P4 direct die water cooling

Volenti · 10-12-2002, 09:21 AM

I've gone for a multiple jet approach over a single jet;

In the background you can see the 2 small 700L/H pumps that I'm using in series for more pressure.

here you can see the spray pattern;

this is a (slightly blurry) shot of the completed "block" showing the spray chamber;

the block next to the naked cpu;

block mounted over cpu with perspex hold down;

though it's only there for a bit of added strength, the block is epoxied to the cpu, making this kinda permanent, good thing it works well

preliminary temps are kind of interesting, during setup I leaked some water on the motherboard and after I dried it out I was getting some stability issues at idle so I've backed it right off so I can post this without getting "interupted". But the quick load tests I did immediatly after I got it going indicated a load temp 10 degrees above coolant.

once I sort out the stability issues (the motherboard may not be completly dry) I'll clock it back up and see what it thinks of it.

A quick and dirty flow rate measurment puts the flow at ~200L/H

this thing just screams out for an ultra bright blue led

Puzzdre · 10-12-2002, 11:16 AM

pretty brave move you made...

nice work also...

Quote:

this thing just screams out for an ultra bright blue led

ABSOLUTELY !!!

waiting for temps...

Simml · 10-12-2002, 11:28 AM

a blue led or an uv-led and uv-active coolant ... that would look great

if you have enough pressure, it should be possible to use a sealant ring instead of epoxid

bigben2k · 10-12-2002, 11:49 AM

Interesting, but I think you'll have to crank up the flow WAY up there. 200lph (50 gph) might work ok with a P4, but I'd shoot for at least 200 gph.

I dunno. You might want to do a graph of temp vs flow rate.

Nice work!

phreenet · 10-12-2002, 12:03 PM

Nice

, looks cool, got temps?

MadDogMe · 10-13-2002, 06:16 AM

Would'nt it of been easier to use holes rather than tubes?. bit like a showerhead?...

Can't wait to see results. How about a mini heatspreader plate(ultra thin between ridges, 0.5mm with 1.5mm+ high ridges) to go on the core?, with ridges or pins to spread the area?. I don't know how you'd fix it though. AS3 mixed with AS epoxy?.
Or individual Cu pins stuck with epoxy?, area & turbulance?...

I like the idea of direct die cause you're hitting the core sides with H20 as well as the top

...

Could you lay 'strips' of Cu across the core to increase surface area?, that way you would'nt need a 'plate' with a minimum thickness base...

PS. would the outlet be better recessed?, for flow patern?. rather than protruding into the 'chamber'...

Volenti · 10-15-2002, 07:17 AM

Ok I've got everything running sweet now, turns out I had a little bit of water trapped inside the CPU socket, gee I wonder why it was crashing

Got all that sorted out though.

I chose that nozzle design for a couple of reasons, it's pretty restrictive so it boosts water volecity a lot (even with these "pathetic" pond pumps) I wanted a jet that would attack the core with high volecity water, but that would also allow the "used" water to get the hell out of the way once it had done it's job of scrubbing a bit of heat from the core. It's also stupidly easy to make and mount.

As for performance you should probably take any temp readings with a grain of salt, since we all know how carefully calibrated the in-die diode is

but that being said this direct die approach kicks the crap out of all my blocks bar one, and is more stable at this speed than it was with the my best copper block (which was a clone of cathar's concept block)

P4 1.8@2.64, 1.8v , 10 degrees above water idle, 10.5 (as best I can judge) under cpu burn.

other points of interest, allthough I havn't yet tried one of my larger head pumps with this block(mainly because when I tested it with the jet it complained loudly) I did kill one pump ( I have 2 in series) and the temp only went up 1.5 degrees under load (that's roughly 100L/H)

The pumps I'm using are pretty cheap (about $15 US I'd imagine) so I may get another 2 and have 4 in series, that should proove interesting...

MadDogMe · 10-16-2002, 04:09 AM

I see what you mean about letting 'the water get out of the way' now, the space behind the outlets lets water 'rise'. Is the outlet flush with the block 'top inside'?, I can't make it out from the pics fuzziness

...

How far above water ambient is the CPU temp?, do you think increasing the surface area of the core will help lower temps?. what do you think about the feasability of AS epoxying 'fins' straight onto the core?, or making a micro thin(thiner than a block would support) plate with fins on to stick to the core if the first is'nt 'do~able'?

...

