Closed loop and water peltier cooled water?

[Mbeach]

Hey all!
My first post here.. been reading quite a bit before deciding to introduce myself.
Well, been addicted to watercooling cpu's almost 2.5years now. A system is never perfect, always new ideas and new projects. I gave up using the peltier on S370 mobo (CUSL2) because of the hazzle to get it air tight and reduce the risk of condensation. Of course it is doable very easy, but i don't wanna add silicone and other non-removal stuff on the mobo. Anyhow, i have to settle with my P3 750 running a 133FSB which is fine for now. Pics from the beginning to the complete setup you find here!

As you can see from the pics i'm using a rather large reservoir (2 litres) and now during the summer the water get of course warmer which conludes that my cpu is getting warmer of course. So i've been thinking in the same directions as Insomnia did here and it seemed to work good. A 16degF decrease, that's about 9degC. 9-10degC would set my watertemps to about 24-25degC (~75degF) which i believe would help me push that damn P3 to a 140-145FSB. But it would take way too long time to cool down 2.5litres 9-10degC, so need to reduce the amount of water or even better remove the reservoir and use a closed loop. For that i need a new pump so i ordered this one:

Eheim 1046


This is my plan:
Since i already have one (drain-)pump, reservoir, peltier and two spare waterblocks i could cool the water down by clamping the two waterblocks together with the 71W peltier in between. The cool waterblock is the water for the cpu and gfx board and completly in a closed loop which has less water without a reservoir. The hot side has it's own pump, smaller reservoir and uses of course the radiator to cool the water.

So what you all guys think? A GO or a NO-GO??

Below is my stuff:
Aopen HX08
Asus CUSL2
P3 750@133FSB, 1.85V
Asus V7700T/Pro 64Mb DDR
2x128Mb PQI Power Module PC-166
40Gb IBM HD
SB live! 1024 value
Realtek NIC
HP7200 CDRW
Hitachi DVD

[Mbeach]

^BumP^

[bach-on]

Some have and are trying setups similar to what you're describing. I tried it myself a few months back. The results were disappointing.

Some theories are that the water just can't seem to take in cool as efficiently as it gives up heat. The water doesn't stay in the cooling block long enough to get very chilled. I spent nearly a hundred bucks experimenting and trying to accomplish this. I never saw any real advantage in temps or in overclocking ability.

Maybe you can do it better. But there was some guy in Australia who had something like three or four drift 180 watt TECs in an attempt to chill the water. He said the problem was that the water on the hot side heated up faster than the water could be chilled on the cool side.

Don't wanna pour rain on your parade, but you asked.

[Amy]

Originally posted by Bach-On!!!:
Some have and are trying setups similar to what you're describing. I tried it myself a few months back. The results were disappointing.

Some theories are that the water just can't seem to take in cool as efficiently as it gives up heat. The water doesn't stay in the cooling block long enough to get very chilled. I spent nearly a hundred bucks experimenting and trying to accomplish this. I never saw any real advantage in temps or in overclocking ability.

Maybe you can do it better. But there was some guy in Australia who had something like three or four drift 180 watt TECs in an attempt to chill the water. He said the problem was that the water on the hot side heated up faster than the water could be chilled on the cool side.

Don't wanna pour rain on your parade, but you asked.

BachOn:

You may have been confused of what needs to be done. Actually 2 80W 8Amp potted pelts
will give you - A very acceptable range of cooling capability.



Cheers Amy -

Don't try this with one pump Curcuit like K did. Duh

[Mbeach]

LOL!!! Great pic!!

[bach-on]

Please allow me to explain a bit.

First, I misread the original post. I thought he was talking about using the chiller concept to overpower a TEC mounted directly on his CPU. That was not the case.

Yes, you can cool the water a bit with the kind of concept Amy showed in her diagram (nice work!). Maybe my water blocks weren't as efficient as those Amy indicated, but I followed pretty much the same overall concept she did with two 72 watt TECs at 14 volts. I DID see a difference of maybe 2-5 degrees F. But that was all. Maybe that kind of difference is sufficient for many. Maybe I was expecting too much.

I just did not see the kinds of cooling gains I had been hoping for - namely temps in the mid to upper 20's at full load. Perhaps I just had unrealistic expectations.

If we live long enough and learn enough, maybe we become wise.

[CapnBry]

I've got a TBird 1.33Ghz on an Abit KT7A-Raid.

I started experimenting with this about 2 or 3 weeks ago. I started by experimenting. I took a 4x4 watertight electrical enclosure and cut a hole in the top just not big enough for a heatsink bottom to fit through. I attached it with plumbers goop. Then I put the 80W peltier on, and strapped a huge (like 6x4) heatsink with a low speed 92mm fan. Filled with water, it got down to 40F in an hour. Good!

Then I fit some 3/8" barbed fittings in the side and attached it to my current H20 system. I ran the water from the current resivour (damn I can't spell that) up to the CPU, down to the radiator, then to the new peltier chiller.

