Heatsink Conversion -> Pelt Waterblock

koslov · 04-25-2003, 12:48 AM

After reading about some people's success with converting pin- or fin-type heatsinks into waterblocks, I decided to give it a try myself. I chose the Dynatron DC1U-B04, for it's high density fins and large surface area. It also has a flat bottom, with no groove for the socket cam. Well, just how big is it?

Full specs:
Material: Copper C1020
Dimension: 100x74x29
Pitch: 1.5 mm
Thickness of fin: 0.45 mm
Number of fins: 49
Height of fin: 9 mm
Height of base: 3.5 mm
Length of base: 100 mm

The plan:
This Socket 940 (Hammer) heatsink will be transformed into a waterblock for a 40 x 40mm peltier, to go on a socket A. As you can see from the last pic, the fin ratio is imperfect for watercooling, but I am still expecting good performance. I plan to chop off most of the fin area, leaving only a 50 x 50 area left in the center. Much of the baseplate will have to be chopped off too, just to fit on a Socket A mobo. The fins will also be cut down to a height of 4-5mm. I will top the block with 1/2" Lexan, which will have two barbs. Each entry will have a flow separator channel carved into the bottom of the Lexan, to try to equalize flow among the fins.

Construction
First, I would like to cut the baseplate to size (probably 80 x 74mm) with a table saw, leaving small pieces with fins still attached. These I'll use for VGA RAM cooling

.

Then I want to trim back the fins to their 50 x 50mm area, probably using my dremel.

Next I'll sand the fins down to a height of 4-5mm using a belt sander.

The Lexan top will be drilled out to make a cap for the block, with the cavity covering the fins. The flow separators will also be made using a drill press. I'll tap the Lexan about halfway in for the Poly barbs, leaving room for the flow separator.

The mounting to the coldplate will be accomplished with 4 screws going through the baseplate.

PS: If you want to order one, you can email or call Dynatron directly, and they will send you your heatsink for a small shipping charge and sales tax if you live in CA. This one was $29, plus $6 UPS ground and sales tax.

jaydee · 04-25-2003, 12:43 PM

Quote:

Originally posted by koslov

PS: If you want to order one, you can email or call Dynatron directly, and they will send you your heatsink for a small shipping charge and sales tax if you live in CA. This one was $29, plus $6 UPS ground and sales tax.

Hummm. $35 + barbs + materials for the top. I think one would be better off just buying a premade block.

bigben2k · 04-25-2003, 12:45 PM

You can also pick up any of the other Dynatron HSF: some of them are skivved, but otherwise have a good ratio of fin-to-channel, for watercooling.

They're relatively cheap too.

jaydee · 04-25-2003, 12:51 PM

Quote:

Originally posted by bigben2k
You can also pick up any of the other Dynatron HSF: some of them are skivved, but otherwise have a good ratio of fin-to-channel, for watercooling.

They're relatively cheap too.

That is sure a nice peice of metal he posted though. I would love to see how they make those fins. It looks like they slit and then bent the fins into place and then mill/grind/whatever the tops flat. Sure would love to see one made...

koslov · 04-25-2003, 04:17 PM

I probably would have gotten a commercial block, but few of them really seem fit for pelt use. Only the Swiftech's (which aren't cheap either) and perhaps #Rotor's design. But I think I can do better than both with this block... I am hoping for a °C/W of .15 or better with it, which is not unreasonable considering the bigger surface area of a pelt.

I would have picked a heatsink with a better fin ratio, but none would have given me as much room to work with as this one. This is how I came up with size requirements: 40 x 40mm area to be cooled, add 5mm to each side to allow for heat spreading in the baseplate, then add 10mm to each side for the Lexan top to attach to. This leaves 70 x 70mm. I guess I could have gotten away with a lot less if I didn't want a Lexan top. I could have had something closer to 50 x 50mm if I had just decided to solder a copper box top on. But I like working with Lexan, and this way I can remove the top for cleaning.

