Water block coming out 5-10C better than the Storm?

@Pete: I thought that too.. So in my minds eye I created a new IHS made it bigger to transfer heat to the water better. Sealed it to make leaks less likely. Then I took a look at my rig. Hehe....

Purely speculation:
Another way to improve the way an IHS would transfer heat to water would be to increase the speed water passes over the IHS. (This adds effective area without actually making the area bigger)
Or create turbulence. Or a mixture of both. I was also thinkin' creating alot of swirl over the IHS might have some merit but I eventually discarded the idea.
Most likely Anderssons drawing is pretty close except. Norths block is obviously at least 2 pieces. A seal plate and an accelleration plate.

I believe it's direct IHS rather than direct die. Simply because soldered IHS's are currently mainstream. And most user's probably are not going to try to remove it.

I looked it up and Copper has a thermal conductivity of 401 W/mk while water has a thermal conductivity of only around 0.6 W/mk. Can the water flow on direct die/IHS really be better than a properly designed water block? Does anyone know enough physics to calculate the thermal conductivity for a specific volume flow rate of water compared to the copper mass of a Storm for example?

Thicker more heat to travle throuhg and same for the cooling effect. Ie i take me a lot of effort to shout to someone in the usa but if you were sat in my front room no effort at all.

Thinner the better.

Removing the IHS easy peasy. What to do with it once off...umm. Maybe some dipples like the base of the storm and have the inlet directly over that, closer to the die that way. lapping the IHS on top till sort it too making it slightly thinner and flatter.

In all honestly you can't make the IHS that much bigger, thicker yeah, but you get the same problems again.

Just now need to find someone with a mill or cnc so i can get coper top block made up. Maybe sliver solder it as to braze for heat reasons on such a thin bit of copper!

The more powerfull the pump to a limit the better the cooling i'd say due to the flow of water over it

Quote:

---

Originally Posted by AzraelDarkangel

I looked it up and Copper has a thermal conductivity of 401 W/mk while water has a thermal conductivity of only around 0.6 W/mk. Can the water flow on direct die/IHS really be better than a properly designed water block?

---

Direct die cooling is way more efficient than a conventional water block could ever be.
Think about what exactly does "thermal conductivity" mean...
;)

Hint:
Water has 11-fold thermal capacity compared to copper.

Quote:

---

Originally Posted by largon

Direct die cooling is way more efficient than a conventional water block could ever be.
Think about what exactly does "thermal conductivity" mean...
;)

Hint:
Water has 11-fold thermal capacity compared to copper.

---

Well, that's why I was asking if anyone knew the physics/math behind it. I'm not saying direct-die isn't better. But the current logic most people are using of getting rid of the block to use water to cool directly is somewhat similar to saying "Hey let's get rid of the heatsink and cool the cpu with the fan directly." Thermal capacity is another issue, but since the block and water are used together it's not that important what the thermal capacity of each is, but what the thermal capacity of the whole system is. The amount of copper used in a water block is fairly small and by itself may have a very small thermal capacity, but the water is there to take it away in a hurry and the radiator is there to cool the water.

Also your statement "Water has 11-fold thermal capacity compared to copper." is a very blanket statement without specifying what volume of water compared to what mass of copper under what conditions. Or you could just link it to a reference.

Also "Direct die cooling is way more efficient than a conventional water block could ever be." is a very presumptious statement. You have no idea what new technologies could be developed. Or what the difference is exactly between the two methods, people have gotten varrying results.

Quote:

---

Originally Posted by AzraelDarkangel

But the current logic most people are using of getting rid of the block to use water to cool directly is somewhat similar to saying "Hey let's get rid of the heatsink and cool the cpu with the fan directly."

---

Erm... That's quite a crude analogy.

