View Full Version : Dual 172 watt pelts in a Maze 2.2, Results and a pic
I finally got the dual 172 pelts up in the Maze2.2. Ambient is 27.8C, case 30.9C, air exiting radiator (drawn from inside the case) 35C, water 52.3C, socket thermistor idles at 9 to 10C and full load at 35C. This is with a Duron at 1016 mHz and a VCore of 2.04 volts, the Birds can wait until this new configuration is sorted out. The pelts are drawing 19 amps at 20 volts.
While I’m pleased that nothing went up in smoke this time, it’s obvious the heat is not getting out of the coolant fast enough. I picked up the parts for a cooling tower yesterday and should have some results to post by the end of the weekend.
A big thanks to Dan, Dennis and Jeremy at Danger Den, the high performance cooling artists.
Here is a pic of the Maze 2.2 installed on my Abit KT7A-R.
[img="http://albums.photopoint.com/j/View?u=1551335&a=11768501&p=48642929"]
I am curious to see the results of the cooling tower as I have been kicking around the same idea of dual 172's/DD Maze 2-2. I have all the parts here but the only holdup is I have been thinking of a way to get the heat out of the water. It is very humid where I am in the summer so I am not sure if evaporative cooling is going to hold up, at least with swamp coolers they seem to work better in dry areas. My chilled water setup is out of the question it is only good for about 100w or maybe a little more of heat.
I think the cooling tower & dual 172 pelts will kick the proverbial fanny out of anything short of a Vapochill.
GERRY136
05-17-01, 02:05 PM
can u guys guide me towards some info on this vapochill process? i have a slot a athon.
thanx.
Tom's Hardware did a very complete review of the Vapochill.
Wow! 52 DegC water temp you need to watch your hoses don't melt !!.
Would a bloody great big radiator be the way to go to cool you water (maybe one off a big motorbike c/w fan or similar)? bit more practical than a tower.
Until you get a better way of cooling the water like you said then you'll probably get MUCH better temps with just one of those peltiers. I saw a review where a guy used either a 156W or 172W and was getting below-freezing temps and rising to about 6C under full load. Btw, what is your water pump rated at? Withg the amount of heat that is creating you'll pretty juch want at least 350Gph. Anything less than that and the water will become oversaturated with heat inside the waterblock and you'll lose thermal efficiency... but then again it won't matter if you're radiator can't handle the load. I look forward to seeing your tower results.
-=mac=-
Mac42 (May 17, 2001 05:17 p.m.):
Until you get a better way of cooling the water like you said then you'll probably get MUCH better temps with just one of those peltiers. I saw a review where a guy used either a 156W or 172W and was getting below-freezing temps and rising to about 6C under full load. Btw, what is your water pump rated at? Withg the amount of heat that is creating you'll pretty juch want at least 350Gph. Anything less than that and the water will become oversaturated with heat inside the waterblock and you'll lose thermal efficiency... but then again it won't matter if you're radiator can't handle the load. I look forward to seeing your tower results.
-=mac=-
I am not an idiot. :) I do understand what is going on here. I have tried both a 156 watt pelt and a single 172. With lower ambient temps, I still could not get to 6C at full load. As I recall it was about 17C. That was three months ago. My pump is a Danner 700.
I planned to use a cooling tower from the beginning of this project. First I wanted to see what would happen in my current system. It also provided a convenient test bed. The cooling tower should be installed tomorrow.
mrpcman
05-17-01, 11:32 PM
I have the exact same peltier/waterblock. it's a nice setup, but it would work better if I had a more powerful PSU.
I would be interested in how this block setup cools a T-bird 1.33ghz.
eradda02
05-18-01, 09:48 PM
you say you have your radiator draw air from the case??? am i right??? if i am that is baaaaad, the air in your case is warmer so the water will cool down slower...
Ms. Nath,
I would rather do the R&D with a $39 CPU. I will put my 1333 Bird in eventually. Hold on to your hat.
eradda02,
Until I tried this dual pelt set up, my case temp was always with in 1C of ambient, which does not have much of an affect on the radiator. Since the heat is no longer being removed from the water, the case heats up a bit. I'll have the cooling tower in tomorrow.
Colin- Where did you get your pelts?
Danger Den.
The Pignato Cooling Tower will be in shortly.
*spazzed*
05-19-01, 05:54 PM
I have that "cooler master" fan on my heatsink, both of which are stock on my HP. I dunno why mines loud though, maybe i've blown the bearings on it :D
that's gonna kick ass with the cooling tower, colin!!!!! get pics and results in here asap! ;) :)
dimmreaper
05-20-01, 02:51 AM
Have you tried running ice water through it's gills Colin?