Anyway, what I was really hoping for was some construction tips. Should I use a table saw to cut the base? Do I need a special blade?

jaydee · 04-25-2003, 04:30 PM

Table saw to cut the base? Wouldn't recommend it. If you do you will need a $50+ saw blade capable of cutting metal. I am sure your familiar with www.overclockers.com. Hoot was making such a block and was using a table saw. PM and ask him which blade he is using. I bet it is $50 or better.

I havn't figured out how you plan on spreading the watter to all those channels effeciently. Your going to need a hell of a flow rate to cool a pelt off so you don't want a to restrictive block unless you are going to use a 500+GPH pump or something.

The best block I can think of for pelts is DTek's big Spiral that Fixittt designed. Deep channels and lots of flow over a lot of surface area. Just my opinion but I don't think micro channels is the best way to go for pelt cooling.

bigben2k · 04-25-2003, 04:37 PM

I disagree, somewhat.

The issue here is that you may have some difficulty getting the right clamping pressure on the pelt. I'm not sure if the fins can actually support the required clamping pressure without collapsing.

The Spir@l is ideal, because you can simply tap into the four corners, which are relatively massive.

Otherwise, if you can distribute the flow evenly, as much as possible, this block would do well with a pelt, but you're looking at a lot of work to get this done.

JFettig made a channel block that looks pretty good for pelt use. Cathar is also working on one.

jaydee · 04-25-2003, 05:03 PM

Quote:

Originally posted by bigben2k
I disagree, somewhat.

The issue here is that you may have some difficulty getting the right clamping pressure on the pelt. I'm not sure if the fins can actually support the required clamping pressure without collapsing.

The Spir@l is ideal, because you can simply tap into the four corners, which are relatively massive.

Otherwise, if you can distribute the flow evenly, as much as possible, this block would do well with a pelt, but you're looking at a lot of work to get this done.

JFettig made a channel block that looks pretty good for pelt use. Cathar is also working on one.

You disagree? You just backed up everything I was getting at.

How do you evenly distribute the flow through all those channels? If you put a inlet on the left and an outlet on the right the water will avoid the outer channels and go strait through the middle. Center inlet? Don;t see that working either. You would need a 40mm inlet. What cathar did to make the channels in the WW work will not work over 40mm with a 1/2" inlet. Probably not even a 3/4"inlet. And if you do that you will need one hell of a pump as 156+watts is alot of heat and your going to need a lot of flow through all of those small ass channels.

I don't see anyway that would make this work any better than the Spiral, Maze 3-1, or TC-4. Not like 1-3C is going to make a whole hell of a lot of difference cooling a 156+ watt pelt anyway. If done right your going to be below the point of diminishing (sp?) returns anyway.

But this is IMO. And IMO pelt cooling is a waste of time, money, and energy aswell.

Better of skipping it and dumping your money into a phase change system and do it right from the get go.

bigben2k · 04-25-2003, 05:12 PM

What I don't like about it is having to design and build:
-a manifold for inlet/outlet
-some way of clamping the cold plate

but it can be done.

I think it would be easiest to sacrifice a few channels, for the sake of structural integrity, but how they're going to be filled, and with what, I don't know.

Either way, 0.15 c/w is too optimistic: the channel ratio is not optimal, and you can't fix it, but it should match a TC-4 (my prediction).

You're gonna have to be pretty creative to get this to work right, and it's going to be a lot of work.

Good luck!

koslov · 04-25-2003, 06:00 PM

Noooooooooo, hit keyboard shortcut for "back" by mistake and lost huge post....

Ok, here we go again:

Here is how I plan to distribute the flow evenly:

I will drill a V-shaped slit into the Lexan, beneath the barb. It is basically like a nozzle on some garden hoses, but the area of the manifold outlet will be about equal to the barb's cross section. The hard part will be drilling into the Lexan at an angle, but it is doable.

JD, I think your evaluation of minichannels for pelt cooling is off. Pelts seem almost better suited for minichannels since you have a larger surface area over which to put fins.