1kg of water has a volume of 1litre. 1kg of air takes 1293litres.
AND
The unit of thermal capacity:

kJ/(kg • K)

=

energy per mass... not volume.
Heatsink is a compulsory component due to the poor thermal attributes of air. Water is a whole another case though.


thermal capacity / resistivity / density
kJ/(kg • K) / W/(m • K) / kg/dm³

air
1.01 / 0.026 / 0.001293
Extremely poor capacity considering matter density.
Air is a thermal insulate. Not a coolant.

water
4.19 / 0.6 / 1.00
Superior capacity considering density. More than adequate conductivity.

copper
0.387 / 400 / 8.96
Low capacity, thus high correlation between the amount of thermal energy and temperature.
Superior conductivity compared to other materials; "good" for minimizing the impact of thermal resistance between heat source and coolant.
Note that direct cooling features no thermal resistance between the heat source and coolant.

The copper bottom of a water block only serves one purpose:

*to make the system closed, thus easier to use*

The second possible purpose would be to act as a heatspreader/exchanger in a scenario where the heat source has too high thermal density for the coolant flow to handle. This would also be the only thing that could render direct die cooling inferior to a conventional block because direct die cooling is very dependant on the amount of coolant flow in contact with the heat source.
BUT thermal density is not a problem, nor will it never become a problem for computer chips when water is used as a coolant, thanks to the die size and consumed wattage being almost constant through generations of chips.

Quote:

---

Thermal capacity is another issue, but since the block and water are used together it's not that important what the thermal capacity of each is, but what the thermal capacity of the whole system is.

---

Actually, the thermal capacity of the whole system is not important at all when we talk about the liquid cooling of a computer. Thermal capacity of a system only acts as a pool of heat. But ofcourse, we don't want to pile up the heat because we don't want the temperature of the system to gradually rise. The only two relevant factors are - less surprisingly - the ability the release thermal energy to the outside world and the coolant's thermal capacity.

Quote:

---

Also your statement "Water has 11-fold thermal capacity compared to copper." is a very blanket statement without specifying what volume of water compared to what mass of copper (...)

---

The whole point in using water as a coolant is that water needs more heat per unit than air for the temperature to rise a single degree. High thermal capacity (or more rapid replacement of the coolant) allows coolant temperature to change as little as possible - yielding lower temperature for the heat source.
Any interface (bottom of the water block) will only steepen the gradient, thus decreasing temperature delta.

Quote:

---

Also "Direct die cooling is way more efficient than a conventional water block could ever be." is a very presumptious statement. You have no idea what new technologies could be developed. Or what the difference is exactly between the two methods, people have gotten varrying results.

---

Note these 3 words:

Quote:

---

Originally Posted by myself

(...) conventional water block (...)

---

Meaning a metal vessel through which water flows. Be it a pin grid or jet impingement like, say, on Storm.

The fun thing is playing with thermal capacity and thermal conductivity as raising one and lowering the other can negate the effects of less ability for one or the other.

Very well written post btw, Largon ;)

Quote:

---

Originally Posted by andersson.j

I belive that a countersunk o-ring pressed against the IHS would seal very well. That's how many blocks (i.e storm) are designed and how many barbs seal. A simple direct IHS block could look something like this:
http://img252.imageshack.us/img252/69/directihs1nu2.jpg

A picture of the countersunk o-ring.
http://img252.imageshack.us/img252/4...ectihs3oy2.jpg

The only advantage with this method is that the heat has one less TIM to pass. But you could achive the same thing with a normal block on a naked core and then you could increase the surface area of the base with some structure like cups which makes it perform even better. Hence I don't think he's using direct IHS cooling.

Maybe if you used the same principle to build a direct die block? But it wouldn't be compatible with soldered IHSes so I doubt that's what he's thinking either.

I wonder what he's up to, if he's actually up to anything...

---

How well would that perform if the channels were milled into the top? I'm thinking turbulance (check), minimal pressure drop (check), broad compatibility (check). Imagine a reverse Apogee...

The biggest issue I can think of with this sort of implementation would be removal/ cleaning/ bleeding the system. That could be one heck of a mess.

Thats direct die water cooling, if you search for it you may find some older projects that use it. (I did it once to a 9700pro and a celeron)

Very well put. Nice post. Largon

Edit: On an AMD IHS you could mod it any number of ways. On the Intels you have fewer options as they are thinner to start with. But thinner is less thermal resistance. How much diffrence will an IHS mod make? Depends on too many things. A simple inlet outlet as depicted won't work too good.
(but the drawing was meant to propose a possible mounting system)

I made a direct die block about two years ago. It worked pretty good and had an impingement design. I wish I had some accurate equipment to test it though. I used a rubber seal like the CAD picture above, and never had a leak.