Volenti · 10-16-2002, 06:14 AM

Quote:

Originally posted by MadDogMe
I see what you mean about letting 'the water get out of the way' now, the space behind the outlets lets water 'rise'. Is the outlet flush with the block 'top inside'?, I can't make it out from the pics fuzziness ...

yea it's flush as well as having fluted edges.

Quote:

How far above water ambient is the CPU temp?, do you think increasing the surface area of the core will help lower temps?. what do you think about the feasability of AS epoxying 'fins' straight onto the core?, or making a micro thin(thiner than a block would support) plate with fins on to stick to the core if the first is'nt 'do~able'? ...

according to the in-die diode it's ~10 degrees (c) over the water temp, but I have no way of knowing how accurate that is.

as for trying to increase the surface area of the core like you said, well i dunno how effective it would be, since the whole point of going direct die in the first place is to avoid having TIM layers between the water and core. You could possibly use a dremel with a fine engraving bit to score the surface of the core, if you're feeling particulary brave

MadDogMe · 10-16-2002, 08:00 AM

I did acyualy think about that too

, but I don't know how a core is constructed, or what from?, if the silvery~ceramicy bit is just put on top as a heatconductor then scoring it would be OK, if there's curcuitry of whatever embedded in it then it's a no~no

...

There's also the problem of ideal pressure with the plate idea i 'spose?...

Volenti · 10-18-2002, 03:38 AM

minor update, Blue LED's! woot

Volenti · 12-03-2002, 03:40 AM

I have successfully removed the perspex block from the cpu (it was epoxied on) On close inspection all that could be seen on the core was very faint...marks... that are in a pattern that would match the contact points of the jets in the block, however as I rubbed the core some more with a peice of clean cloth to see them more clearly they slowly faded away, interesting.

I have also found a suitable, re-useable and safely non-permenent sealant to use in future direct die cooling, with that in hand I replaced the multiple jets in the block with a single 3/8'' intake(as an expriment)

As expected the flow rate shot up (I havn't measured it, though at least triple the flow of the multiple jets) and also (mostly expected) the block performed worse;

mulit jet, (temps in degrees above water temp)
idle 10
load 10.5

single 3/8''
idle 12
load 15

interesting to see the jet impingment theory in practice...

bigben2k · 12-03-2002, 08:37 AM

That's most interesting! Thanks, I can use that info!

What distance did you have between the 3/8 opening and the top of the CPU? Some of the info I've seen points towards a ratio of 10:1 (distance to opening size).

Albigger · 12-03-2002, 03:30 PM

Thanks for taking the time to test that out Volenti. man you must really spend A LOT of your time on all this.

Was this test conducted with your new 3 stage pump or with a single pump? Are you going back to the jet design or a newer jet design?

Quote:

Originally posted by bigben2k
That's most interesting! Thanks, I can use that info!

What distance did you have between the 3/8 opening and the top of the CPU? Some of the info I've seen points towards a ratio of 10:1 (distance to opening size).

Interesting. I'm curious as to why such a large ratio? i would have thought (to a point) closer to the object is better (less chance that the surrounding water will slow down the velocity of the incoming water)?

Links to some of this??

bigben2k · 12-03-2002, 03:41 PM

I've got it, somewhere... I'll post it later. It was related to a Korean nex-gen nuclear reactor emergency system.

Volenti · 12-04-2002, 01:46 AM

Quote:

Originally posted by bigben2k
That's most interesting! Thanks, I can use that info!

What distance did you have between the 3/8 opening and the top of the CPU? Some of the info I've seen points towards a ratio of 10:1 (distance to opening size).

it was about 6mm above the core, the multiple jets were about half that away.

Quote:

Thanks for taking the time to test that out Volenti. man you must really spend A LOT of your time on all this.

Was this test conducted with your new 3 stage pump or with a single pump? Are you going back to the jet design or a newer jet design?

The 3 stage pump is currently in bits while I work on a better valve setup, for the test I used a (surprisingly un-modified) 740GPH/8.5' head pond pump. ( have used that pump before with the mulitple jets as well)

I'll make a new block that will have provision for (hopefully) easy jet changes as well as stepping up to 1/2'' inlets/outlets (to make my plumbing a bit easier more than anything)

MadDogMe · 12-04-2002, 04:06 AM

Is that ratio for submerged usage BB2K?...

What is the mystery sealant Volenti?, is it silicone based?, I'd've thought it would be OK (silicone) cause you have a clamping pressure as well, I'd use epoxy if there was'nt...