As for pressure drop, consider this:

Cross section of 1/2" tubing: 500mm²
Cross section of 1/2" poly barb: 314mm²
Cross section of flow manifold outlet (6 x 50mm): 300mm²
Cross section of block (5mm fins x 50mm width): 250mm²

Water flows from 500mm² to 314mm² to 300mm² to 250mm². As you can see, most of the pressure drop comes from going from the hose to the barb.

Ben: I don't think .15°C/W is that unreasonable. Remember, I am cooling a 1600mm² area, not the tiny 100mm² area of a CPU. °C/W decreases as surface area increases.

Clamping will be done through bolts going directly through the baseplate. They will not be attached to the top at all, so therefore structural integrity wrt to clamping force is a non-issue.

edit: One more thing I wanted to address:

Hot side thermal resistance greatly affects the core temp of pelt-cooled system:

This is the difference between a .20 °C/W block and a .15 °C/W block. The .02 added on is estimated °C/W of the radiator.

BTW, I used .1 °C/W for cold side resistance, but this was just a wild guess. Anyone have any clue as to what the real value might be for a 100mm² CPU and 1/4" copper cold plate?

jaydee · 04-25-2003, 06:17 PM

Quote:

Originally posted by koslov

JD, I think your evaluation of minichannels for pelt cooling is off. Pelts seem almost better suited for minichannels since you have a larger surface area over which to put fins.

I am not arguing that, I am saying you will not be able to make use of those fins. That manifold with slit isn't going to work. It is to long with to small of an inlet. Maybe if you used three 1/2" inlets with the same slit? With that big slit the water will still just use the middle channels and not the outter ones. Unless you can create enough pressure to force water through all those channels, but then you would need outlets that would be the same as the inlet. Because trying to force to water out two 1/2" outlets on the side is going to make the water in the outer channels just fill up and do nothing. Water will take the route of least resistance which will be strait down and left to right in the middle of the block assuming the outlets are in the middle of the block. If you had 3 inlets in line, and 3 outlets in line on both sides you might be able to get the water to go somewhat evenly through all the channels.

koslov · 04-25-2003, 06:32 PM

F*ck, I hit back again... was the Intellimouse

JD, I see what you're saying, but I think it is just a matter of matching the manifold outlet shape to the target flow rate (lol, did I say "just"?). Imagine a 1 x 50mm outlet for a moment... don't you think that even at low flow rates, the water would still flow pretty evenly?

I will try to find the right shape by first making a test manifold, then hooking my pump to it (plus rad), and seeing how well the water flows evenly.

Oops, I forgot to mention I am using a pump with 8.5m of head

. CSL Velocity T-3.

jaydee · 04-25-2003, 07:01 PM

Quote:

Originally posted by koslov
[b]F*ck, I hit back again... was the Intellimouse

Hate it when that happens! When I am using the touchpad on my laptop similar things happen..

Quote:

JD, I see what you're saying, but I think it is just a matter of matching the manifold outlet shape to the target flow rate (lol, did I say "just"?). Imagine a 1 x 50mm outlet for a moment... don't you think that even at low flow rates, the water would still flow pretty evenly?

I will try to find the right shape by first making a test manifold, then hooking my pump to it (plus rad), and seeing how well the water flows evenly.

Oops, I forgot to mention I am using a pump with 8.5m of head

. CSL Velocity T-3.

Sounds like your all over it! Let us know how that comes out. That is just a sweet peice of copper. Almost be a shame to hack it up.....Almost.

Thats a sweet looking pump to!

MadDogMe · 04-26-2003, 04:08 AM

The only shame IMO is that it's soldered on fins, not skived...

Personaly I think 50mm is'nt that wide. What barb/tube are you using as an inlet?. I would'nt use a threaded barb as they're so restrictive, I'd use 13mm Cu pipe(15 outer) matched to 13mm I/D silicone tubing. Then I'd work out the internal surface area of the 13mm circle and make my inlet 'channel' area slightly smaller than (or equal to) the pipes area. You would have to try and maintane(sp?) this 'area size' through the transition from circle/pipe to slit/channel to keep it optimised though...

