The next evolution to cathar's microchannel design

[hara]

Although the title will sound like bs to some of you, I've spent lots of time thinking about this design. As the title suggests, it isn't built from the ground up by me. Before explaining anything, I'll get strait to some pics:

[hara]

The difference to cathar's design is that as the distance from the cpu core gets larger, the base gets thicker, aiding in better conductivity. Here, the are above the core is cooled by a microchannel which extends deep into the copper.

[hara]

Another shot. The base thickness is nearly proportional to the distance from the heat source.

[hara]

Shot from above. Also I'm desinging a top with jet impingement. It will only require one layer opposed to 2 on cathar's design. For those who are wondering, this was derived from this block

[chewyboy]

hara, good job on that. could you throw up a cad (dwg/dxf/other) file of that. I'm not 100% quite sure where some things are on your design. it even looks like some of it is offcentered. thanks.

[hara]

Here it is...

As I said, It has one inlet, at the centre and two outlets which converge to one. Here is what I mean. It's 1am. Tomorrow, I'll draw the top which is different than cathar's.

[Cathar]

Originally posted by hara
The difference to cathar's design is that as the distance from the cpu core gets larger, the base gets thicker, aiding in better conductivity.

In which case you've totally misunderstood my design and the reasons for why things are done the way they are...

Still, looks nice. On the surface it looks almost identical to morphling1's implementation with the center inlet coming out and around to a single outlet above the inlet. Problem is that approach imbalances the flows, but I guess not everyone wants a 3-barb block...

Looks good for a start. Keep it up.

[hara]

Still, looks nice. On the surface it looks almost identical to morphling1's implementation with the center inlet coming out and around to a single outlet above the inlet. Problem is that approach imbalances the flows, but I guess not everyone wants a 3-barb block...

That will be compensated by means of the special top

More on that later

[chewyboy]

thanks, i get it now. the machining down into the block is hat confused me. a question though, i believe cathar designed the block with an optimal base thickness as for maximum cooling. taking more of the metal out like you did wouldn't that reduce some of the heat transfer? if i'm wrong on any of that let me know.

[jaydee]

Originally posted by hara
The difference to cathar's design is that as the distance from the cpu core gets larger, the base gets thicker, aiding in better conductivity. Here, the are above the core is cooled by a microchannel which extends deep into the copper.

Base gets thicker? I am not to sure but it seems to me the thicker the material the harder it is to extract the heat, and the denser the material the worse it gets. Copper being pretty dense the thicker you make it the harder it will be to get heat out of the thicker areas. I do not know if there is any real scientific logic to back that up, but that is how I visulize it in my head. Maybe someone that does know can shead some light on that.

And you do not want it to thin either, because if there is not enough material to absorb the heat and transfer it to the coolant that is also bad. And you want enough material to keep the thing from breaking under the 20lbs or so amount of pressure that is recommended from AMD on that little core. Where these points are is trivial to the design as I am sure you know.

Good Luck, I hope you can find someone capable of milling that.

[SysCrusher]

Originally posted by hara
The difference to cathar's design is that as the distance from the cpu core gets larger, the base gets thicker, aiding in better conductivity. Here, the are above the core is cooled by a microchannel which extends deep into the copper.

This is how I built my block though it's made from Al, minus the channels and nozzle and add three barbs. If you were to look at mine face down you would see a two stepped hole and it tapers off to a fine point in the center just above the die with swirling turbulators running up the sides. The tapered idea allows you to get as thin of a base in the center of the heat area without comprimizing the surface strength of the metal touching the die. I managed to get within 1mm in the center before I quit. I thought of the idea after watching about fire fighting. Kill the source of heat which is dead center in the middle. If you do that then the block should be able to cool the surrounding area effeciantly. It works great for me with Al to aleast 7C delta - I like to try it in copper and see if I could knock that down by 2C more to tie or come close to Cathar's. Though it won't work well for a pelt.

I doubt you would be able to beat Cathar's design but you never know till you give it a whirl. I think it will work well but I think you will have problems with flow rates. I don't understand how flow works in your design.