I've toyed with Pelts and Ice water in the past with some good results. One of these days I'd like to try running dry ice cooled alcohol through a waterblock with a pelt underneath. That would be something to see . . . .
Keep us posted matey.
I have to apologize for the slow response on this project. I expected to have the cooling tower in Friday evening. Friday was my brother-in-law’s funeral. Then Saturday, my niece who has been in China for the last year and a half showed up and most of today will be spent working on my church’s radio station. Stay tuned.
Jeff,
I have not tried ice water. Too much of a hassle pulling the radiator out of the case. I don’t want to hack up this H20 rig. I will probably end up giving it to my wife and putting the pelt setup in another case.
Quick results: ambient 26.3C, water load 36.2C, CPU load 19C, water idle 35.5C, CPU idle -3C. Not quite what I expected. It seems the pelts put out a lot of heat to dispose of.
My wife has named the cooling tower El Fuente! Off to work on the radio station.
iznogoud
05-20-01, 03:22 PM
instead of 2 172 watt peltiers , wouldnt be more efficient to have 1 , say 400 watt peltier? I mean it must be possible and i bet they do exist. This way one eliminates the loss of energy transfer from one peltier to the other.
That's rather disapointing =(. I guess chilling the water is necessary for that much heat.
Carmine_Paterno
05-20-01, 08:04 PM
I think if you had a smaller pelt, it would cool better than the bigger ones. The hot side would get cooler, meaning the cold side would get cooler. Therefore a more efficent and cooler processor. Why did you put 2 172watt pelts on a Duron @ 1064? Shouldn't you be able to do that with air? Try 1 smaller pelt, and see how it reacts, maybe like a 128watt pelt. Or, get your water alot cooler before it hitts the pelts, and maybe put another radiator before it goes into the pump.
Back from working on the radio station tower.
First thing to try is a taller cooling tower and perhaps a different shower head. There is a 1250 GPH pump in Bakersfield. I may pick it up to try instead of the current 700 GPH one.
For those who can't be bothered to read the whole thread, I have tried smaller pelts and I am using a Duron for various reasons. One is I don't want to toast a Bird doing R&D.
Also to the best of my knowledge there is no 400 watt pelt availble. Current state of the art is a dual 172 watt setup. If we have a hard time dealing with the heat from about 344 watts, how do you expect to dispose of 400 watts of heat?
Lasse S.
05-21-01, 12:12 AM
Colin (May 20, 2001 11:32 p.m.):
Back from working on the radio station tower.
First thing to try is a taller cooling tower and perhaps a different shower head. There is a 1250 GPH pump in Bakersfield. I may pick it up to try instead of the current 700 GPH one.
For those who can't be bothered to read the whole thread, I have tried smaller pelts and I am using a Duron for various reasons. One is I don't want to toast a Bird doing R&D.
Also to the best of my knowledge there is no 400 watt pelt availble. Current state of the art is a dual 172 watt setup. If we have a hard time dealing with the heat from about 344 watts, how do you expect to dispose of 400 watts of heat?
you have already passed the 400w mark ;) 19amps*20volts=380Watts and you have to add your CPU to that too (say 50W probably more)...so maybe you are at 430W already =) hehe...you are dealing with an oven here...remember the pelt Qmax value is cooling power, not the power it radiates/consumes....
you need an awesome radiator for that...I like this one:
http://www.heatercore4u.com/2-746.htm
I'm using a similar one (same manufactorer) I found in a car wreck (85 model BMW 325i) it has an amazing fin density and 20 pipes....couple this with two powerful 120mm fans and things should look better....
also...pelts rated at Vmax=24Volts should be runned at 17-20V for maximum efficiency (85% of rated Vmax)...try it ;)
Lasse S.
05-21-01, 12:22 AM
pics of rad mentioned above
GERRY136
05-21-01, 12:23 AM
Colin (May 17, 2001 02:19 p.m.):
Tom's Hardware did a very complete review of the Vapochill.
thanx colin!
Lasse S.
05-21-01, 12:25 AM
another one
325 heater cores are aluminum. Not going to mix well with an copper water block.
20 volts is the sweet spot. Raise or lower the voltage and the temps climb.
dimmreaper
05-21-01, 12:34 AM
Hey Colin, I've got the transcooler off a Peterbilt that's about 4' x 5' (yes I do mean feet, not inches). Perhaps you could find something similar at a local bone yard?
Lasse S.
05-21-01, 12:52 AM
Colin (May 21, 2001 12:33 a.m.):
325 heater cores are aluminum. Not going to mix well with an copper water block.