Going to have a block made to fit on my LGA7&5 IHS which has been poped off, qill AS5 the die. On the top of the IHS i'll probs have some small pits centraly over the dia and along the dia with the inlet being centered. The IHS will be sliver soldered to the block i think...brase too much heat and over kill really. Just tin the edges then palce togther and clamp it and super heat it should be fine.

Quote:

---

Originally Posted by Clay

Very well put. Nice post. Largon

Edit: On an AMD IHS you could mod it any number of ways. On the Intels you have fewer options as they are thinner to start with. But thinner is less thermal resistance. How much diffrence will an IHS mod make? Depends on too many things. A simple inlet outlet as depicted won't work too good.
(but the drawing was meant to propose a possible mounting system)

---

With an IHS, you end up with the same effect of storm vs. conventional block: Buffer zone.

Direct Die is where the fun is.

I found this article you guys might find interesting, scroll down to the 2nd part:
http://xbitlabs.com/news/coolers/dis...031075540.html

http://xbitlabs.com/images/news/2006...mpingement.jpg

btw, nice explanation largon

That looks very nice and the numbers they quoted are more than good.
This will be an interesting cooling technique, but it seems to be too hard to homemake something similar.
What does the jet stream do and why is it better than a big water stream flowing over the die?
Does the jet plate accelerate the water?

Quote:

---

Originally Posted by AzraelDarkangel

I found this article you guys might find interesting, scroll down to the 2nd part:
http://xbitlabs.com/news/coolers/dis...031075540.html

http://xbitlabs.com/images/news/2006...mpingement.jpg

btw, nice explanation largon

---

I like the first pic in that link. (Have been looking for it for a while now to post in here.)

Take a new block that would run water over the rods to cool it. Like in a nuclear reactor.

I'd like to wish the guy (NortHWizarD) whose threads are the original focus of this thread the best of luck.

I think it's great that his efforts seem to have sparked off some new interest in waterblock design, and that is a very good thing. Some of the responses in the latter half of this thread show some good thought and a few decent ideas.

I think that any discussion, experiments and developments are good. Regardless of how the block performs, the fact that people are talking about block design aspects is a good thing.

Quote:

---

Originally Posted by Fr3ak

That looks very nice and the numbers they quoted are more than good.
This will be an interesting cooling technique, but it seems to be too hard to homemake something similar.
What does the jet stream do and why is it better than a big water stream flowing over the die?
Does the jet plate accelerate the water?

---

Same principal as the storm: concentrated cooling due to velocity and turbulence of the water not only ripping through the jets but slamming into the base (in this case the core). Violence is good! :slobber: :D


Slots may work similarly or better, direct die hasnt been tested thoroughly really.. but pressure and velocity are going to help a lot. Thus a small gap between the die and top of the block is probably necessary to facilitate high velocity and decent pressure.

Quote:

---

Originally Posted by AzraelDarkangel

I found this article you guys might find interesting, scroll down to the 2nd part:
http://xbitlabs.com/news/coolers/dis...031075540.html

http://xbitlabs.com/images/news/2006...mpingement.jpg

btw, nice explanation largon

---

Would'nt the seal in this picture burn? or at least wear over time from the heat from the side of the core?

One of the main issues with direct die is that CPU silicon isn't wholly water-proof. Over an extended period of time water molecules will permeate the silicon substrate and render the CPU inoperable. Of course this can be addressed by providing some additional layer atop the silicon upon which the jet streams will act, but that then introduces additional thermal resistance and additional concerns with respect to the longevity of the insulative layer under the effects of a water jet stream. Given enough time, water will abrasively wear away pretty much anything, it all comes down to time vs expected CPU life-span vs added thermal resistance. Still, the guys at IBM are pretty smart and I'm sure that they're on top of it already.

A heat dissipation of 375W/cm² is a fairly decent and is right up there, assuming a 100C peak silicon temperature. I calculated that the Storm block on a bare-die CPU on a good mount is capable of around the 450W/cm² sort of range. Of course, cost and convenience are the final deciding factors.

