return of the silver plated cpu

well folks i doubt many of you remember when i first had the idea to silver plate the top of a cpu in order to water cool it using a d2d block and not have the ill effects of water coming into direct contact with the cpu substrate and die. but that's basically the idea. there would be NO HEATSINK. let me say one more time NO HEATSINK. ok thats out of the way. some people seem to have trouble visualizing what i mean so i will explain in great, and eloquent might i add, detail. /sarcasm. ill block it out with ### in case you already understand.

######################explaination of idea########
start with a direct die waterblock, one that sprays water directly onto the die. in the regular direct die setup, the water comes in, contacts the cpu die (core) itself, swishes around a bit to pick up extra heat, and subsequently leaves the block. the same ideas apply as in regular waterblocks, such as turbulence etc, but now there is no bottom, so you have reduced the thermal gradient. (no need for that messy cpu paste either... )

the substrate around the actual die will be coated first with a sealant, courtesy of my new workplace, that is electrically non-conductive and almost completely non-reactive to any chemical. good stuff to have if you ask me. also, it will hold up to 450 celcius, so no problem there. the point of this is to seal the bridges and components that there may be on top of the processor from the next step so as not to bridge contacts that should remain unbridged.

next, add conductive silver paint. ok the way silver plating works, is you immerse the victim in a "bath" that has silver atoms dissolved in it and run a current across it, with the anode being pure silver (dipped into the solution) and the piece to be plated being the cathode (also dipped into solution). the solution is an electrolyte, so the current is carried across it and you complete the circuit. this current deposits a thin layer of silver on the cathode (your work piece). because of the need for continuity of a circuit, the cathode must be conductive...not hard to figure out. heres where the (very expensive) conductive silver paint comes in. but i dont want to bridge the bridges on top of the processor, hence the sealant. so after putting on sealant on everything but the die, spray it down with a very little bit of conductive paint. that stuff was like 50 bucks for 8 oz.

finally comes the electrolyte bath and the current, applied with my work's handy (also very expensive) power supply. and voila, i have a processor that should be something close to flat on top, but with a direct bond, (if you will) of silver between the heat dissipating die and the coolant that will be sprayed on the top.

the whole point of this huge ordeal, as i mentioned, is that the water never comes in contact with the actual cpu, though it almost does becuase the layer of silver is so thin that its thermal gradient is negligable. basically, its "direct die" cooling without the dangers of water absorbtion.

#############################

glad you're still with me here. well the whole point of this thread is that the idea has been revived because of my new job in an electrical engineering lab that has all of the equipment that i was too poor and too uninformed to buy. i was wondering what all of you thought about it this time around. needless to say, i am pretty excited about the prospect of actually carrying out this experiment. i feel sure that the processor will emerge unharmed. I almost forgot to mention that NeoMoses was also very intrumental in helping me develop the idea the first time. he appears to have dropped it as well though. so neomoses... if you see this maybe you could get back at it too. ok now ill quit rambling and leave it up to yall to talk about. i hope ive explained it well enough.

jungle

CRAZY IDEA not many ppl have the resources of knowledge to atempt such a task it shoundz great post pics as soon as its all serup

sounds good on paper, i'd like to see it actually done though...

i believe Cather is making a G5 which uses silver instead of copper for a WB

i don't see the point of coating the core with silver when you could just have it come in direct contact with the water? wouldn't that be just one more layer between the core and the water?

Never really understood why anyone would want to do direct die cooling??

Everyone thats posted about doing it before has pretty crappy results...

Very little surface area will just require even more flow.

You could plate the die in diamond if you want, but you are still only taking heat away from that little 100 mm^2 core...

If you want to improve anything just remove the IHS and mount the WB very carefully to the die...

Sorry if I missed someone else's post or article or whatever but Ive never seen anyone do direct die cooling and get better results than just a good WB.

Edit--

Oh yeah you didnt say anything about the thermal conductivity of the "sealant" you are going to use. Sure it might not melt before 450c but if its a thermal insulator then you are still screwed...

Rather than use your conductive paint, you should use Tollen's reagent to get the initial layer of silver.

he he he this topic always sparks much debate. id like to address some points with my rebuttals in backwards order since that is how it appears here. obviously there will be no hard evidence until i have some numbers, which may or may not ever happen. im really doing this as much for the novelty and fun as for the better temps. nonetheless....

