My direct-die cooling system

S

Sniper X

Limp Gawd
Joined
Feb 13, 2006
Messages
240
A while back i had inferior air cooling for my cpu and gpu. My load temps were 58C for my cpu and 80C!!! for my unlocked x850pro. My 60mm cpu fan was as loud as hell and the stock cooling on my video REALLY limited its overclockabiliy. So i decided to do something about it, I decided that i would finally watercool my pc(i had been thinking about doing it for about a year or so). I got a bonneville radiator from autozone($19), 250gph waterpump from petsmart($30), copper parts from homedepot($12), clamps and 1/2 clearflex hose from mcmastercarr.com(20$), two 120mm fans and a heatercore shroud from dangerden($40). Total cost was about $121 ,Including the shipping and handling i had to pay on some parts.
I will not go into the details of the whole project, but i will talk about what i did with my cpu waterblock that i made. My cpu waterblock does not have a little copper maze inside of it for water to travel through but instead i have my waterblock directing water over the actual cpu die.
diagram5fj.jpg

I sprayed a clear protective coating over the area around the die to prevent shortcircuting the resistors around it.
My gpu waterblock is simple and it is a copper pipe end-piece with two side holes for 1/2 copper pipes .The inlet pipe goes in all the way to the other end and has holes drilled in it to disperse the water onto the copper plate on its bottom. My gpu is not direct die cooled but i may try it in the future.
My temps for my watercooling system are 40-42C idle,44C load CPU(the temp probe is under the cpu so it is inaccurate, the die temp should be lower). 33C idle,48C load GPU. I think i could get better temps out of my direct die cooling by creating more water pressure over the cpu die.

My loop order is T-line----->pump---->cpu--->gpu---->rad---->

NOW for some pics

dsc009529sl.jpg


dsc009547kd.jpg


dsc009568vi.jpg


dsc009594pn.jpg


dsc009643lt.jpg


dsc009659ms.jpg
 
I think we have our new GOCA spokesperson! :)

Very nicely done, and I'll bet it outperforms most pro blocks, to boot!
 
How does it perform? I'd expect the die to benefit from the increased surface area of more traditional watercooling methods, just the way the die benefits from the increased surface area of a traditional heatsink.
 
You're already over your bandwidth as well (read: pics are down)- how big were they? They didn't look that big before...
 
kumquat said:
How does it perform? I'd expect the die to benefit from the increased surface area of more traditional watercooling methods, just the way the die benefits from the increased surface area of a traditional heatsink.
Click to expand...

Direct die cooling is better because moving water can take heat away faster than if the heat traveled through thermal paste then a copper base then into the water. with direct die cooling you are getting rid of 2 of the thermal mediums that the cpu heat has to travel through. You also remove the possiblity of not having a flat heatsink or waterblock.

My motherbloard(lol funny typo had to leave it in) sensor says that my idle temp for my cpu is 42C but i dont really trust it. It should be lower.

it could be that my 2600+ Thoroughbred gives off way too much heat, my brother has a 2600+ barton with a copper/aluminum heatsink and it idles at 46C
 
Sniper X said:
Direct die cooling is better because moving water can take heat away faster than if the heat traveled through thermal paste then a copper base then into the water. with direct die cooling you are getting rid of 2 of the thermal mediums that the cpu heat has to travel through. You also remove the possiblity of not having a flat heatsink or waterblock.
Click to expand...
Getting rid of 2 thermal mediums, sure.

But think about air cooling. By using a thermal coupling to a high surface area block of metal, you greatly increase the ability of the air to remove heat. Why would the same not be true of water?

Sniper X said:
My motherbloard(lol funny typo had to leave it in) sensor says that my idle temp for my cpu is 42C but i dont really trust it. It should be lower.
Click to expand...
So, er... sure that's a typo? I'd like to see this compared to a commercial solution with a waterblock. I also think that if direct water cooling was a better solution, someone would have marketed it by now.
 
kumquat said:
Why would the same not be true of water?
Click to expand...

Water has a higher density than air so cool water contacting a hot surface will cool down the hot surface faster than cool air. In fact the water will cool down the CPU better than having a waterblock over the CPU. Someone did this a while back and the temperatures were much much better.

kumquat said:
So, er... sure that's a typo? I'd like to see this compared to a commercial solution with a waterblock. I also think that if direct water cooling was a better solution, someone would have marketed it by now.
Click to expand...

It depends on what you mean by "better". Better as in more marketable, safer and easier to use for the average person? No, it isn't, which is why currently no blocks out there are direct die.

What direct watercooling does is offer you better performance at the cost of ease of use and easy installation. The average person doesn't want to seal their waterblock to their CPU, nore spray on sealant to protector the resistors. In fact, most people don't even want to install their own heatsink paste.