Volenti · 12-04-2002, 04:59 AM

Quote:

Originally posted by MadDogMe
Is that ratio for submerged usage BB2K?...

What is the mystery sealant Volenti?, is it silicone based?, I'd've thought it would be OK (silicone) cause you have a clamping pressure as well, I'd use epoxy if there was'nt...

It's a re-usable temporary adhesive called Blu-tack (made by Bostic) use it in a thin layer like a gasket, you will still need mechanical clamping (think of it as a better O-ring for our purposes)

it's readily available in australia, I don't know of it's availability overseas.

I've been using it (literally) for years to seal subwoofers in enclosures, can't believe I didn't think of using it earlier.

hara · 12-04-2002, 06:08 AM

Blu-tack??? U-tack is better.

I use it to stick posters on my wall. It's a bit like plasticine or bubble-gum. I wouldn't trust it if i were you. Btw, as for availability, I live near tunez (tunisia)

Volenti · 12-04-2002, 06:18 AM

Quote:

Originally posted by hara
Blu-tack??? U-tack is better.

I use it to stick posters on my wall. It's a bit like plasticine or bubble-gum. I wouldn't trust it if i were you. Btw, as for availability, I live near tunez (tunisia)

this is exactly the reason why I didn't mention that I was using blu-tack in the first place.

I may be crazy, but I'm not stupid, it does work, other wise I wouldn't be using it.

Puzzdre · 12-04-2002, 06:41 AM

Blue tack is great! I used it to stick my loudspeaker cabinets onto the sand filled stands (30 kg each), I could still tip over the whole assembly just by pushing it on the highest side of the speaker cabinet. It sticks and holds like hell

...

Great thinking Volenti!!!

Les · 12-04-2002, 09:03 AM

Some possibly "best case" sums may interest.
Using 2mm OD and photo of nozzle estimate Nozzle 1mm ID.
Offset (H) of 3mm
Taking 200 LPH ( 28.5714 LPH per Jet) which gives a Velocity of ~ 10.1m/s
Calculated for a Circular Die using Flomerics Jet Impingement calculator.

The Flomerics calculator does show an increase in Heat Transfer Coeff with decreasing offset(within the calculator limits of 2D <= H <= 12D) .
This is an Offshoot of other play and am not(ATM) pursuing.

dax · 12-04-2002, 09:49 AM

I wonder if i could get something like Blue-Tack around here, would come in handy! Ought to check that someday, hmm so much work, so little time... hard to sort out priorities too: exams vs. computers.

And les i absolutely LOVE your charts... very nice work

Volenti · 12-06-2002, 02:32 AM

I've altered the single jet slightly, reduced it's diameter from 7.5mm to 5mm and brought it closer to the core (3mm away)

Temps did improve, by 2 degrees (load), but it's still lagging behind the multiple jet version, also I seem to have killed my secondary ram stick (won't boot with it in) will look into that later.

I'll make a new block over the weekend with a grid pattern of jets over the core, and more even and flow friendly outlets, see how that goes.

oops forgot pics;

MadDogMe · 12-06-2002, 03:40 AM

Is it watertight?(with the pressure?), cause I tried sealing something with bluuu tack and it leaked (no pressure though!!).
I'm sure there are better (specialised) sealants, but if it works!, I would'nt have trusted it though, (though I was gonna use it to unlock my CPU

)...

How long have you had it running/tested?...

10-12-2002, 09:21 AM	#1
Volenti Cooling Savant Join Date: Jan 2002 Location: in a nice cool spot Posts: 427	P4 direct die water cooling I've gone for a multiple jet approach over a single jet; In the background you can see the 2 small 700L/H pumps that I'm using in series for more pressure. here you can see the spray pattern; this is a (slightly blurry) shot of the completed "block" showing the spray chamber; the block next to the naked cpu; block mounted over cpu with perspex hold down; though it's only there for a bit of added strength, the block is epoxied to the cpu, making this kinda permanent, good thing it works well preliminary temps are kind of interesting, during setup I leaked some water on the motherboard and after I dried it out I was getting some stability issues at idle so I've backed it right off so I can post this without getting "interupted". But the quick load tests I did immediatly after I got it going indicated a load temp 10 degrees above coolant. once I sort out the stability issues (the motherboard may not be completly dry) I'll clock it back up and see what it thinks of it. A quick and dirty flow rate measurment puts the flow at ~200L/H this thing just screams out for an ultra bright blue led __________________ feel free to icq/msn me, I'm always willing to toss around ideas.