20 volts is the sweet spot. Raise or lower the voltage and the temps climb.
yeah...sadly it's ALU (though it works very well)....do you know abt any good CU ones?
ah...I see, you have already tried the adjusting the voltage....it's a common mistake to think that Vmax is ideal...do you think the MAZE2 waterblock is efficient enough for the task? The drop from 53°C (coolant) to 35°C (CPU at load) seems VERY little....those pelts should be able to maintain a much higher deltaT even if the coolant is hot....try measuring the waterblock temp (at the top of it)...if it is much higher than the coolant temp it's not suited for the task...
Lasse S.
05-21-01, 01:11 AM
btw...I have a hard time understanding the logic behind a maze water-block design....sure, it will work good in a CPU only setup....but give it 200W or more (typical TEC setup's) and the coolant will heat up pretty quickly in there....as a result the coolant will travel around in there not doing anything good...wouldn't it be much better to new fresh and cold coolant in there instead?
maybe it's time to look at the coolant exchange rate of water-blocks??! and I really think site's like overclocker.com should start looking at water-block performance when used it TEC setup's too....as the CPU only test's are getting pretty boring (within 1-3°C difference)
With H20 at 37C the block is 40C.
Lasse S.
05-21-01, 01:25 AM
Colin (May 21, 2001 01:20 a.m.):
With H20 at 37C the block is 40C.
only 3°C difference? that's REALLY good....but where is the pelt cooling power lost? The drop from 53°C (coolant) to 35°C (CPU at load) seems VERY little....something is fishy here....
Lasse S. (May 21, 2001 01:27 a.m.):quote]
only 3°C difference? that's REALLY good....but where is the pelt cooling power lost? The drop from 53°C (coolant) to 35°C (CPU at load) seems VERY little....something is fishy here....
I agree. I have had similar results with just one 156 watt pelt.
I believe we may be expecting too much. The pelts are side by side. Running at 20 volts we can expect a Delta T of about 57C across the pelts. This is before thermal interface losses. Two pelts side by side increase the cold plates ability to handle higher heat loads but do not increase the Delta T, it stays at a constant 57C minus any losses. To get colder temps, the pelts need to be stacked or the water needs to be cooler. The benefit of running two pelts side by side should show up when I throw a hot Bird in the system. The temps should be close to the Durons.
Lasse S.
05-21-01, 11:49 AM
Colin (May 21, 2001 01:59 a.m.):
I agree. I have had similar results with just one 156 watt pelt.
you are only trying to cool down a 60W CPU...right? that should be a piece of cake for those two 172W pelts (even though the coolant temp isn't ideal)....I have used several different CPU/TEC setups in the past, and never seen performance as lousy as that (keeping the pelt Qmax/load ratio in mind)....I would suspect two things:
1. Either something is wrong with the installation (bad thermal contact)
2. Wrongly reported temps....
I know several people have had problems when using multiple TEC's with one water-block and coldplate....a friend of mine tried to build a chiller using two big waterblocks and three 80W TE Dist TEC's....he actually had four of those TEC's, but no matter what he did, he couldn't get three in there with a good mounting pressure...one would always be "loose"...he later figured out that there were height differences between the pelts....that's probably not a big issue when using only two...I would take everything apart and inspect it throughly, and try reinstalling it again....
Lasse S.
05-21-01, 11:53 AM
Colin (May 21, 2001 11:44 a.m.):
I believe we may be expecting too much. The pelts are side by side. Running at 20 volts we can expect a Delta T of about 57C across the pelts. This is before thermal interface losses. Two pelts side by side increase the cold plates ability to handle higher heat loads but do not increase the Delta T, it stays at a constant 57C minus any losses. To get colder temps, the pelts need to be stacked or the water needs to be cooler. The benefit of running two pelts side by side should show up when I throw a hot Bird in the system. The temps should be close to the Durons.
yeah...but are you getting a deltaT of 57°C?!? looks more like 15-30°C to me...what coldplate temps do you get?
Lasse S.
05-21-01, 11:59 AM
stacking TEC as big as these sounds like a bad idea....you get huge deltaT (theoretical) but lousy Qmax (less than the lowest Qmax rated pelt in the stack)
I tried that for fun once...a P3 1GHz, 80 TEC, and 156W TEC stack...performed worse than a single 80W TEC setup....
Lasse S.
05-21-01, 12:18 PM
and yes, this is expecting too much =) I can't think of any convient way of cooling a 100W CPU using TEC's directly on top and get good results (below 0°C at 100% load)...a good chiller setup might work though....
You are way beyond the limits of standard water cooling system here...thus you get HOT coolant =) hehe...but a NICE experiment anyway!!!
btw...I here's a chiller setup that gave me some interesting results...it invloves two water cooling loops:
Primary loop (CPU cooling):
1. Eheim 1048
2. Chiller block
3. CPU Water block
4. everything insulated
note! no rad in that loop...and keep the tubing as short as possible, and use as little coolant as possible...if too much coolant, it will take a LONG time for the temps to drop....
