Waterblock question

I'm going to be building myself a watercooling setup soon and I had this idea:

What if the core of the CPU was exposed in the waterblock? Let's assume it's water tight. What would the performance be like? I know that the pump would have to be on before the computer was turned on, but it seems to me that it would work well.

Yeah, it's called direct die cooling...a think quite a few people have tried it, it's supposed to produce pretty good results. There might be some threads in Extreme Cooling...not sure. Try using the search for direct die cooling.

I'm pretty sure that's not too uncommon, it's usually referred to as "direct die" contact. The performance gain is quite nice because you have no type of insulater against the heat transfer of the water, so , not extactly but pretty close, the C/W of the block would be close to that of water it's self. I have actually been thinking on doing this for a while but have not gotten anything up in CAD yet, something like where the opening to the core would just be within thousands of an inch smaller than the core itself, and with a nice flat/lapped surface and good mounting fource it would be watter tight, you could even use a little RTV to be safe. But yeh it's very possible / has been done.

Thanks for the info, guys. I can't wait for my new shop to be done so I can make one of these suckers.

One thing you need to be careful of though...using a cooling setup like that can result in the actual die corroding...and that's not a good thing.

actually, I read an article where the guy had problems with water saturating the core. I mean the core was acting like a sponge and slowly sucking the water in. might wanna watch out for that...

From what i have read diect die is NEVER long term

as littlepiggie said the die soaks in the water and dies (haha)

if you have a good supply of processors go for it but you will kill the chip

I personally don't see how direct die could perform better than a cascade waterblock. It seems pointless to me, especially when you consider the fact that any direct die cooled cpu WILL eventually die. There are just so many better ways of cooling.

squeakygeek said:

I personally don't see how direct die could perform better than a cascade waterblock. It seems pointless to me, especially when you consider the fact that any direct die cooled cpu WILL eventually die. There are just so many better ways of cooling.

Click to expand...

it'll die?? Cool!
never knew that.
yep, direct die sucks. copper transfers heat faster then water does, waters just good for its thermal capacity. conduction, convextion, radiation, i think it was. water is a bad thermal conductor. put a pot of water on a stove top and it'll take a freakin hour to boil (i hate waiting for my morning tea to boil!). metals on the other hand... damn, hot stuff. If only we could use mercury in w/c.. then you could direct die cool, but u have to be pretty insane to try it, stupid as well.. but i guess stupidity isn't a factor in insanity, yes/no?

a pot of water on a stove top and it'll take a freakin hour to boil

Click to expand...

That would be a result of heat capacity.

If only we could use mercury in w/c

Click to expand...

Mercury is too heavy - you wouldn't be able to pump it enough. Water is king.

Ask JFettig about it. His core lasted a couple of weeks before going off to AMD heaven. Either the core is soaking up the water, or the epoxy holding it in isn't waterproof as it once was.

I tried it once, and the chip lived (only just). I heard a high pitched rumble like boiling water and shut down. Replaced that setup with a real block. One of the surface capacitors went up in smoke though, don't know why it still worked.

Also to note, the substrate isn't water tight as it had been on earlier chips, so that's something to design for as well.

Order some keychain making supplies, you'll have some shortly

squeakygeek said:

That would be a result of heat capacity.

Mercury is too heavy - you wouldn't be able to pump it enough. Water is king.

Click to expand...

awww, but .. but .. but
mercury is so fun...
all the idiots that experimented with it loved it before they turned into vegetables. i'd do it, but all i have is a 50ml bottle works nicly in my vintage blood pressure thingy, its actually pumped by air pressure.
good stuff

Korndog said:

awww, but .. but .. but
mercury is so fun...
all the idiots that experimented with it loved it before they turned into vegetables. i'd do it, but all i have is a 50ml bottle works nicly in my vintage blood pressure thingy, its actually pumped by air pressure.
good stuff

Click to expand...

ah where did you get that?

my dad is a doctor
he has a few, but i only found one in the garage.
i remember when i was little the glass broke and i poured it out on my hands, it was heavy..
it ran through the cracks of my fingers and into the carpet it disappeared.
i think that little incident would explain why i'm so stupid now (i got a C in precal.. GRAHHH!)

So it always kills the CPU? Well, that sucks. I guess I'll just have to make a real waterblock then.

good thinking

It used to work, but I think they changed materials. Besides with the wattage of the cpu's now .vs the shrinking die size, a chip doesn't have enough surface area to be directly cooled with water well.
There are still a few articles around about direct die cooling, but watch what speed the chip is to determine the age of the article.

Some things about the good 'ol days are to be missed...sigh.

Water-block technology has also progressed dramatically since "the good ol' days".

Back then, there wasn't a whole lot of difference in terms of waterblock technology between running some water over the CPU, as opposed to running some water over a piece of aluminium (which is fairly common back then) or copper sitting on the CPU. The waterblock of course incurred the extra cost of additional thermal gradients through the thermal paste and the waterblock material, typically meaning that direct die was better.

Nowadays, waterblocks have progressed to the point where their internal shape and flow patterns are specifically crafted to maximise thermal transfer between the waterblock metal and the water. A flat die CPU has none of that at all, and in fact, simply can't because it is a flat piece of silicon.

So where the wins used to be for direct die was simply due to there not being something else between the water and the CPU, that story has changed where what a waterblock loses in extra thermal gradients, it can more than make for in terms of improved thermal transfer efficiency as opposed to simply working with a small flat piece of silicon which does not lend itself well to the waterblock techniques being employed today.

Worse, the CPU's porosity is such that even if one attempted to apply modern waterblock techniques to the CPU, the CPU will die very quickly.