[winewood]

Hara,

I think this design totally rocks. I like the 1 mm shallow divot in the center that is over the proc. It looks like it will be a great block! I can't think of a single thing that I would change. The crazy principles of thermodynamics really allow for some designs that are unique. It seems like this is like an aircooled fin system, cept the fins are small where they need to be to allow the water to quickly wisk away the heat. This design may be very useful at a high flowrate concept. GJ! Please post some benchmarks soon. I wont even compare to someone elses blocks, as numbers are the only thing that matter, and speculation is.. just that.
I am looking forward to that 2 barb outlet that is joined in the cap. VERY promising.

Maybe this would work really well with Alu since the base retains much of its thickness...

[hara]

Thanks for your comments! I'll get to the top now.

[hara]

Copper being pretty dense the thicker you make it the harder it will be to get heat out of the thicker areas

I've read that this is a myth. Temperature is inverseley proportional to the area and proportional to the distance.

[hara]

This is the block with the top which will compensate for the uneven waterflow, making it even again.

[SysCrusher]

Is the block off center? If it is I would cut out a step on the base to clear the socket.

[hara]

It is off centre. The Inlet nozzel is directly over the CPU core. There is no need to cut a step on the base. It has been taken into consideration already.

Will this block perform well?

[bigben2k]

IMO, I think you've got the channel depth upside down: it should be lower around the core, and higher over the middle. It would keep an equal distance between the bottoms of the channels, and the middle of the core.

That would "aid in better conductivity", since the heat disperses in a radial pattern, from that center point.

[hara]

IMO, I think you've got the channel depth upside down: it should be lower around the core

No I haven't. Think again. Why does the maze 3 have water entering from the centre and exiting from the sides not the opposite? The same holds here, only in the 3rd dimension. It may seem confusing at first but if you think about it, you'll understand what I mean.

What do you think about the top?

[SysCrusher]

Originally posted by hara
It is off centre. The Inlet nozzel is directly over the CPU core. There is no need to cut a step on the base. It has been taken into consideration already.

Will this block perform well?

The reason I ask is because alot of motherboard manufactures crowd the space just under the socket with big capasitors and such so it wouldn't fit. It won't fit on my board at all unless you reverse the block and add the step on the base. If it was me I would make the longer side of your block sit over the longer side of the socket. The again, I could be totally confused as to how your block sets over the socket.

I think it would perform well with what you got now. I'v been able to get within 2C of the WW block with my tapered base block but I never thought of microchannels like you have with a tapered base. Make it from copper and I believe you will have something to actually compete with the WW.

Despite how others think the depth of your channels should be reversed, I think you would want more depth in the center to whisk away the heat more effectively and faster. Then a higher depth on the ouside of the center so that the depth tapers off increasingly getting deeper at the middle. The heat is radial - yes that is true. A tapered depth across the channels I believe would make that radial even smaller because your effectively carrying the heat away from the center alot better and faster. If you can do that then the heat generated from outside the center would be less because the center would be cooler adding less heat to the edges thus you would need a smaller depth to cool the sides effectively. A flat base would just make that radial flat spreading the heat out over the edges. A tapered base would make that radial even smaller without spreading the heat out to the edges because the thicker sides of the taper would allow the heat to travel up and not to the sides. Think of it as a fire. To put out a fire you would toss water directly on the center of the fire. If you would to toss water around the edges of the fire you would never put out the fire because the heat from the center would just heat the edges back up and burn again. Just one big perpetual situation.

Also you want to make sure your water flow/velocity is able to reach the very bottom of the tapered channels for it to work correctly otherwise your stagnant water will just heat up and sit there.

[jaydee]

Originally posted by hara
I've read that this is a myth. Temperature is inverseley proportional to the area and proportional to the distance.

That made no sence to me. Seems you just backed up what I said though.
Thicker = more distance? So if you are not appying any more coolant to that thicker (as I cannot see that being possible in a CPU WB) area it has more distance for the heat to travel causing less efficent cooling of the thicker area? If what I said was a myth then it would be just as easy to cool a 2 inch square copper block as it would be an infinate amount of copper with the same amount of coolant capacity. :shrug:

Anyway you state

Originally posted by hara
The difference to cathar's design is that as the distance from the cpu core gets larger, the base gets thicker, aiding in better conductivity. Here, the are above the core is cooled by a microchannel which extends deep into the copper.