Secondary loop (Chiller cooling):
1. Eheim 1250
2. Chiller block
3. Good rad ;)
I only played with a 60W CPU and a 156W TEC at 13.8V....but coolant in the primary loop dropped to 7°C...the fans I used in this test wasn't powerful enough so the coolant in the secondary loop was 35-40°C....so there's a lot of room for improvement here....
Colin (May 21, 2001 11:44 a.m.):
I believe we may be expecting too much. The pelts are side by side. Running at 20 volts we can expect a Delta T of about 57C across the pelts. This is before thermal interface losses. Two pelts side by side increase the cold plates ability to handle higher heat loads but do not increase the Delta T, it stays at a constant 57C minus any losses. To get colder temps, the pelts need to be stacked or the water needs to be cooler. The benefit of running two pelts side by side should show up when I throw a hot Bird in the system. The temps should be close to the Durons.
Colin,
Great thread, really interesting stuff but disapointing results I'm sure you agree.
Have you tried monitoring the cold plate temp? If you have a peltier Delta T of 58 C and a water/cpu delta T of only 17 C, you must be getting poor heat transfer somewhere.
If there is a big Delta T between cpu and coldplate it would point to a problem tranfering the heat between the cpu to the cold plate (possibly because of the small contact area of the Duron ?). If the coldplate is similar temp to the cpu the problem must be heat transfer accross the pelts.
Just my thaughts.
Good Luck !
An interesting test would be to hook a garden hose up to the intake of the Maze block and see what the results are. At least the water temp will be good and you can get an idea of how much more effort it is worth to put into the progect.
There could also be a problem with flow through the block or with the block dissipating that much heat with the current flow rate. I wonder what the differance is between the in and out water temps of the block. DD does make there blocks available with 1/2" fittings. Changing the fittings could help net a higher flow rate if the passageways inside are also 1/2"
Lasse S. (May 21, 2001 11:49 a.m.):
[quote]Colin (May 21, 2001 01:59 a.m.):
I agree. I have had similar results with just one 156 watt pelt.
you are only trying to cool down a 60W CPU...right? that should be a piece of cake for those two 172W pelts (even though the coolant temp isn't ideal)....I have used several different CPU/TEC setups in the past, and never seen performance as lousy as that (keeping the pelt Qmax/load ratio in mind)....I would suspect two things:
1. Either something is wrong with the installation (bad thermal contact)
2. Wrongly reported temps....
quote]
The pelts and the Duron make nice Artic Silver impressions on the copper surfaces so I don’t think this is an issue.
As far as temps go, this is an in socket thermistor, so any temps are suspect. Cooling one side of a something that is hot in the middle and then reading the temps on the opposite side is not optimum. The thermistor is bent upward and centered under the die with Artic Silver on it. There is also neoprene inside the socket which further increases the contact pressure. It is possible that a part of the temperatures is because of this.
ken257 (May 21, 2001 06:22 p.m.):
An interesting test would be to hook a garden hose up to the intake of the Maze block and see what the results are. At least the water temp will be good and you can get an idea of how much more effort it is worth to put into the progect.
There could also be a problem with flow through the block or with the block dissipating that much heat with the current flow rate. I wonder what the differance is between the in and out water temps of the block. DD does make there blocks available with 1/2" fittings. Changing the fittings could help net a higher flow rate if the passageways inside are also 1/2"
The difference between water temp and the block is 3C which indicates the block is handling the heat ok. I tried inserting the Cooling Cube in the loop before the cooling tower. My water temps dropped 4.4C but the CPU temps are up a bit. I believe this is due to restriction in flow caused by the Cube. This leads me to believe a higher volume pump like the 1250 could make a significant difference.
Lasse S. (May 21, 2001 11:53 a.m.):
yeah...but are you getting a deltaT of 57°C?!? looks more like 15-30°C to me...what coldplate temps do you get?
57C is theoretical at 20 volts. Then there are losses in the cold plate and the three thermal interfaces, not to mention the steel bolts connecting the cold plate to the water block. The math does not quite add up but when one considers the inaccuracies of the socket thermistor, it starts to make sense.
The results I am getting are similar to the single 172 watt pelt tried. Two pelts side by side should not drop the temps much more than one, and then you have to dispose of all the heat they generate. What two pelts side by side will do is increase the load capacity of the system. I don’t believe a hot running Bird will raise the temps much over the Duron.
I have not taken the block off the mother board to measure the cold plate temps yet. I think that will have to wait until tomorrow.
My plan to increase flow hit a setback today. I drove 60 miles to Bakersfield to pick up a Danner 1200. When I got home, I noticed the impeller housing is cracked. I’ll see if I can locate another pump tomorrow. If all goes well, the new pelts should be here tomorrow too.