Very little surface area will just require even more flow.

Click to expand...

i think you are missing a detail that i believe will nullify this problem. the die is not the only thing plated: the entire top of the proc is plated. with such a tiny layer of silver, i suspect that heat will travel extraordinarily well to the outer edges of the plating. what i am attempting is, in essence, a waterblock whose baseplate is only a few microns thick and the rest is constructed of poly. the whole purpose, if im not mistaken, of cups in normal waterblocks is to minimize the thickness of metal between the source of heat (the cpu die) and the coolant (the water). i actually think that this bottom will have as much surface area as a normal block, just with a much lower thermal resistance because of the decreased thickness. the only reason people havent made the entire bottom of the wb as thin as the bottom of the cups is that it would make the whole thing structurally unstable and incapable of withstanding mounting pressures. this bottom, on the other hand, is directly interfaced with the processor itself, eliminating this problem.

Oh yeah you didnt say anything about the thermal conductivity of the "sealant" you are going to use. Sure it might not melt before 450c but if its a thermal insulator then you are still screwed...

Click to expand...

i have thought about this. and that is the reason i didn't seal the actual die, which is the only significant source of heat. since the sealant is not covering the die, there will be no problem with its thermal transfer properties.

i don't see the point of coating the core with silver when you could just have it come in direct contact with the water? wouldn't that be just one more layer between the core and the water?

Click to expand...

agreed, except that direct exposure of the core to water has been shown to cause cpu failure. that is the reason for the plating, to eliminate actual contact but keep most of the benefits. because the plating is only a few microns thick, the difference should be negligable. i actually plan to seal the area around the core and not plate it at some point to see if i get better results (in case it is only the substrate that absorbs water), but i suspect that the added surface area of the plating would make for better numbers.

update: the sealant i used, i discovered from the guy that gave it to me, does not dry hard, but is more flexible like caulk. my only worry about this is that the plating might crack if enough pressure were put on the top of the processor, maybe not much at all. if worse comes to worse, i can scrape off this sealant and go for the one i originally intended to use.

and swampthing, what is this stuff you speak of? i am just following directions from the plating company with the paint. it doesnt necessarily have to be silver, thats just what they sell.

Why not gold? much more resistant to corrosion, at a few microns thick there should not be a diff in conductivity. Isnt the electric potential between silver and copper like 1v? With just a coating it wont take much to corrode away all of the silver......

The reason for the cupping on waterblocks isn't to decrease the thickness of the plate...it's to create turbulence. The only way your idea will work better than that of a regular waterblock is if you somehow create turbulence. You need to create turbulence over the CPU die for water to absorb heat. The only way you could make this design work, is if you somehow built a waterblock with the bare CPU die (with your coating) as the base of the waterblock and the top of the block had channels etc. to create turbulence over the CPU die.

Tollen's reagent is a mix of chemicals that deposits silver on damn near any surface, commonly used in mirror making. It won't get you as thick a layer as you might need, but would be an excellent starter for the electroplating.

Jungle said:

next, add conductive silver paint. ok the way silver plating works, is you immerse the victim in a "bath" that has silver atoms dissolved in it and run a current across it, with the anode being pure silver (dipped into the solution) and the piece to be plated being the cathode (also dipped into solution). the solution is an electrolyte, so the current is carried across it and you complete the circuit. this current deposits a thin layer of silver on the cathode (your work piece). because of the need for continuity of a circuit, the cathode must be conductive...not hard to figure out. heres where the (very expensive) conductive silver paint comes in. but i dont want to bridge the bridges on top of the processor, hence the sealant. so after putting on sealant on everything but the die, spray it down with a very little bit of conductive paint. that stuff was like 50 bucks for 8 oz.

finally comes the electrolyte bath and the current, applied with my work's handy (also very expensive) power supply. and voila, i have a processor that should be something close to flat on top, but with a direct bond, (if you will) of silver between the heat dissipating die and the coolant that will be sprayed on the top.