I dont see a problem with the coldplate mounting at 3.5mm, it may take more than 4 bolt/screws to achieve 100PSI needed (112 IIRC) though

, as 3.5mm is'nt much thread depth. I'd thread them into the B/P only, not the poly top!, outside of the waterseal (O~ring?)...

Nice to see a TEC head here, I disagree with JD on this, I think they're more 'efficient' than phasechange. Phasechange is alot of work IMO. TEC's are such a small bit more than normal H2o, another PSU, a coldplate and some condensation proofing...

The only thing putting me off of a 220watt TEC is getting my paws on a PSU for it!, (13~15V, 30amps: continuous not max peak)...

Les · 04-26-2003, 04:55 AM

koslov
As indicated in your other thread* some of my relevant thoughts and experimental results may interest:
http://forums.procooling.com/vbb/sho...5&pagenumber=1

*http://forums.procooling.com/vbb/sho...&threadid=6471

koslov · 04-26-2003, 06:43 AM

The heatsink is skived, you can tell when the base of the fins are bent.

Thanks Les. I plan on running the pelt at 20v perhaps. I could run them at Vmax, but Kryotherm turned up higher temps.

Hmm, come to think of it, I could run two in series, consuming only 144W at 20V, and still get a very nice temp of around 5°C. Well, I certainly have enough surface area to work with. Hell, I could *almost* fit four on there! And power supplies aren't a problem, I have one 24V 600W PSU, one 24V 240W PSU, and one 14V 320W'er. Not that I plan on using every one!

If I used more than one pelt, I would have to be concerned about the heatspreading in the coldplate. Would a hotspot form in the junction between the two TECs (ie directly above the CPU, the last place you want a hotspot)? Do you have any clues?

koslov · 07-19-2003, 01:21 AM

My progress:

Obviously this is not a finished product; in fact, I might redo the Lexan top.

Tools: 3-axis NC mill, Pencil grinder, Air motor with cutoff wheel, Dremel mounted in Router Table, 3/8" NPT tap, Sandpaper 600-2000 grit.

Issues with current proto:

One thread is counter-threaded, causing barb to sit slightly crooked. Possibly not making a watertight seal either.
Flow manifold ( channel connecting barb to fin pocket) too wide, water is not channelled evenly.

To be done:

Create second flow manifold
Drill clearance for bolts into Lexan top
Tap baseplate for bolts to Lexan
Tap baseplate for bolts to coldplate
Cut barb threads down so they sit flush when threaded in
Bolt and seal top to base with Marine Goop
Lap base flat on calibrated granite slab
Thread in barbs with Teflon tape
Leak test

After I finish the second flow manifold with the pencil grinder, I will know whether or not I need to make a new top.

JFettig · 07-19-2003, 08:11 AM

it looks like you need a plate with a manifold in it. Kinda like my gpu block but I make mine with 1 plate because of the side enterance.

Basicly cut it out like you have it. but at the ends, make 1/16" or 3/32" slits that go all the way thru, and mill down so its really thin area like that, with a 3/8 end mill so theres area for the water to go side to side.

sorry, my dad took my digital camera on a motorcycle trip, so I had to use the cheap webcam.

basicly make one plate that looks like that with the botttom milled out for the fins, then make the top with the hosebarbs in it, and a slot for water to travel to the ends of it easily. Neither the way you have it and the way I told you will flow perfectly, but I beleve mine will flow more decently with lots more flow.

Thats the first pic of my gpu manifold on the web

Jon

koslov · 08-04-2003, 02:04 AM

Well, the block and coldplate are going in the oven tomorrow for annealing (1400F!). I will be using argon to keep out impurities.

Everything is milled and tapped, and I milled out that ugly "flow manifold" I made with the pencil grinder. There is now a special inlet and outlet channel on each side.

Once annealed, I just need to lap then seal up the block. I am not quite ready to mount it yet, I am still deciding on how I am going to go about condensation-proofing.