As for your temperatures, Socket A procesors don't have an on Die thermal probe so the temperature your motherboard is giving you is the temperature in the cavity under the CPU. It detects the temperature of the air there. Since your CPU still generates heat and you have no fan blowing over the area, the temperature reported by your motherboard will be higher than your CPU in this case.

Anyways, good job!
 
DaLurker said:
As for your temperatures, Socket A procesors don't have an on Die thermal probe so the temperature your motherboard is giving you is the temperature in the cavity under the CPU. It detects the temperature of the air there. Since your CPU still generates heat and you have no fan blowing over the area, the temperature reported by your motherboard will be higher than your CPU in this case.
Click to expand...

That makes sense because i was wondering what the thermal probe that stuck out of my cpu socket was for, I wanted to think it was for case temps. Im really curious now to know what my actual die temps are.
 
i have seen many posts from people that used direct die and the proc died, not immediately, but after a few months. I lapped the die of my proc and i say direct die WC is too risky ;)
 
kemist1117 said:
i have seen many posts from people that used direct die and the proc died, not immediately, but after a few months. I lapped the die of my proc and i say direct die WC is too risky ;)
Click to expand...

If it dies i have an excuse to upgrade. :D
 
It's not the prettiest thing to look at, but I bet it works well. I salute you for getting the tubing over all of those copper pipe parts.
 
what if instead of water he used non conductive fluid?? that would save the need to secure parts which could be damaged by water.
 
thewhiteguy said:
It's not the prettiest thing to look at, but I bet it works well.
Click to expand...

i think it looks cool in the blue light, kinda industrial, and the green antifreeze glows.

thewhiteguy said:
I salute you for getting the tubing over all of those copper pipe parts.
Click to expand...

yeah, getting the tubing on the pipes was probably the most stressful part of the project.
 
IffY said:
what if instead of water he used non conductive fluid?? that would save the need to secure parts which could be damaged by water.
Click to expand...

non conductive fluid doesn't have as good thermal properties as distilled water and its more expensive if you want the good stuff.
 
DaLurker said:
Water has a higher density than air so cool water contacting a hot surface will cool down the hot surface faster than cool air. In fact the water will cool down the CPU better than having a waterblock over the CPU. Someone did this a while back and the temperatures were much much better.
Click to expand...


Um, this is not at all how cooling works. Basically, the basic equation for a heat exchanger is that Q (Heat Removed) = U (overall heat transfer coefficient) * A(Contact Surface Area) * Delta T (Temperature difference). This why almost all heat exchangers have fins and such to increase the contact surface area. The reason that water cools better than air, and removing the IHS improves cooling comes in with the U value, the overall heat transfer coefficient.

For computer cooling, air or water, 1/U = 1/h + sum(b/k) (1/U is overall heat resistance) h is the fluid heat transfer coeffecient, b is thickness and k is thermal conductivity for each solid in the process. The reason that water cools much better than air is that the h value for water is much much higher. While some of this is due to the density, there are many more factors such as thermal conductivity, heat capacity, fluid velocity, fluid flowrate. Also, for most waterblock w/ decent thermal compounds, the b/k values little resistance comparatively (<10%). Removing the IHS improves temps by removing a very thick, fairly resistive solid layer, but still doesnt compare to the improvement you gain by going water from air.

What this all means is there is a reason why companies dont produce on-die coolers. They dont work as well and are more dangerous to your system.
 
Cool system!,something from the norm!
Im pretty sure water removes the same volume of heat from an equal area at 15 time's that of air ..maybe much higher.try tweaking your flow rate with a small valve too.
 
Not to be too argumentative, but I wouldn't use an almost 5 year old article as a source anymore for comparing water-blocks to direct-die, things can change a lot in 5 years.

Using the C/W values direct from that site, and comparing for example to the published Swiftech Storm values shows that the storm has C/W values ranging from 70% to 33% of that direct die, depedning on flow rate. Even the older, cheaper MCW6002 has values 70% to 50% of the direct die method. This is all not to say that direct-die cooling is necesarrily bad. I mean, it is significantly cheaper, and gives you a more unique system, not to mention the pride of making it all yourself, but a modern water block should cool better, and be somewhat safer.

(Just for reference, C/W values are combined U*A values that I mentioned in my previous post. Even if your U value is lower (higher resistance) you can combat that by increasing the surface area exposed to the water)

Edit(Just a note, the testing is done on a newer processor, so the die sizes will not be the same, but that in fact means the storm is even better than it seems. With smaller die sizes, direct die cooling becomes less and less effective.)
 