10-12-2002, 11:49 AM	#4
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Interesting, but I think you'll have to crank up the flow WAY up there. 200lph (50 gph) might work ok with a P4, but I'd shoot for at least 200 gph. I dunno. You might want to do a graph of temp vs flow rate. Nice work! __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

10-12-2002, 12:03 PM	#5
phreenet Cooling Savant Join Date: Feb 2001 Location: Gloucester, Virginia Posts: 356	Nice , looks cool, got temps? __________________ Dual Pentium!!! 933@1107 Liquid Cooled.

10-13-2002, 06:16 AM	#6
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	Would'nt it of been easier to use holes rather than tubes?. bit like a showerhead?... Can't wait to see results. How about a mini heatspreader plate(ultra thin between ridges, 0.5mm with 1.5mm+ high ridges) to go on the core?, with ridges or pins to spread the area?. I don't know how you'd fix it though. AS3 mixed with AS epoxy?. Or individual Cu pins stuck with epoxy?, area & turbulance?... I like the idea of direct die cause you're hitting the core sides with H20 as well as the top ... Could you lay 'strips' of Cu across the core to increase surface area?, that way you would'nt need a 'plate' with a minimum thickness base... PS. would the outlet be better recessed?, for flow patern?. rather than protruding into the 'chamber'... Last edited by MadDogMe; 10-13-2002 at 06:23 AM.

10-15-2002, 07:17 AM	#7
Volenti Cooling Savant Join Date: Jan 2002 Location: in a nice cool spot Posts: 427	Ok I've got everything running sweet now, turns out I had a little bit of water trapped inside the CPU socket, gee I wonder why it was crashing Got all that sorted out though. I chose that nozzle design for a couple of reasons, it's pretty restrictive so it boosts water volecity a lot (even with these "pathetic" pond pumps) I wanted a jet that would attack the core with high volecity water, but that would also allow the "used" water to get the hell out of the way once it had done it's job of scrubbing a bit of heat from the core. It's also stupidly easy to make and mount. As for performance you should probably take any temp readings with a grain of salt, since we all know how carefully calibrated the in-die diode is but that being said this direct die approach kicks the crap out of all my blocks bar one, and is more stable at this speed than it was with the my best copper block (which was a clone of cathar's concept block) P4 1.8@2.64, 1.8v , 10 degrees above water idle, 10.5 (as best I can judge) under cpu burn. other points of interest, allthough I havn't yet tried one of my larger head pumps with this block(mainly because when I tested it with the jet it complained loudly) I did kill one pump ( I have 2 in series) and the temp only went up 1.5 degrees under load (that's roughly 100L/H) The pumps I'm using are pretty cheap (about $15 US I'd imagine) so I may get another 2 and have 4 in series, that should proove interesting... __________________ feel free to icq/msn me, I'm always willing to toss around ideas.

10-12-2002, 11:28 AM	#3
Simml Cooling Neophyte Join Date: Oct 2002 Location: Austria Posts: 1	a blue led or an uv-led and uv-active coolant ... that would look great if you have enough pressure, it should be possible to use a sealant ring instead of epoxid

10-16-2002, 04:09 AM	#8
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	I see what you mean about letting 'the water get out of the way' now, the space behind the outlets lets water 'rise'. Is the outlet flush with the block 'top inside'?, I can't make it out from the pics fuzziness ... How far above water ambient is the CPU temp?, do you think increasing the surface area of the core will help lower temps?. what do you think about the feasability of AS epoxying 'fins' straight onto the core?, or making a micro thin(thiner than a block would support) plate with fins on to stick to the core if the first is'nt 'do~able'? ...

10-16-2002, 08:00 AM	#10
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	I did acyualy think about that too , but I don't know how a core is constructed, or what from?, if the silvery~ceramicy bit is just put on top as a heatconductor then scoring it would be OK, if there's curcuitry of whatever embedded in it then it's a no~no ... There's also the problem of ideal pressure with the plate idea i 'spose?...

10-18-2002, 03:38 AM	#11
Volenti Cooling Savant Join Date: Jan 2002 Location: in a nice cool spot Posts: 427	minor update, Blue LED's! woot __________________ feel free to icq/msn me, I'm always willing to toss around ideas.