04-25-2003, 12:48 AM	#1
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	Heatsink Conversion -> Pelt Waterblock After reading about some people's success with converting pin- or fin-type heatsinks into waterblocks, I decided to give it a try myself. I chose the Dynatron DC1U-B04, for it's high density fins and large surface area. It also has a flat bottom, with no groove for the socket cam. Well, just how big is it? Full specs: Material: Copper C1020 Dimension: 100x74x29 Pitch: 1.5 mm Thickness of fin: 0.45 mm Number of fins: 49 Height of fin: 9 mm Height of base: 3.5 mm Length of base: 100 mm The plan: This Socket 940 (Hammer) heatsink will be transformed into a waterblock for a 40 x 40mm peltier, to go on a socket A. As you can see from the last pic, the fin ratio is imperfect for watercooling, but I am still expecting good performance. I plan to chop off most of the fin area, leaving only a 50 x 50 area left in the center. Much of the baseplate will have to be chopped off too, just to fit on a Socket A mobo. The fins will also be cut down to a height of 4-5mm. I will top the block with 1/2" Lexan, which will have two barbs. Each entry will have a flow separator channel carved into the bottom of the Lexan, to try to equalize flow among the fins. Construction First, I would like to cut the baseplate to size (probably 80 x 74mm) with a table saw, leaving small pieces with fins still attached. These I'll use for VGA RAM cooling . Then I want to trim back the fins to their 50 x 50mm area, probably using my dremel. Next I'll sand the fins down to a height of 4-5mm using a belt sander. The Lexan top will be drilled out to make a cap for the block, with the cavity covering the fins. The flow separators will also be made using a drill press. I'll tap the Lexan about halfway in for the Poly barbs, leaving room for the flow separator. The mounting to the coldplate will be accomplished with 4 screws going through the baseplate. PS: If you want to order one, you can email or call Dynatron directly, and they will send you your heatsink for a small shipping charge and sales tax if you live in CA. This one was $29, plus $6 UPS ground and sales tax. Last edited by koslov; 04-25-2003 at 12:53 AM.

04-25-2003, 12:45 PM	#3
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	You can also pick up any of the other Dynatron HSF: some of them are skivved, but otherwise have a good ratio of fin-to-channel, for watercooling. They're relatively cheap too. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

04-25-2003, 04:37 PM	#7
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	I disagree, somewhat. The issue here is that you may have some difficulty getting the right clamping pressure on the pelt. I'm not sure if the fins can actually support the required clamping pressure without collapsing. The Spir@l is ideal, because you can simply tap into the four corners, which are relatively massive. Otherwise, if you can distribute the flow evenly, as much as possible, this block would do well with a pelt, but you're looking at a lot of work to get this done. JFettig made a channel block that looks pretty good for pelt use. Cathar is also working on one. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

04-25-2003, 05:12 PM	#9
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	What I don't like about it is having to design and build: -a manifold for inlet/outlet -some way of clamping the cold plate but it can be done. I think it would be easiest to sacrifice a few channels, for the sake of structural integrity, but how they're going to be filled, and with what, I don't know. Either way, 0.15 c/w is too optimistic: the channel ratio is not optimal, and you can't fix it, but it should match a TC-4 (my prediction). You're gonna have to be pretty creative to get this to work right, and it's going to be a lot of work. Good luck! __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