Direct die is better.... Because with not direct die it has to travel through the paste and the copper medium then into the water. Direct die, the heat is transfered directly to the water.
 
sabrewolf732 said:
Direct die is better.... Because with not direct die it has to travel through the paste and the copper medium
Click to expand...
By that logic, then directing air directly over the die instead of forcing the heat to travel through the paste and copper is better.

It's obviously not.

The goal is to get the heat away from the die. If you move the heat into the water, it's the same as moving the heat into a copper block.

The question is which is more efficient - the copper block with a much higher surface area than the die, or the water itself. It would seem to me that a well designed copper block properly themally coupled to the die would have a far higher heat removal capability than water alone, but that's just my gut instinct. This would require a controlled comparison between the most efficient modern copper blocks and direct die.

I'd like to see one.
 
kumquat said:
By that logic, then directing air directly over the die instead of forcing the heat to travel through the paste and copper is better.

It's obviously not.

The goal is to get the heat away from the die. If you move the heat into the water, it's the same as moving the heat into a copper block.

The question is which is more efficient - the copper block with a much higher surface area than the die, or the water itself. It would seem to me that a well designed copper block properly themally coupled to the die would have a far higher heat removal capability than water alone, but that's just my gut instinct. This would require a controlled comparison between the most efficient modern copper blocks and direct die.

I'd like to see one.
Click to expand...

Air is a horrible heat transfer medium. You can't move air over fast enough and in enough directed quantity, whereas w/ direct die water it's directed at the core and is being continuously pumped over. But you're right, a direct comparision would be better.
 
sabrewolf732 said:
Air is a horrible heat transfer medium.
Click to expand...
Right. Water is obviously better than air.

But is it better than copper with a highly efficient thermal compound?
 
Calth said:
With smaller die sizes, direct die cooling becomes less and less effective.
Click to expand...

i think what you ment to say was "With smaller die sizes, any type of cooling becomes less and less effective." if the same amount of heat is generated

What is the limiting factor in waterblocks today? What makes a successful waterblock?
 
kumquat said:
Right. Water is obviously better than air.

But is it better than copper with a highly efficient thermal compound?
Click to expand...

Yes.
 
Sniper X said:
i think what you ment to say was "With smaller die sizes, any type of cooling becomes less and less effective." if the same amount of heat is generated

What is the limiting factor in waterblocks today? What makes a successful waterblock?
Click to expand...

While waterblocks become less effective with die size decreasing, it is too a much lesser extent than direct die cooling. The limiting factors for water blocks are surface area to weight, surface area to cost, and surface area to pressure drop ratios. Ideally, you want as large as surface area possible with minimum weight, but this leads to more difficult machining and increased manufacturing costs, as well as increased pressure drops. Increasing the pressure drop is bad, as it effectively decreases the heat transfer coefficient of the water by decreasing the overall liquid flowrate.

So the challenge in waterblock design is balancing all of these factors.

As to whether or not water is more efficient than a well-applied thermal compound and water block, my opinion from what I've seen in my schoolwork (I've taken several classes involving heat transfer, one of which was primarily heat exchanger design), thermal conductivity is almost never the main source of thermal resistance, most times <10%, in a heat exchanger, especially considering the main material is copper which is very conductive. Now the thermal compound can cause difficulties if it is not applied right. If you have a very thick layer it can probably begin to have a significant effect, but that is the fault of the applier not the waterblock. With a proper thin coat, the thermal compound will have a very slight effect. (Heat flow is proportional to 1/thickness, that is the thinner the layer the higher the heat flow)
 
kumquat said:
OK.

Show me.
Click to expand...

it's common sense. The water will be the medium removing the heat either way, and it's flow rate is the same either way. Thermal paste is actually a lot less efficient than you believe it to be.
 
sabrewolf732 said:
it's common sense. The water will be the medium removing the heat either way, and it's flow rate is the same either way. Thermal paste is actually a lot less efficient than you believe it to be.
Click to expand...
How is it common sense?

The same could be said of air.

"The air will be the medium removing the heat either way, and it's flow rate is the same either way."

It's not obvious to me, as a person who studied 4 years to be an engineer, that the water will "obviously" be better.
 
kumquat said:
How is it common sense?

The same could be said of air.

"The air will be the medium removing the heat either way, and it's flow rate is the same either way."

It's not obvious to me, as a person who studied 4 years to be an engineer, that the water will "obviously" be better.
Click to expand...

Because you can't direct the air and the air doesn't conduct well enough to have much effect...
 
sabrewolf732 said:
Because you can't direct the air and the air doesn't conduct well enough to have much effect...
Click to expand...
How well does the water conduct? How well does the copper with a quality thermal paste conduct?

Give some numbers for your layman "common sense."
 