Click to expand...

how will you conduct the current through the CPU die?

gongo56 said:

i believe Cather is making a G5 which uses silver instead of copper for a WB

Click to expand...

jungle is talking about direct die, not a WB

Jungle said:

i think you are missing a detail that i believe will nullify this problem. the die is not the only thing plated: the entire top of the proc is plated. with such a tiny layer of silver, i suspect that heat will travel extraordinarily well to the outer edges of the plating. what i am attempting is, in essence, a waterblock whose baseplate is only a few microns thick and the rest is constructed of poly. the whole purpose, if im not mistaken, of cups in normal waterblocks is to minimize the thickness of metal between the source of heat (the cpu die) and the coolant (the water). i actually think that this bottom will have as much surface area as a normal block, just with a much lower thermal resistance because of the decreased thickness. the only reason people havent made the entire bottom of the wb as thin as the bottom of the cups is that it would make the whole thing structurally unstable and incapable of withstanding mounting pressures. this bottom, on the other hand, is directly interfaced with the processor itself, eliminating this problem.

Click to expand...

I suggest you read up on baseplates, there is an optimum thickness and it is not 0. The cups are there for turbulence, not to reduce thickness.


Good ideas, experimentation is the only way to know for sure

Jungle said:

he he he this topic always sparks much debate.

Click to expand...

while Im all for experimenting and having fun making stuff, but I like to keep the moto of "work hard AND smart" at the top of my list.
This is more of a correction of misconceptions rather than a debate.

Jungle said:

the die is not the only thing plated: the entire top of the proc is plated. with such a tiny layer of silver, i suspect that heat will travel extraordinarily well to the outer edges of the plating.

Click to expand...

To quote from a horrible movie "Assumption is the mother of all ___-ups."

As greenman100 already said the ideal thickness for any metal to conduct heat efficiently is definitely not 0.
Not to mention the fact it doesnt matter if you go from covering the die to covering the die and the are around it. The surface area you would be moving water over is still a fraction of any current well built waterblock. So to stress it even more, you'd need a lot of flow to get a good C/W...

Basically you are just going to be making a low resistance silver waterblock if you want decent performance. I still dont see any advantages to this idea.

No need to mention baseplate thickness now I guess. I'm just curious if the silver really will stick to the surface as it does when plating metal. With the optimum baseplate theory shot down, I don't think there is any reason for it now. AFAIK, the core is water-impermeiable, but the organic substrate isn't. In the past most direct-die was used with the ceramic substrate chips, which can wick water fairly well.

I also don't see how ordinary silver paint will work. Sure, the paint will be conductive, and the silver will stick to it very well, but in the end the paint [which is now plated] is sticking to the core not because it's fused with the core, but because of the adhesion of paint. Tollen sounds like it might work due to chemically being deposited. Best of luck to you, of course, but I don't think it will pan out too well. I hope I'm proven wrong of course, since I'm not well-versed in the world of electorlytic plating

As for direct die being better than waterblocks, due to the fact certain areas of the core heating up more than other areas, direct-die can't hope to beat an excelent waterblock. Thats where the baseplate rules get you again - the hotter parts get cooled better due to the heat spreading out. I look at it like shooting at a large crowd without moving your gun. Close up, you'll just be clearing a straight line - far away the bullets, while not moving as fast, will spray a greater area. Not that I think of that stuff regularly...[the voices just tell me these things ]

HiProfile said:

...I look at it like shooting at a large crowd without moving your gun. Close up, you'll just be clearing a straight line - far away the bullets, while not moving as fast, will spray a greater area. Not that I think of that stuff regularly...[the voices just tell me these things ]

Click to expand...

WOW

Man... dont even know how to respond to that on a sane level...

I like to think of thermal dissipation as people leaving a ball game. The more gates there are the quicker people leave the park, the shorter the path to the cars the faster they get in the car, the wider the roads the faster the cars leave, etc etc... And whenever people stop and have to wait they get mad and angry

Must admit though, I have never thought of killing a large amount of people when I think of heat passing through copper...

maybe a nice jet block glued to the the cpu substrate would be best to get good water velocity?

hmmm i am not used to being "shot down." but im not quite convinced that the idea is as flawed as you all seem to think. i am interested in this optimal baseplate thickness. it seems to me that the ability of a metal to conduct heat would be inversely proportional to its thickness because the heat source would have to heat more volume of metal before moving farther away from its source point. like water in a big tube and water in a small tube. a small tube will fill with water much faster than a big tube? is the analogy making sense? do you have a link about this optimal baseplate thickness?

edit: just found a scientific paper at this site that seems to support my understanding of conductance.

from paper: "First, an ideal heatsink with infinite fin thermal conductivity is considered, providing simple power-law prediction of performance. Next, fins of finite thermal conductivity and thickness, as well as effects of developing flow are included in the analysis, permitting prediction and minimization of weight."