If there is less heat in those areas would'nt a thinner base be needed instead of a larger one? Why would you want to add more mass to the cooler areas to build up heat?

[hara]

It is funny that some things can be interpreted in different ways.

1. If the mobo fits a maze3, this wb will fit
2. You need to cool the heat generated by the cpu core. It is that what matters.
3. The "extra cooling" is different than cathar's because the heat passes through a much larger area (because the fin thickness is uneven). Heat passes better through a larger area. Temperature is inversely proportional to the cross sectional area.
4. Thicker = more distance? True but according to pythagoras theorem the distance is the root of Both the adjacent squared added to the opposite squared meaning that the distance will not be much larger but the area will be much larger.
5. Why would you want to add more mass to the cooler areas to build up heat? The heat will have already been absorbed by the water.

Any suggestions to the design?

[SysCrusher]

Originally posted by hara
It is funny that some things can be interpreted in different ways.

1. If the mobo fits a maze3, this wb will fit
2. You need to cool the heat generated by the cpu core. It is that what matters.
3. The "extra cooling" is different than cathar's because the heat passes through a much larger area (because the fin thickness is uneven). Heat passes better through a larger area. Temperature is inversely proportional to the cross sectional area.
4. Thicker = more distance? True but according to pythagoras theorem the distance is the root of Both the adjacent squared added to the opposite squared meaning that the distance will not be much larger but the area will be much larger.
5. Why would you want to add more mass to the cooler areas to build up heat? The heat will have already been absorbed by the water.

Any suggestions to the design?

Sorry guys, I took my time posting and by then you posted another.

Check my post above. I think I tryed to explain it. Either way I think it will work great for you.

[jaydee]

Originally posted by hara


5. Why would you want to add more mass to the cooler areas to build up heat? The heat will have already been absorbed by the water.

If it was already absorbed by the water then you wouldn't need any material there at all, let alone set different depths? :shrug: The more mass you have the slower the heat from the center of that mass will make it to the outer edges where the coolant is. The convection is the same but it is the same principle as driving. it takes you longer to drive 1 mile than it does 1.5 miles. The same applies to heat. Making a longer path for it to travel is not what you want at any temperature. But then again you don't want it to short either or there will not be enough mass to absorb the amount of heat being put off.

Reread what you said in #2 there. Worry less about what is going on farther away from the core and worry what is going on at the core. I think you are trying to make it far more complicated than it needs to be in order to do the same thing with varing depths around the core.

And as for varying the channel higths that sit ontop of the core, I have no clue how you can calculate the warmer spots of the core. The core itself is not evenly hot. The left side can be warmer than the right, or the middle warmer than the edges and hotspots here and there. It depends on what is going on inside the core. So trying to figure out what part of the core is the hottest and adjusting the design around that is nearly impossible. It would be better to just assume that the heat is spread 100% evenly across the core and make you diesign around that as it will not be possible to measure the temps on top of the core as it will fry with no HS.

So in other words what you said in #3 there is useless. Your proprotional argument fails as you do not know where the heat really is coming off the core. It is not going to spread evenly to your proprtional concept unless you are 100% certain where the heat levels are. If one fin is higher where the core is cooler it will not work well, nor will having the fin to short where it is warmer. Make them all the same higth. You also have to cosider the TIM joint will not be perfect enough to make use of your porprtional design (even if it was right).

If you knew where the most heat was, and you where capable of a perfect TIM joint (which there is no such thing) then making various levels to correspond with the heat maybe an idea. But this info is not known so IMO your wasting you time trying to do that. Make the fins the same higth, do not worry about the heat around the outside of the core as the water will do plenty good with just a flat bottom of the block to absorb it. No special stuff would be needed as there isn't enough heat left over to really worry about.

Thats all I have, I maybe off base. Best thing to do would make a few and trial and error it a bit. Try your different fin hights and try it with the same size fins ect... There is to many unknown variables to make it perfect by calculation IMO.

[hara]

I assumed the core is all heating the same. I placed the thin base exactly on the centre because heat it is surrounded by heat. It's in the middle of the heat source. The reality is that half the area of the cpu core is L2 cache which doesn't produce much heat but anyways... I'll start researching where to machine it cause I'm sure it's a bitch.

I thought about a compination of thorns, mills and drill presses. What do you think?