Since my cooling tower is now 7.8 feet tall, I have considered running the 1200 to pump the tower and the current Danner 700 to handle the Cooling Cube. They would both be in the same loop. With my simple mind, the math seems to work out. Any thoughts on this would be appreciated.
William
05-22-01, 09:16 PM
that might work really well, having a big pump to deal with only pumping the water up into the tall tower. Instead of the cube, which restricts the flow rate, what about a second, smaller cooling tower? That would be as good or better, and not require a pump with a ton of head.
Ok, here is a mind blower. The other cooling tower temps previously posted were with a 26” long 4” pipe in the cooling tower and a Danner 700 pump. For review they are: ambient 26.3C, H2O 36.2C, CPU load 19C, CPU idle –3C.
With a 26” pipe and a Danner 1800 pump the numbers are: ambient 27.2C, H2O 35.4 C, CPU load 35C and CPU idle 0C.
With a 64” pipe and a Danner 1800 pump the numbers are: ambient 26.8C, H2O 33.6C, CPU load 35 C and CPU idle 0C.
So what have we learned? It is possible to provide too much flow through a Maze 2.2 to extract the heat efficiently. This has a serious impact on the “flow threads”.
I’m not sure where to go now. I bigger cooling tower may help. A monster radiator might help. I will try the new pelts out of obligation to Danger Den. Getting rid of all of this heat is a problem. Perhaps dual 122 pelts are a better solution. The Vapochill is looking pretty damn sweet right now.
colin, wher are you located at ( kinda off the subject) and i believe that you are loosing efficiency when you uppued the flow because you are not giving the water enough time to absorb the heat properly, I gained most of my cooling experience with my pro street nova, huge coling problem, tried restrictor plates in the water to slow down the flow and when i got it optimal is was using a reduce( mounted in the thermostat housing) that changed the 1 7/8" hole down to a 3/4" inch hole, even better results( more stable at operationg temp) when i put in the 1/2" reducer. different subject all together but same concept, too much flow is bad and not enough is bad too, try puting a cheap valve somewhere inline( where i dont know but somewhere) and then see what different flow rates give you as far as getting the heat out of the water.
I have been having trouble getting the heat out of my water, as a result i am going to good ol' u-pick u-save and finding me a nice alum. rad out of a car, then i am going to get a few more fans, possibly one big one ( like a 15" perma cool that flows 2790cfm and I thing 16 or 18 volts) and then see what i get, also how can i up the voltage that i am supplying my fan just short of buying a high voltage external psu from all electronics.
well enough for now,
email me if you want and maybee we can meet up( i live in ontario,ca)to discuss or trade some parts around and experiment a bit, as i do have access to cnc mills and am an excelent cad designer.
Later
Tim
sorry the permacool fan draw is only 9.8amps
here is a pic also
i`ve lurking here for ages and it seems to me the cold plate is the problem.
have u tried a thicker cold plate???
or even getting one made out of cusil?
(22%copper,78%silver).
about 2cm thick.
we had this big discussion in OCAU about the optimal cold plate material and thickness.
and the cusil is rated the best thermal conductor there is .
i had one made up even though it costed me an arm and a leg but it was a top performer.
was going to get a water block made up also but never got to save up enough cash to get the project going.
if u r not strapped for cash and want an all out cooling performance then u should try it...
it will drop about +/-6degC off the full load temp just with the cusil coldplate...
GOD....
I agree the water is flowing too fast to extract the heat from the block. Kind of blows the theory of faster flow in the block and slower in the radiator apart.
As for the cold plate, I did not measure the thickness. I will when I try the new pelts. I am not dumping any more funds into an exotic cold plate. Two pelts on a hot processor may just be generating too much heat to deal with.
Just a thought but would'nt a very fast flow through a water block raise the possibility of cavitation occuring. This would cause the efficiency to drop dramatically (bubbles don't conduct/absorb heat). Any sharp edges or turns in the block could give rise to this problem.
Good point, especially considering the design of the block.
inertia
05-24-01, 02:05 PM
What did you pay for the thing? I know you said an arm and a leg, but . . .
Why is the coolant temp sooo gosh darned high? I've been lurking for a very long time and I've never seen anything that high. Let's hear some further details about your specific cooling tower - how you built it, fan or no fan, the show head - or the radiator you've been using.
inertia (May 24, 2001 02:05 p.m.):
What did you pay for the thing? I know you said an arm and a leg, but . . .
Why is the coolant temp sooo gosh darned high? I've been lurking for a very long time and I've never seen anything that high. Let's hear some further details about your specific cooling tower - how you built it, fan or no fan, the show head - or the radiator you've been using.