I also don't see how ordinary silver paint will work

Click to expand...

this isnt ordinary silver paint, or at least i hope not. it was bought from the plating company and developed for the express purpose of plating non-conductive substrates.

while Im all for experimenting and having fun making stuff, but I like to keep the moto of "work hard AND smart" at the top of my list.

Click to expand...

easy tiger, no need for veiled insults to my intelligence, though perhaps what you said about lack of surface area is correct. again, i have thought about this some, but did not mention it because it was more on the subject of waterblocks. but the top of the waterblock would have small jets that came in at opposing angles. about 5 jets on either side, some angled to intersect just above the cpu die. i think that this opposed waterflow will create a lot of turbulence. also, i have closed off part of the area through which the water can flow out by leaving a fairly wide partition going down the center of the outlets. i hope that this will cause even more turbulence in the water, allowing as much of the surface area of the water itself to come in contact with the heat source.

The reason for the cupping on waterblocks isn't to decrease the thickness of the plate...it's to create turbulence

Click to expand...

i was basing my understanding of the purpose of the cups on http://www.procooling.com/reviews/html/waterblock_comparison-_rbx_whi.php

from the article:
"The jet-in-a-cup effect creates a double-impingement action of cooling the very base of the cups that are directly above the CPU. The water then slams into the base of the walls of the cups cooling any heat that conducts up the cup walls. Almost all of the CPU's heat gets dissipated within 2-3mm from the CPU itself [emphasis mine], this living up to the initial stated design philosophy. When the water exits the cups it flows away without affecting any other jet's activity. Much experimentation was required to find the correct balance of jet width and height, cup width and height, inter-cup geometry, and then find a way to get the whole lot affordably machined."
-Cathar



more discussion welcome.

Your getting close but not close enough. You get to thin with the base plate, paper thin as you have suggested, the more you rely on water alone to handle the heat load. The thermal properties of water is way worse than silver. Thermal conductivity of silver 430 W/m-k compared to water 0.6 W/m-k. The heat would travel so fast through that thin silver the water couldn't effectively absorb it away fast enough causing higher temps. The reason why direct die water cooling sucks. You would have to have a huge flow rate to get anywhere.

You guys say he needs a huge flow rate to get good performance. Why is that a bad thing. For the money that someone is spending to do this you could well aford a big pump and a low restriction loop.

The water block is the most restricive part of the loop. He can design his block to have a very low resistance and then run paralle heater cores. This with a big pump will give killer flow rates.

About thermal codutivity. Water doesn't need to have good conduction because we pump it. The water has to pick up heat and hold it until it gets to the rad. The pump circulates the water so the heat doesn't have to move from one moulcle to another be cause they themselves are moving

Jungle said:

hmmm i am not used to being "shot down." but im not quite convinced that the idea is as flawed as you all seem to think. i am interested in this optimal baseplate thickness. it seems to me that the ability of a metal to conduct heat would be inversely proportional to its thickness because the heat source would have to heat more volume of metal before moving farther away from its source point...

Click to expand...

No.

An engineer needs to help me on this but I am pretty sure thermal conductivity is proportional to thickness much like electrical conductivity.

The thicker the better conductivity but the longer the distance the higher the resistance.

zebkoolindc said:

You guys say he needs a huge flow rate to get good performance. Why is that a bad thing. For the money that someone is spending to do this you could well aford a big pump and a low restriction loop.

The water block is the most restricive part of the loop. He can design his block to have a very low resistance and then run paralle heater cores. This with a big pump will give killer flow rates.

About thermal codutivity. Water doesn't need to have good conduction because we pump it. The water has to pick up heat and hold it until it gets to the rad. The pump circulates the water so the heat doesn't have to move from one moulcle to another be cause they themselves are moving

Click to expand...

Beeeecause you can either do all this work, spend all this cash and get decent performance, or you can just buy a water block, buy the same huge pump and get very good performance.

Having huge flow is always good. But having a system that requires tons of it because its poorly designed is not...