With the system off, and the pump running, the water gets about 10F below ambient. I think this is because the radiator lets any coolness out just like it lets the heat out. With the system on, and running at full throttle, the temp gets up to ~130F (compared to about 150F earlier). The system runs at under 98F at idle.

Something about my temperatures. I think the sensor on my Abit is a big fubared, because with a Thermoengine with a Delta 7k, it would still get up to 150F under high load.

I've got two new peltiers on order, because I think I fried my current one when some neoprene melted inside of it (don't ask). I'll keep you guys posted if you want.

Cap'n Bry

[Mbeach]

Yeah, pls keep us posted...
I haven' implemented my test setup yet... but i don't think it won't fit in my case anyway...

[Joe]

Amy I like that design layout!

[Amy]

Originally posted by Joe:
Amy I like that design layout!

Thanks Joe, a compliment coming from you.

This is a second pass of January's experiment. But with higher clamping pressure on TEC's
http://www.octools.com/ocforum/cgi-b...num=1000329765

Cheers Amy - A.C.

[ 09-18-2001: Message edited by: Amy ]

[jastrckl]

Amy, after looking over your experiments and setup, I have to say that as long as you look better than a damn poodle, you are most definitely life partner material. I doubt I'd be the first to find you though :/ I'm pretty into this stuff, talking shop some day would be fun. :P I have a tbird 850 I plan on pelting with a pair of 172's @ 24V... I found a pretty good linear supply to power it all too, and not too expensive. Just waiting for someone to design a decent dual pelt waterblock-right now, NO ONE gets them right. As you've noticed, clamping pressure of pelts is _critical_. I wish I had my own shop to make one. Anyway, happy hunting.
-Josh

[jastrckl]

lol, I meant to type "damp poodle". I suppose both apply-'damn damp poodle'.

[resago]

2 questions, whats wrong with the maze2.2?

and

where did you find your power supply?
do you have a link?

[jastrckl]

The maze 2.2 doesn't provide for adequate clamping pressure of pelts, restricts flow too much (I've seen some really promising tests of flood vs maze type blocks, and the flood types seem to do better) and it's made out of the wrong material. Silver wouldn't be practical because of the loads involved, but CuSil would be ideal (it's an alloy of copper and silver that conducts better than either and I think it's stronger than copper as a bonus-it's mad expensive though due to patents). As for the power supply issue, check out car battery chargers. Advance Auto Parts in my town sells a 5/10/20A DC supply @ 13.8V for about $40 on sale right now. They're bulky, but you can't beat the price. It will be a cleaner output than a switching supply, but not as nice as some of the fancy regulated supplies people use. Peltiers aren't picky enough to really need super-clean supplies so I don't care much.

[Jonas]

my personal take on chillers...not using it because i had leaks from sloppy cutting...it'll be round 2 some time next week.



and from the top


with this setup i was going for 2 things. the first is a counter to the problem of most chillers: the water doesn't stay around longenough to lose its heat. hence the stackedplate rad. then i used my old mc2001 from swiftech and lopped it into 2 peices. with an 80w pelt sandwiched between. then i was worried about the water not being moved around inside the container, so i added a tiny rio 40gph pump that "draws" through the pins of the inside hs, and "blows" out onto the radiator. then i filled the whole shebang up with a mix of water wetter/antifreeze/water and fired it up. the temps inside reached about 10c in under 2mins with no load. never got the chance to hook it to my system because the damn thing leaked. but it shows it works! it was about 2 mins into it that the damn thing started leaking so i am not sure how low it would have gone.
its possible with enough tecs and a large enough containerthat you could dump a larger rad into it and get rid of the air cooled rad altogether. but i think it would be a bad idea. and be really inefficient...the whole goal of this was just to reduce the coolant temp by around 5-10c.

[rayita]

jastrckl, the maze2.2 not provide enough claping pressure for TEC? With 4 screws? are you sure?

ok, whats the correct design for a WB you suggest to provide the necessary clapping pressure?

[Butcher]

correct clamping pressure for a TEC is about 200 psi, I don't think the dd maze2 allows that much whithout breaking the block - the sttel (I'm assuming they're steel) screws would tear the threads out of the softer copper block.

I'm planning to clamp peltiers with steel bar and steel screws / nuts, should allow very high clamping pressures as steel is a lot harder than copper.

[Amy]

Originally posted by jastrckl:
Amy, after looking over your experiments and setup, I have to say that as long as you look better than a damn poodle, you are most definitely life partner material. I doubt I'd be the first to find you though :/ I'm pretty into this stuff, talking shop some day would be fun. :P I have a tbird 850 I plan on pelting with a pair of 172's @ 24V... I found a pretty good linear supply to power it all too, and not too expensive. Just waiting for someone to design a decent dual pelt waterblock-right now, NO ONE gets them right. As you've noticed, clamping pressure of pelts is _critical_. I wish I had my own shop to make one. Anyway, happy hunting.
-Josh

jastrckl:

I have five grand children.. Please cool yourself, not your PC... Ha !

Cheers Amy Croft
-