04-25-2003, 06:00 PM	#10
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	Noooooooooo, hit keyboard shortcut for "back" by mistake and lost huge post.... Ok, here we go again: Here is how I plan to distribute the flow evenly: I will drill a V-shaped slit into the Lexan, beneath the barb. It is basically like a nozzle on some garden hoses, but the area of the manifold outlet will be about equal to the barb's cross section. The hard part will be drilling into the Lexan at an angle, but it is doable. JD, I think your evaluation of minichannels for pelt cooling is off. Pelts seem almost better suited for minichannels since you have a larger surface area over which to put fins. As for pressure drop, consider this: Cross section of 1/2" tubing: 500mm² Cross section of 1/2" poly barb: 314mm² Cross section of flow manifold outlet (6 x 50mm): 300mm² Cross section of block (5mm fins x 50mm width): 250mm² Water flows from 500mm² to 314mm² to 300mm² to 250mm². As you can see, most of the pressure drop comes from going from the hose to the barb. Ben: I don't think .15°C/W is that unreasonable. Remember, I am cooling a 1600mm² area, not the tiny 100mm² area of a CPU. °C/W decreases as surface area increases. Clamping will be done through bolts going directly through the baseplate. They will not be attached to the top at all, so therefore structural integrity wrt to clamping force is a non-issue. edit: One more thing I wanted to address: Hot side thermal resistance greatly affects the core temp of pelt-cooled system: This is the difference between a .20 °C/W block and a .15 °C/W block. The .02 added on is estimated °C/W of the radiator. BTW, I used .1 °C/W for cold side resistance, but this was just a wild guess. Anyone have any clue as to what the real value might be for a 100mm² CPU and 1/4" copper cold plate? Last edited by koslov; 04-25-2003 at 06:19 PM.

04-25-2003, 04:17 PM	#5
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	I probably would have gotten a commercial block, but few of them really seem fit for pelt use. Only the Swiftech's (which aren't cheap either) and perhaps #Rotor's design. But I think I can do better than both with this block... I am hoping for a °C/W of .15 or better with it, which is not unreasonable considering the bigger surface area of a pelt. I would have picked a heatsink with a better fin ratio, but none would have given me as much room to work with as this one. This is how I came up with size requirements: 40 x 40mm area to be cooled, add 5mm to each side to allow for heat spreading in the baseplate, then add 10mm to each side for the Lexan top to attach to. This leaves 70 x 70mm. I guess I could have gotten away with a lot less if I didn't want a Lexan top. I could have had something closer to 50 x 50mm if I had just decided to solder a copper box top on. But I like working with Lexan, and this way I can remove the top for cleaning. Anyway, what I was really hoping for was some construction tips. Should I use a table saw to cut the base? Do I need a special blade?

04-25-2003, 04:30 PM	#6
jaydee Put up or Shut Up Join Date: Dec 2001 Location: Spokane WA Posts: 6,506	Table saw to cut the base? Wouldn't recommend it. If you do you will need a $50+ saw blade capable of cutting metal. I am sure your familiar with www.overclockers.com. Hoot was making such a block and was using a table saw. PM and ask him which blade he is using. I bet it is $50 or better. I havn't figured out how you plan on spreading the watter to all those channels effeciently. Your going to need a hell of a flow rate to cool a pelt off so you don't want a to restrictive block unless you are going to use a 500+GPH pump or something. The best block I can think of for pelts is DTek's big Spiral that Fixittt designed. Deep channels and lots of flow over a lot of surface area. Just my opinion but I don't think micro channels is the best way to go for pelt cooling.

04-25-2003, 06:32 PM	#12
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	F*ck, I hit back again... was the Intellimouse JD, I see what you're saying, but I think it is just a matter of matching the manifold outlet shape to the target flow rate (lol, did I say "just"?). Imagine a 1 x 50mm outlet for a moment... don't you think that even at low flow rates, the water would still flow pretty evenly? I will try to find the right shape by first making a test manifold, then hooking my pump to it (plus rad), and seeing how well the water flows evenly. Oops, I forgot to mention I am using a pump with 8.5m of head . CSL Velocity T-3.

04-26-2003, 04:08 AM	#14
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	The only shame IMO is that it's soldered on fins, not skived... Personaly I think 50mm is'nt that wide. What barb/tube are you using as an inlet?. I would'nt use a threaded barb as they're so restrictive, I'd use 13mm Cu pipe(15 outer) matched to 13mm I/D silicone tubing. Then I'd work out the internal surface area of the 13mm circle and make my inlet 'channel' area slightly smaller than (or equal to) the pipes area. You would have to try and maintane(sp?) this 'area size' through the transition from circle/pipe to slit/channel to keep it optimised though... I dont see a problem with the coldplate mounting at 3.5mm, it may take more than 4 bolt/screws to achieve 100PSI needed (112 IIRC) though , as 3.5mm is'nt much thread depth. I'd thread them into the B/P only, not the poly top!, outside of the waterseal (O~ring?)... Nice to see a TEC head here, I disagree with JD on this, I think they're more 'efficient' than phasechange. Phasechange is alot of work IMO. TEC's are such a small bit more than normal H2o, another PSU, a coldplate and some condensation proofing... The only thing putting me off of a 220watt TEC is getting my paws on a PSU for it!, (13~15V, 30amps: continuous not max peak)...