I am not sure what thing you are refering to. All of the specs are in the thread. As for the heat load, it's close to 450 watts, that's why it's so hard to cool the water.
Slain (May 24, 2001 12:46 p.m.):
Just a thought but would'nt a very fast flow through a water block raise the possibility of cavitation occuring. This would cause the efficiency to drop dramatically (bubbles don't conduct/absorb heat). Any sharp edges or turns in the block could give rise to this problem.
Cavitation is a possibility but you would sure be aware of it if it were happening, lots of noise and vibration.
If the water was not picking up the heat you would notice a sharp drop in water temp and a sharp increase in block temp, your water temp is down but this is maybe due to your tower, has the block temp climbed?.
For me the big mystery in all of this is why your not getting a more Delta T between your block and cpu temps.
The Delta T between the block and water is still 3C. There is no noise or extra vibration but if you look at the picture, the block is well damped. If there were vibration, Nevin pointed out that it would cause intermittant contact with the CPU an temps would go up. But this would be for both idle and load.
If the new pelts show up today, I will have more to say tomorrow.
William
05-24-01, 08:54 PM
what about introducing two resevoirs? One for the tower to get the water up into it, and another going from the tower to the cpu. Use the smaller pump on the computer circuit, and the larger one to pump up into the tower. What about adding a second tower into the path. Have a "Y" at the top and go to two shower heads, that would definately cool the water better.
Don't shoot me if the following has been mentioned before:
Have you tried dropping the power of the peltiers down to about 85% (not sure if it's possible depending on your power supply). One guy tried something like this with a 172W... he found the best results were between 85%-90% power... anything above that didn't change the CPU temps but increased the coolant temps quite a bit.
Another thing... I picked up a 1/2"-thick copper cold plate from Danger Den and actually improved the cooling versus my 1/4" cold plate. I can't remember how think you said you were using the standard peltier Mase2 or not.
Btw, don't think I'm talking to you like you're a newbie or anything... I'm just throwing out ideas. Something tends to get lost in the text a lot of times.
-=mac=-
William
05-24-01, 10:43 PM
this has become a brainstorming post, don't be afraid to post ideas, even if you think they are dumb. Maybe the whole thing can't be used, but part of it could be invaluable when solving a problem.
All ideas are welcome. That’s what this forum is about. Despite my best intentions, engineering background and all precautions, this has turned into a real conundrum. The only thing I ask is that you read the thread to find out what has transpired.
I have tried dropping the voltage and the temps rise. The PSU I have has front panel controls and meters for both voltage and current . Very handy. I have not tried a thicker cold plate. Since one of the pelts was suspect before this experiment, I will try two fresh pelts. If I don’t get better results or a revelation, it’s time to punt and there is still the issue of how to deal with so much heat.
Back to cavitation again.
Is your water block silent? any hissing noise will indicate cavitation. Cavitation does'nt alway produce massive vibration, we are not talking outboard motor here. You could be producing alot of very small bubbles effectivly making a foam in the block.
I just thought of something. Do you remember the article on the front page a while back called something like "Near Vapochill performance for $150"? I looked around for it but I couldn't find it. Anyway, someone found a "water chiller" and used it as a reservoir. It's basicly the same principle as a mini-fridge exept it's made to only hold a liquid. I think the guy had a TBird of some sort, but with normal watercooling his water temps were around 0C. He also mentiontioned just placing an aluminum reservior in the cold liquid in order to keep from having antifreeze and such in the cooler. He also said something about being able to rent one of these, though I'm not sure where you'd find one.
Oh yea, what kind of additives are you using in your water? I read through everything but don't remember that issue specificly. I know AutoZone sells stuff you add to a car's water coolant to keep the engine from overheating in high-temp enviroments (basicly lie water-wetter, but possibly better). It cuts down the surface tention in the water and allows it to transfer better in both the radiator and waterblock (might help flow-rate as well). The only major problem I see with this is that you wouldn't want to use a cooing tower with this stuff since it would be disolving tin the air... plus it's not as cheap as straight water.
Just ideas...
-=mac=-
If your water/additive ratio is wonky cooling is affected. Also howmany radiator/coolin towers are you using I would recomend having at least two.
Slain (May 25, 2001 12:20 p.m.):
Back to cavitation again.
Is your water block silent? any hissing noise will indicate cavitation. Cavitation does'nt alway produce massive vibration, we are not talking outboard motor here. You could be producing alot of very small bubbles effectivly making a foam in the block.
Yeah it's silent. I spent some time with a mechanic friend who is into computers today. Apparently diesels can suffer from microscopic bubbles due to the vibration. The same thing could be happening inside the Maze 2.2. If you look at the pic on the first page, you can see it's well damped, so it may not be possible to detect miniscule vibrations with a calibrated finger.