12-03-2002, 03:40 AM	#12
Volenti Cooling Savant Join Date: Jan 2002 Location: in a nice cool spot Posts: 427	direct die discection I have successfully removed the perspex block from the cpu (it was epoxied on) On close inspection all that could be seen on the core was very faint...marks... that are in a pattern that would match the contact points of the jets in the block, however as I rubbed the core some more with a peice of clean cloth to see them more clearly they slowly faded away, interesting. I have also found a suitable, re-useable and safely non-permenent sealant to use in future direct die cooling, with that in hand I replaced the multiple jets in the block with a single 3/8'' intake(as an expriment) As expected the flow rate shot up (I havn't measured it, though at least triple the flow of the multiple jets) and also (mostly expected) the block performed worse; mulit jet, (temps in degrees above water temp) idle 10 load 10.5 single 3/8'' idle 12 load 15 interesting to see the jet impingment theory in practice... __________________ feel free to icq/msn me, I'm always willing to toss around ideas.

12-03-2002, 08:37 AM	#13
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	That's most interesting! Thanks, I can use that info! What distance did you have between the 3/8 opening and the top of the CPU? Some of the info I've seen points towards a ratio of 10:1 (distance to opening size). __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

12-03-2002, 03:41 PM	#15
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	I've got it, somewhere... I'll post it later. It was related to a Korean nex-gen nuclear reactor emergency system. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

12-04-2002, 04:06 AM	#17
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	Is that ratio for submerged usage BB2K?... What is the mystery sealant Volenti?, is it silicone based?, I'd've thought it would be OK (silicone) cause you have a clamping pressure as well, I'd use epoxy if there was'nt...

12-04-2002, 06:08 AM	#19
hara Cooling Savant Join Date: Oct 2002 Location: Malta, Mediterranean Posts: 662	Blu-tack??? U-tack is better. I use it to stick posters on my wall. It's a bit like plasticine or bubble-gum. I wouldn't trust it if i were you. Btw, as for availability, I live near tunez (tunisia)

12-04-2002, 06:41 AM	#21
Puzzdre Cooling Savant Join Date: Aug 2002 Location: Croatia Posts: 969	Blue tack is great! I used it to stick my loudspeaker cabinets onto the sand filled stands (30 kg each), I could still tip over the whole assembly just by pushing it on the highest side of the speaker cabinet. It sticks and holds like hell ... Great thinking Volenti!!! __________________ 'Out of cheese error... ...please reboot the universe (press the GBL to continue)'

12-04-2002, 09:03 AM	#22
Les Cooling Savant Join Date: Oct 2001 Location: Wigan UK Posts: 929	Some possibly "best case" sums may interest. Using 2mm OD and photo of nozzle estimate Nozzle 1mm ID. Offset (H) of 3mm Taking 200 LPH ( 28.5714 LPH per Jet) which gives a Velocity of ~ 10.1m/s Calculated for a Circular Die using Flomerics Jet Impingement calculator. The Flomerics calculator does show an increase in Heat Transfer Coeff with decreasing offset(within the calculator limits of 2D <= H <= 12D) . This is an Offshoot of other play and am not(ATM) pursuing.

12-04-2002, 09:49 AM	#23
dax Cooling Neophyte Join Date: Nov 2001 Location: Ghent, Belgium Posts: 83	I wonder if i could get something like Blue-Tack around here, would come in handy! Ought to check that someday, hmm so much work, so little time... hard to sort out priorities too: exams vs. computers. And les i absolutely LOVE your charts... very nice work

12-06-2002, 02:32 AM	#24
Volenti Cooling Savant Join Date: Jan 2002 Location: in a nice cool spot Posts: 427	I've altered the single jet slightly, reduced it's diameter from 7.5mm to 5mm and brought it closer to the core (3mm away) Temps did improve, by 2 degrees (load), but it's still lagging behind the multiple jet version, also I seem to have killed my secondary ram stick (won't boot with it in) will look into that later. I'll make a new block over the weekend with a grid pattern of jets over the core, and more even and flow friendly outlets, see how that goes. oops forgot pics; __________________ feel free to icq/msn me, I'm always willing to toss around ideas.

12-06-2002, 03:40 AM	#25
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	Is it watertight?(with the pressure?), cause I tried sealing something with bluuu tack and it leaked (no pressure though!!). I'm sure there are better (specialised) sealants, but if it works!, I would'nt have trusted it though, (though I was gonna use it to unlock my CPU )... How long have you had it running/tested?...

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