04-26-2003, 04:55 AM	#15
Les Cooling Savant Join Date: Oct 2001 Location: Wigan UK Posts: 929	koslov As indicated in your other thread* some of my relevant thoughts and experimental results may interest: http://forums.procooling.com/vbb/sho...5&pagenumber=1 *http://forums.procooling.com/vbb/sho...&threadid=6471

04-26-2003, 06:43 AM	#16
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	The heatsink is skived, you can tell when the base of the fins are bent. Thanks Les. I plan on running the pelt at 20v perhaps. I could run them at Vmax, but Kryotherm turned up higher temps. Hmm, come to think of it, I could run two in series, consuming only 144W at 20V, and still get a very nice temp of around 5°C. Well, I certainly have enough surface area to work with. Hell, I could almost fit four on there! And power supplies aren't a problem, I have one 24V 600W PSU, one 24V 240W PSU, and one 14V 320W'er. Not that I plan on using every one! If I used more than one pelt, I would have to be concerned about the heatspreading in the coldplate. Would a hotspot form in the junction between the two TECs (ie directly above the CPU, the last place you want a hotspot)? Do you have any clues?

07-19-2003, 01:21 AM	#17
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	My progress: Obviously this is not a finished product; in fact, I might redo the Lexan top. Tools: 3-axis NC mill, Pencil grinder, Air motor with cutoff wheel, Dremel mounted in Router Table, 3/8" NPT tap, Sandpaper 600-2000 grit. Issues with current proto: One thread is counter-threaded, causing barb to sit slightly crooked. Possibly not making a watertight seal either. Flow manifold ( channel connecting barb to fin pocket) too wide, water is not channelled evenly. To be done: Create second flow manifold Drill clearance for bolts into Lexan top Tap baseplate for bolts to Lexan Tap baseplate for bolts to coldplate Cut barb threads down so they sit flush when threaded in Bolt and seal top to base with Marine Goop Lap base flat on calibrated granite slab Thread in barbs with Teflon tape Leak test After I finish the second flow manifold with the pencil grinder, I will know whether or not I need to make a new top. Last edited by koslov; 07-19-2003 at 01:46 AM.

07-19-2003, 08:11 AM	#18
JFettig Cooling Savant Join Date: Feb 2003 Location: Willmar MN/Fargo ND Posts: 504	it looks like you need a plate with a manifold in it. Kinda like my gpu block but I make mine with 1 plate because of the side enterance. Basicly cut it out like you have it. but at the ends, make 1/16" or 3/32" slits that go all the way thru, and mill down so its really thin area like that, with a 3/8 end mill so theres area for the water to go side to side. sorry, my dad took my digital camera on a motorcycle trip, so I had to use the cheap webcam. basicly make one plate that looks like that with the botttom milled out for the fins, then make the top with the hosebarbs in it, and a slot for water to travel to the ends of it easily. Neither the way you have it and the way I told you will flow perfectly, but I beleve mine will flow more decently with lots more flow. Thats the first pic of my gpu manifold on the web Jon

08-04-2003, 02:04 AM	#19
koslov Cooling Savant Join Date: Jul 2002 Location: Posts: 264	Well, the block and coldplate are going in the oven tomorrow for annealing (1400F!). I will be using argon to keep out impurities. Everything is milled and tapped, and I milled out that ugly "flow manifold" I made with the pencil grinder. There is now a special inlet and outlet channel on each side. Once annealed, I just need to lap then seal up the block. I am not quite ready to mount it yet, I am still deciding on how I am going to go about condensation-proofing.

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