it is actually .370 wide, not the .420 I originally said. That being said the above area .230 in^2 is based on .420 wide. If I use the .370 as the width the area actually shrinks .200in^2 making it MUCH more than double.However, to be fair, you wouldn't be able to use the exact foot print of the silicon as your pocket size. (As I mentioned above)

this is what I am thinking. Feel free to give your thoughts.

I am going to use a CMM to measure the silicon height. I am interested to see just how much "headroom" that intel figured into the design. It also tell what the average thickness of the thermal paste would be based on the depth of the IHS pocket.

I was planning on using CLU on the silicon face but use a non metallic based paste between the phenolic board and the newly designed IHS. I sill plan to keep the outside lip of the IHS to the inside pocket depth the same (to prevent crushing the core, as Syn pointed out) I plan to use the same "headroom" between the board and the IHS that Intel used on the IHS to core. I wasn't planning on re bonding the IHS back to the chip. Anyone see any issues with that?

I plan to install the CPU and see what temps are. If they have made a significant drop then it would be proof of concept for attempting direct cooling the silicon.

Thanks-

 

[Quickcorepro]

http://www.bit-tech.net/news/hardware/2015/10/06/direct-to-silicon-liquid-cooling/1 interesting read