It's not obvious, because there are many factors that come into play in any heat transfer system.

Think back, probably when most of you were playing with Legos (so was I, don't get offended). CPUs WERE cooled with no heatsink, sometimes without any kind of forced convection. Yes, there are situations where direct fluid impingement can cool better than a heatsink/waterblock.

Sniper X: Nice job on this stuff. How close is the outlet to the CPU die? I probably would have put some turbulence-inducing features at the end of the nozzle, but who knows how much difference it would actually make.
 
^^ I still play with lego

According to Wikipedia (and my geography texbook)... Water has a thermal conductivity of 0.6 W/mk. Air, on the other hand has a thermal conductivity rating of 0.0262. By these figures, water is 23x better at removing heat than air. Waterblocks generally contain a fraction of the surface area as today's leading copper heatsinks. Yet, waterblocks consistantly achieve better temperatures than today's heatsinks. This would lead me to believe that regardless of size, air cooling can remove less heat than watercooling. If you soldered the top of the IHS to the base of your copper HS, then maybe things would be alot closer, but you can always solder your IHS to a waterblock too. Please, don't do that though...

Numbers? Sure.

Copper conducts at just over 400 W/mk, and Arctic silver contains a W/mk rating of 350, at a 25 micrometer layer. Considering the thinnest layer you're probably going to get between the CPU and the HSF is 1/2 a mm thick, and that's 500 micrometers... well, you can be the judge. Needless to say, arctic silver is not very efficient at moving heat. It's a hell of a lot better than other pastes, but nothing spectacular.

Do I think direct die cooling is a good idea? Well, no, because I've heard about water seepage into the die of the processor. And you thought current induced electrical migration was bad, just think about water induced electrical migration :eek:

To the OP. Your mod is awesome, don't get me wrong. If I had the experience needed to pull something like this off, I totally would too. And considering you chose an older processor, well, who cares about all the threadcrappers huh.
 
Arcygenical said:
^^ I still play with lego

According to Wikipedia (and my geography texbook)... Water has a thermal conductivity of 0.6 W/mk. Air, on the other hand has a thermal conductivity rating of 0.0262. By these figures, water is 23x better at removing heat than air. Waterblocks generally contain a fraction of the surface area as today's leading copper heatsinks. Yet, waterblocks consistantly achieve better temperatures than today's heatsinks. This would lead me to believe that regardless of size, air cooling can remove less heat than watercooling. If you soldered the top of the IHS to the base of your copper HS, then maybe things would be alot closer, but you can always solder your IHS to a waterblock too. Please, don't do that though...

Numbers? Sure.

Copper conducts at just over 400 W/mk, and Arctic silver contains a W/mk rating of 350, at a 25 micrometer layer. Considering the thinnest layer you're probably going to get between the CPU and the HSF is 1/2 a mm thick, and that's 500 micrometers... well, you can be the judge. Needless to say, arctic silver is not very efficient at moving heat. It's a hell of a lot better than other pastes, but nothing spectacular.

Do I think direct die cooling is a good idea? Well, no, because I've heard about water seepage into the die of the processor. And you thought current induced electrical migration was bad, just think about water induced electrical migration :eek:

To the OP. Your mod is awesome, don't get me wrong. If I had the experience needed to pull something like this off, I totally would too. And considering you chose an older processor, well, who cares about all the threadcrappers huh.
Click to expand...
1/2 a mm is way, way, way, way more than you're going to get between a decent heatsink and the die. Think an order of magnitude less than that.

A 25 micrometer layer is much closer to what you're going to get than half a millimeter.

Go grab a ruler and draw two little lines on a piece of paper with 1/2 a mm space between them and tell me there's that much space between the die and a HSF.
 
Arcygenical said:
^^ I still play with lego

According to Wikipedia (and my geography texbook)... Water has a thermal conductivity of 0.6 W/mk. Air, on the other hand has a thermal conductivity rating of 0.0262. By these figures, water is 23x better at removing heat than air. Waterblocks generally contain a fraction of the surface area as today's leading copper heatsinks. Yet, waterblocks consistantly achieve better temperatures than today's heatsinks. This would lead me to believe that regardless of size, air cooling can remove less heat than watercooling. If you soldered the top of the IHS to the base of your copper HS, then maybe things would be alot closer, but you can always solder your IHS to a waterblock too. Please, don't do that though...
Click to expand...

Unfortunately, those numbers on their own are meaningless in this discussion. Sure, water conducts heat better than air. But, you're comparing apples to oranges... As stated before, there are many other materials properties (density, heat capacity, viscosity...) as well as SYSTEM properties that dictate heat transfer efficiency.
 
You must log in or register to reply here.
Back
Top