I have a question,

With the shrinking of chips, is there going to be a point where chips may even not need after market cooling?


Like on the old INtel/AMD chips where you could just use a small HS or hell maybe even nothing as the heat output would not be that hi because chips would be efficient and able to pump out CPU cycles with out eathing mad juice?

Quote:

---

Originally Posted by Cathar

I'd like to wish the guy (NortHWizarD) whose threads are the original focus of this thread the best of luck.

I think it's great that his efforts seem to have sparked off some new interest in waterblock design, and that is a very good thing. Some of the responses in the latter half of this thread show some good thought and a few decent ideas.

I think that any discussion, experiments and developments are good. Regardless of how the block performs, the fact that people are talking about block design aspects is a good thing.

---

:fact: you're right. The whole watercooling discussion has kind of gotten a little stale. It would be great to see how the end product compared to your designs :toast:

Quote:

---

Originally Posted by phelan1777

I have a question,

With the shrinking of chips, is there going to be a point where chips may even not need after market cooling?


Like on the old INtel/AMD chips where you could just use a small HS or hell maybe even nothing as the heat output would not be that hi because chips would be efficient and able to pump out CPU cycles with out eathing mad juice?

---

The older silicon processes were less energy hungry because they suffered less from thermal leakage, and what thermal leakage they did have was confined to a relatively small number of transistors (1/100th the amount of today's CPU's).

As CPU manufacturing processes have shrunk and more transistors get packed into the same area, it's my understanding that the CPU manufacturers are doing all that they can to keep a lid on CPU thermal output as it is. This is the main problem, that being transistor density. Each transistor that's in use leaks some amount of energy as heat, and (speaking extremely simplistically) when you stick 100x a many transistors in the same area, what was a 1W problem before now becomes a 100W problem.

Again, hugely simplified, but overall it does highlight the nature of the problem that CPU manufacturers face. As it stands, it would appear that thermal issues appear to be the main reason why CPU's can't be made any faster. Sure, there are breakthroughs here and there, and some CPU's run fairly cool, but the moment you try to ramp the speed up to the point that the CPU process should be capable of supporting, heat skyrockets, so instead we have CPU's that run fairly cool only because they run somewhat slowly compared to what they can do. For example, Conroe's running at 2.4GHz stock, but given sufficient cooling will do 4-5GHz.

Cathar,

Just wondering what you think of an anderson like direct-IHS "block". With a storm like design, i.e. a centered jet impingement inlet, do you think it would be possible to create enough turbulence to create better performance than traditional blocks despite the flat IHS? Removing the TIM boundaries seems like a big win.

Quote:

---

Originally Posted by Cathar

The older silicon processes were less energy hungry because they suffered less from thermal leakage, and what thermal leakage they did have was confined to a relatively small number of transistors (1/100th the amount of today's CPU's).

As CPU manufacturing processes have shrunk and more transistors get packed into the same area, it's my understanding that the CPU manufacturers are doing all that they can to keep a lid on CPU thermal output as it is. This is the main problem, that being transistor density. Each transistor that's in use leaks some amount of energy as heat, and (speaking extremely simplistically) when you stick 100x a many transistors in the same area, what was a 1W problem before now becomes a 100W problem.

Again, hugely simplified, but overall it does highlight the nature of the problem that CPU manufacturers face. As it stands, it would appear that thermal issues appear to be the main reason why CPU's can't be made any faster. Sure, there are breakthroughs here and there, and some CPU's run fairly cool, but the moment you try to ramp the speed up to the point that the CPU process should be capable of supporting, heat skyrockets, so instead we have CPU's that run fairly cool only because they run somewhat slowly compared to what they can do. For example, Conroe's running at 2.4GHz stock, but given sufficient cooling will do 4-5GHz.

---

I have a question. Why do they have to shrink it and cram the transistors? I don't mind a cpu the size of a playing card or even a cd. The rest of the computer is still a lot bigger. Do they have to shrink it for performance or are they doing it just to make things smaller?

PS:
hey Cathar, do you have any more videos of you on your bike at the track? :D