Direct Die Cooling Project. - Page 5

jaydee · 08-31-2003, 08:26 PM

Quote:

Originally posted by satanicoo
hummmnnn, maybe, just maybe, the TIM joint isnt 8 degrees after all...

My temps are higher than the Maze 4 so I am not sure we can really judge the TIM. In fact by that logic we would be going negative numbers, in other words the TIM adds efficency. Which is just not the case.

I think our previous theories where pretty correct that the die is just to small and to much wattage to cool without having a way to spread the heat first.

I don't know though.... Maybe someone will pop in with a new theory or idea.

MMZ_TimeLord · 08-31-2003, 09:33 PM

Excellent work Jaydee... I will have to go with the "this just proves my theory that Direct Die Cooling is WAY too risky and NOT benificial enough for me to persue." statement.

Those non-believers in that "other" forum can put that in their pipe and smoke it...

Zhentar · 08-31-2003, 10:51 PM

Quote:

Originally posted by jaydee116

I think our previous theories where pretty correct that the die is just to small and to much wattage to cool without having a way to spread the heat first.

But even if we halved the wattage, wouldn't the efficiency of both systems remain the same, still leaving the direct die worse?

I think its just flat out surface area. I've been trying to think a way to get more surface area on a direct die cooler and I'm not coming up with anything.

jaydee · 08-31-2003, 11:09 PM

Quote:

Originally posted by Zhentar
But even if we halved the wattage, wouldn't the efficiency of both systems remain the same, still leaving the direct die worse?

Sure but thats not the point. It is the size of the die. If you applied 75 watts to a 3/8" x 1/2" die it would be a lot harder to cool than a die 1"x1" with the same watt output. The heat is much more concentrated and much harder to cool because of less surface area for the coolant to do it's job on the smaller die.

Quote:

I think its just flat out surface area. I've been trying to think a way to get more surface area on a direct die cooler and I'm not coming up with anything.

Exactly. And there is NO way to change the surface area of the die unless your AMD! So I am not sure what your even thinking about there? Surface area is not controlable by us in direct die cooling. It is set by the size of the die on the CPU.

bigben2k · 09-01-2003, 12:09 AM

JD: maybe there's something awry in the implementation? What kind of flow rates are you getting? What's the flow geometry? (A quick MS paint job with measurements would help)

Scrizt · 09-01-2003, 06:32 AM

Just a guess out of no knowledge:
Direct die cooling could be better then a copperblock.
But the erode issue would then kill the CPU just to quick. (I'm talking really high pressure)

Volenti · 09-01-2003, 08:27 AM

Quote:

Originally posted by jaydee116
I am going ot use the T-Bird bretb sent me. This time I am going to remove the jets and try and make one big cone style jet right over the die and see what happens.

When I had my p4 direct die setup going I tried the single jet , it was far worse then the multiple jets, by 3-4 degrees.

I wouldn't be too concerned by the relitive lack of flowrate, mine only had ~150L/H. (down from a free flow of 700, 2x 700L/H pumps in series)

bretb · 09-01-2003, 10:16 AM

Quote:

Originally posted by bigben2k
JD: maybe there's something awry in the implementation? What kind of flow rates are you getting? What's the flow geometry? (A quick MS paint job with measurements would help)

From looking back , isn't he using 6 impingment jets directly onto the die then pulling that off the side, ala WW/Cascade.

Good to see the tests being done, good work JD.

jaydee · 09-01-2003, 10:38 AM

Quote:

Originally posted by bretb
From looking back , isn't he using 6 impingment jets directly onto the die then pulling that off the side, ala WW/Cascade.

Good to see the tests being done, good work JD.

Yes, pretty much sums it up. Jet ID is about .0625" OD is .125". Flow is 100GPH with the HydroThruster, 25GPH with the 170GPH Little Giant alone, and 40GPH with the 170 and the 200GPH together. Didn't see much change in temps either way.

bigben2k · 09-01-2003, 03:02 PM

Ok, but what's the height? How far apart are they?

Althornin · 09-01-2003, 03:04 PM

Quote:

Originally posted by satanicoo
hummmnnn, maybe, just maybe, the TIM joint isnt 8 degrees after all...

or maybe, the surface area of water/chip contact is so small that the delta T must be much higher....

jaydee · 09-01-2003, 03:32 PM

Quote:

Originally posted by bigben2k
Ok, but what's the height? How far apart are they?

There is 1/16" from the jets to the Die and the jets are 3/16" apart (center of hole to center of hole). I can bore the jet holes out I think without breaking them. Only way I can see to make this work better is to either add more jets (which I can't do) or to make the jets bigger to cover more of the die. The question is: Is it worth doing? I pulled the block off yesterday and it is ready for mods. I can give it another shot if people are really that interested...