BTW, the 1800 GPH pump went back to the Koi shop today.
Regarding cooling towers, I have just one. The water loss is a quart every 45 minutes. It's working hard.
No additives, just straight distilled water.
William
05-28-01, 01:13 AM
have you tried adding something to it, water wetter or a little bit of antifreeze? That might help a little.
as far as using different types of water weter i would suggest that you only use the redline stuff. i have tried everything that kragen,autozone, and summit racing sells and after many hours of testing (in my pro street nova) I found that redline water weter (about 3 bottles in my car) worked better than any amount of the other stuff. so just dont get fooled with the other stuff, we also used only redline ww at the race company i worked for, they built 24hr LeMans cars, so I think they too have done adequit testing.
Hope this may help anyone who is thinking about or is using water weter.
Tim
Just so you all don't think I quit, the system is back to a 1333 Bird and a Maze 2 for now. Until the new pelts show up, there is not much else to do. I is cool to see the Bird running at or below ambient with the cooling tower and no pelts.
Hope the new pelts do the trick, or we have just proved that the maze 2.2 is plain crap.
inertia (May 24, 2001 02:05 p.m.):
What did you pay for the thing? I know you said an arm and a leg, but . . .
Why is the coolant temp sooo gosh darned high? I've been lurking for a very long time and I've never seen anything that high. Let's hear some further details about your specific cooling tower - how you built it, fan or no fan, the show head - or the radiator you've been using.
for 4cmx4cmx1.5cm = AUS$200=/=US$100.
GOD...
I ordered a maze2-2 a while ago (the 13th) and they accidently sent me a normal maze2 instead. Perhaps I should tell them that I'll keep what they sent and give me the $15 difference back...
BTW- Here's an idea I was playing with, (it's probably retarded and wrong), but I figure it this way- Your trying to handle the output of your pelts and your cpu with that one lonesome waterblock. So, lets say you put a waterblock on the cpu, and two pelts on top of the waterblock, then the maze2 or whatever waterblock on top. Perhaps like two 120w's or maybe just one 170w. Just a thought.
As far as the flow-rate being too fast goes, that's totally bunk, EXCEPT for that bubbles theory. The faster the water is in and out of that block, the less heat the individual particles of water will absorb, thus keeping thermal efficiency high. If there was no thermal efficiency loss when water temperature rose, then flow-rate wouldn't matter, but if your water rises say, 5'C while running through that block, the thermal efficiency is dropping the whole way. If you can make the water run through faster, new cool water will be in before the old water gets warm enough to affect anything that much.
Regarding the above post, I'd like to expand on that idea. What if the stack was setup as CPU-Peltier-Waterblock-Peltier-Waterblock. That way, you'd have 2x the amount of water at any point in time ready to absorb heat, and each waterblock would be respectively absorbing less heat than the single one had been before, thus reducing thermal efficiency loss. I'm not entirely sure how you'd set this up as far as the flow is concerned, but it's something I'd consider trying if I had two peltiers to play with :)
Damn, just looked at the maze 2-2, you won't be able to stack waterblocks cuz of how the inlets are...if it would fit, it might be worth buying one designed for slot coolers, cuz you could stack on top of that.
Slain (May 30, 2001 03:08 a.m.):
Hope the new pelts do the trick, or we have just proved that the maze 2.2 is plain crap.
The Maze 2.2 is definitely not crap. Any block that can deal with over 400 watts of heat and consistently be with in 3C of the water temp is doing quite well. I have had a 156 watt pelt cool better with higher water temps than with the dual 172 pelts. The fickled finger of fate is pointing to the pelts. That’s where there is a low Delta T.
Colin,
For what it's worth I agree with you. The maze block appears to be doing a good job of taking the heat from the pelts and tranferring it to the water. Also your water cooling tower is working pretty good, 35 C is a fair temp considering the heat going into the system. What really does not appear to be working is the generation of a Delta T between the block and the cpu. Duff pelts or poor thermal contact between the cpu/coldplate/pelt would appear to be the issue.
Did you ever get the chance to take a reading of the coldplate temp? because if the coldplate was showing say minus 20C and the cpu is still around plus 20 C it could simply be poor seating of the coldplate on the cpu.
Somebody mentioned the cold plate a while back and it seems like a good possibility to me.
It kind of seems like the setup is not working any better then what you would get from a single 172w tec. I wonder if the outside edges are contributing much at all to the cooling as the heat from the core has to travel quite a bit to the sides to take advantage of the large surface area that the tecs are cooling. It just could be that only a small square in the center right above the core is doing all the work and the rest just acting as a water heater. A thick cold plate maybe around 1/2" might be needed.