Cathar · 09-01-2003, 07:35 PM

Jaydee, I found a reject middle plate if you're interested. It isn't fantastic, but it's the best of the bunch. PM me your address and I'll send it to you. The jets are ~3/32" apart from centre to centre, and so offer around 4x the jet density that you're presently using. You can fully utilise 24 jets in the space of a Barton sized core.

Dunno · 09-29-2003, 04:38 AM

I have experimented with DD cooling as well and agree that Athlons just dont have enough surface area for water to cool.

I cooled a P1 233@292 with DD made from a plastic garden chair foot with an inlet through the centre. Got better than WB results but could not get similar results with an Athlon.

How about making a tiny copper WB with a base the same area as the CPU top surface; Thin pins about 1cm long and a very thin base. Attach it with a non soluble TIM and perhaps a spring loaded centering mechanism to hold it down in place. I know you no longer really have DD cooling but you still get to DD cool the sides of the processor and increase the surface area without large delt T's from a thick base.

Just a thought.

funktional · 09-29-2003, 10:05 AM

My direct die Setup cools much better than any other watercooler i ever used.

My WW clone was my best ordinary Watercooler.

My first direct cooler gave me nearly the same temps but it performes a bit better at high power dissipation. It doesnt cool the whole die direkt only the center. It is applied just like an ordinary watercooler to prevent the water from coming into contact with the PCB. I used thermal paste as sealant and it worked quite well as long as the thermal paste wasnt too liquid.

The new Model is full direct die and uses multiple jets. it performs much better than the old modell.

Arcturius · 09-29-2003, 10:44 AM

That's a very interesting way of going direct-die, with the center directly cooled by water, and the edges by the block->water.

I am a bit leery, though, of the sealing method. It looks like you must be very precise in lining up the block to make sure the opening is centered on the die, along with high, even pressure to make sure the block doesn't tip, and make a leak.

Have you had any problems with leakage due to a slightly improper tightening of the block, or cracking the corners off a processor due to placing all of the mounting pressure strictly on the edges of the die?

EDIT: Also, what is 'fukk' cooling?

funktional · 09-29-2003, 11:50 AM

Yes, lining up the Block is not that easy but it is possible. I drilled the mounting holes quite big (7mm). than i applied spacers to the screws until the hole was right above the die.

the sealing didnt leak but i used heavy pressure, the springs were quite big and long and i doubt that i was within AMD specs

Even though the Corners didnt break.

Here is a pic i shot after i dismountet the cooler.

fakk should be full but i missed the 'l' key and hit the 'k'

funktional · 09-29-2003, 11:51 AM

Forgot the pic...

Pritorian · 10-05-2003, 04:35 PM

Quote:

Originally posted by jaydee116
There is 1/16" from the jets to the Die and the jets are 3/16" apart (center of hole to center of hole). I can bore the jet holes out I think without breaking them. Only way I can see to make this work better is to either add more jets (which I can't do) or to make the jets bigger to cover more of the die. The question is: Is it worth doing? I pulled the block off yesterday and it is ready for mods. I can give it another shot if people are really that interested...

I´m interested!

jaydee · 10-05-2003, 06:46 PM

Quote:

Originally posted by Pritorian
I´m interested!

It is in the works. Going to take a while before I get to it though...

HighFlowRod · 10-05-2003, 07:53 PM

jaydee, I have also been working on a Direct Die Waterblock. I've just been to lazy this past month. I just haven't drilled the middle hole, tapped the inlet and exit holes, and milled the o-ring groove. Take a look at it and see what you think. It's for a P4, btw.

jaydee · 10-06-2003, 10:37 AM

Quote:

Originally posted by HighFlowRod
jaydee, I have also been working on a Direct Die Waterblock. I've just been to lazy this past month. I just haven't drilled the middle hole, tapped the inlet and exit holes, and milled the o-ring groove. Take a look at it and see what you think. It's for a P4, btw.

Thats looks good. Not 100% sure I follow how the water exits but I get the general idea.

FL3JM · 10-06-2003, 11:15 AM

Direct Die Cooling will probably be alot more usefull with the new Athlon64 processors, they have heatspreader all over the top of the whole thing, like the P4 but even bigger.
The core is alot bigger too so that will prolly aid in direct die cooling.

---------
Pics of the nude Athlon64 for you strange people who havnt seen it allready:
http://www.tbreak.com/forums/showthr...5&pagenumber=1

jaydee · 10-06-2003, 12:16 PM

Quote:

Originally posted by FL3JM
Direct Die Cooling will probably be alot more usefull with the new Athlon64 processors, they have heatspreader all over the top of the whole thing, like the P4 but even bigger.
The core is alot bigger too so that will prolly aid in direct die cooling.

---------
Pics of the nude Athlon64 for you strange people who havnt seen it allready:
http://www.tbreak.com/forums/showthr...5&pagenumber=1

Heat spreader is rally irrelevant, if you want to direct die cool it has to go. BUT, direct IHS cooling has some possibilities. We discussed it before for the Intel. The AMD IHS looks bigger and I do not see any holes in it like the P4 one. It should be possible to make an O-ring fit around the outside of the IHS. Biggest problem here is pressure to make the O-ring seal. If I was to do it I would use RTV Silicone anyway which doesn't need as much pressure and it is an adhesive.... I will not be getting a A64/Opteron anytime soon though so that's about all I can add to that.

08-31-2003, 09:33 PM	#102
MMZ_TimeLord Cooling Savant Join Date: Apr 2003 Location: San Mateo, CA, USA, Earth Posts: 433	Excellent work Jaydee... I will have to go with the "this just proves my theory that Direct Die Cooling is WAY too risky and NOT benificial enough for me to persue." statement. Those non-believers in that "other" forum can put that in their pipe and smoke it... __________________ MMZ>TimeLord "Oooooooooh... that's gonna leave a mark!"

09-01-2003, 12:09 AM	#105
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	JD: maybe there's something awry in the implementation? What kind of flow rates are you getting? What's the flow geometry? (A quick MS paint job with measurements would help) __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

09-01-2003, 03:02 PM	#110
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Ok, but what's the height? How far apart are they? __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

09-01-2003, 07:35 PM	#113
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	Jaydee, I found a reject middle plate if you're interested. It isn't fantastic, but it's the best of the bunch. PM me your address and I'll send it to you. The jets are ~3/32" apart from centre to centre, and so offer around 4x the jet density that you're presently using. You can fully utilise 24 jets in the space of a Barton sized core. Last edited by Cathar; 09-01-2003 at 07:41 PM.

09-29-2003, 10:05 AM	#115
funktional Cooling Neophyte Join Date: Jan 2003 Location: Germany Posts: 17	My direct die Setup cools much better than any other watercooler i ever used. My WW clone was my best ordinary Watercooler. My first direct cooler gave me nearly the same temps but it performes a bit better at high power dissipation. It doesnt cool the whole die direkt only the center. It is applied just like an ordinary watercooler to prevent the water from coming into contact with the PCB. I used thermal paste as sealant and it worked quite well as long as the thermal paste wasnt too liquid. The new Model is full direct die and uses multiple jets. it performs much better than the old modell. Last edited by funktional; 09-29-2003 at 11:32 AM.

09-01-2003, 06:32 AM	#106
Scrizt Cooling Neophyte Join Date: Mar 2003 Location: Sweden Posts: 6	Just a guess out of no knowledge: Direct die cooling could be better then a copperblock. But the erode issue would then kill the CPU just to quick. (I'm talking really high pressure)

09-29-2003, 04:38 AM	#114
Dunno Cooling Neophyte Join Date: Sep 2003 Location: South Africa Posts: 82	I have experimented with DD cooling as well and agree that Athlons just dont have enough surface area for water to cool. I cooled a P1 233@292 with DD made from a plastic garden chair foot with an inlet through the centre. Got better than WB results but could not get similar results with an Athlon. How about making a tiny copper WB with a base the same area as the CPU top surface; Thin pins about 1cm long and a very thin base. Attach it with a non soluble TIM and perhaps a spring loaded centering mechanism to hold it down in place. I know you no longer really have DD cooling but you still get to DD cool the sides of the processor and increase the surface area without large delt T's from a thick base. Just a thought.

09-29-2003, 10:44 AM	#116
Arcturius Cooling Savant Join Date: Aug 2001 Location: Chippewa Falls, Wisconsin Posts: 294	That's a very interesting way of going direct-die, with the center directly cooled by water, and the edges by the block->water. I am a bit leery, though, of the sealing method. It looks like you must be very precise in lining up the block to make sure the opening is centered on the die, along with high, even pressure to make sure the block doesn't tip, and make a leak. Have you had any problems with leakage due to a slightly improper tightening of the block, or cracking the corners off a processor due to placing all of the mounting pressure strictly on the edges of the die? EDIT: Also, what is 'fukk' cooling?

09-29-2003, 11:50 AM	#117
funktional Cooling Neophyte Join Date: Jan 2003 Location: Germany Posts: 17	Yes, lining up the Block is not that easy but it is possible. I drilled the mounting holes quite big (7mm). than i applied spacers to the screws until the hole was right above the die. the sealing didnt leak but i used heavy pressure, the springs were quite big and long and i doubt that i was within AMD specs Even though the Corners didnt break. Here is a pic i shot after i dismountet the cooler. fakk should be full but i missed the 'l' key and hit the 'k'

10-06-2003, 11:15 AM	#123
FL3JM Cooling Neophyte Join Date: Jul 2003 Location: Sweden Posts: 66	Direct Die Cooling will probably be alot more usefull with the new Athlon64 processors, they have heatspreader all over the top of the whole thing, like the P4 but even bigger. The core is alot bigger too so that will prolly aid in direct die cooling. --------- Pics of the nude Athlon64 for you strange people who havnt seen it allready: http://www.tbreak.com/forums/showthr...5&pagenumber=1

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)