The block was seated on the CPU nice and firm. There is an impression of the code from the die on the block. The whole assembly is potted in neoprene and silicone and I have not bothered to take it apart yet. I torqued the cold plate down and then pulled it apart to check the pelts contact area before sealing it up. It was fine.
I never did get around to measuring the temp at the cold plate. I should have done that before I switched back to the Bird and Maze 2. I guess I was getting impatient. Loafing along at 1gHz does seem slow when have been spoiled at 1.5.
This has probably been asked, but have you tried the setup with just one peltier? The extra peltier may just be adding more heat to the system than the cooling tower can handle, and thus decreasing efficiency. It's not an endless cycle. Stacking 8 peltiers on top of each other won't do you any good, unless you can take the heat of 8 peltiers out of the water just as efficiently as you could take the heat of one. Anyway, if you haven't, try it with just one of those 172w peltiers.
I tried one 172 watt pelt with a Maze 2. I got better results with a 156 watt pelt. This is what led to the Maze 2.2 and dual pelts. The Delta-T across the pelts won't be any higher but the load capacity should be higher. BTW, that 172 watt pelt that did not perform well is one of the ones I used in this test.
Colin (May 30, 2001 09:19 p.m.):
I tried one 172 watt pelt with a Maze 2. I got better results with a 156 watt pelt. This is what led to the Maze 2.2 and dual pelts. The Delta-T across the pelts won't be any higher but the load capacity should be higher. BTW, that 172 watt pelt that did not perform well is one of the ones I used in this test.
Is the 172watt pelt really a 172watt animal or a 120watt animal?
I bought a 24v Drift-0.8 172watt Peltier.However it's volts and amps charactoristics(4 amp at 12volt, 6amp at 16.75volt) are more akin to those of a 24v Drift-1.15 120watt Peltier,as shown by the Kryotherm program( http://www.kryotherm.spb.ru/soft.htm ).
The only difference in appearance,that I know, is that the 172watt is 3.2mm thick whereas the 120watt is 3.4mm.Without a pair of micrometers I am unable to substantiate my suspicions.I wonder if your volts/amps are similar?
I should be able to tell you real soon. I got word from Danger Den last Thursday that two fresh 172 and two 120 watt pelts were ready to ship. Apparently the first set of replacement 172 watt pelts were lost in the mail. I will measure all of them with a pair of dial calipers.
One more thing. The 172 watt and 120 watt pelts have the same Delta T and will probably perform the same since the heat load from the CPU is under half the combined wattage of the two pelts. Since the cooling system will have 100 watts less heat to dispose of with the120 watt pelts, they may actually produce better results.
wow... i have been following this thread... and it is way to confusing for me... where can i read up on all of this peltier stuff you guys are talking about? what is deltaT???
Hey it confused me for a while too! :D Delta T is the difference in temperature across a thermal interface. You can read more about pelts here, (http://www.heatsink-guide.com/applying-peltiers/)
Thank you Colin for an excellent thread and your response.
I will be most interested in knowing the results of tests on the new modules,particularly so since my 172watt Peltier was also purchased from Dangerden.
As far as perfomance of the two is concerned Kryotherms program predicts that using a good waterblock(K/W=0.15) running at 19v(parallel) 2x120watt will be better than 2x172watt at loads upto about 40watt.However at higher loads the 172watt will come out on top.The actual predictions for a 100watt cpu with a good waterblock(Rh=K/W=0.15),nearly perfect cpu thermal contact and heat spreading(Rc=K/W=0.001),and average insulation(Rins=K/W=2.0) are 5.7C for a 120watt and 2.2C for a 172watt.
Hoping to hear more from you ,
Les
Les,
THANK YOU for the Kryotherm program, the numbers and encouragement. I ran a search for Kryothrem and just started playing with the program. Way cool! The only variable I can see to the 172s vs the 120s is disposing of all the heat. For the time being I will be limited to a bong cooler. I do have a line on a couple of surplus bottled water coolers but this may take some time.
Colin
Kryotherm- the numbers I quoted are obtaied using the KryoTESC portion of the program."Select module","High Effect Single Stage TEM", select the peltier,"Return" then feed in numbers to one heart's content. There is a goldmine of information.
ken257 (May 17, 2001 12:58 p.m.):
I am curious to see the results of the cooling tower as I have been kicking around the same idea of dual 172's/DD Maze 2-2. I have all the parts here but the only holdup is I have been thinking of a way to get the heat out of the water. It is very humid where I am in the summer so I am not sure if evaporative cooling is going to hold up, at least with swamp coolers they seem to work better in dry areas. My chilled water setup is out of the question it is only good for about 100w or maybe a little more of heat.
How can the water be minus seventeen degrees??? wouldn't it be frozen???
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.