TEC assisted CPU waterblock concept

[Scarlet Infidel]

I think we are talking about different things here. Lets start with a magic system that cools the hotside to a constant 30c regardless of load. To get a COP of 2 you are limited to a dT of around 15c. This means with no thermal resistance (not possible) the cold side loop would be 15c (ie probably just above dew point for me). these are the kind of numbers I was playing with in my system.

Now lets say you want to reach zero. Now you need a dT of 30c (actually more because of the extra heat form the pelts but we will assume you can compensate). Now you get a COP of around 0.75. In reality with the added heat load you would probably only get 10c cooler for 2-3 times the power. It just doesnt seem worth it to me.

[zipdogso]

I think we are talking about different things here. Lets start with a magic system that cools the hotside to a constant 30c regardless of load. To get a COP of 2 you are limited to a dT of around 15c. This means with no thermal resistance (not possible) the cold side loop would be 15c (ie probably just above dew point for me). these are the kind of numbers I was playing with in my system.

Now lets say you want to reach zero. Now you need a dT of 30c (actually more because of the extra heat form the pelts but we will assume you can compensate). Now you get a COP of around 0.75. In reality with the added heat load you would probably only get 10c cooler for 2-3 times the power. It just doesnt seem worth it to me.

Yes I see what your saying and it is perfectly true for a single TEC.

That is not what I was talking about...I was talking about multiple undervolted TECs. Each one carries a portion of the heat load, the hotside temp is a lot lower (so long as you give it good cooling.), the delta max is lower, the delta temp is lower. The general theory is you need a good number of TEC's dependent on your heatload...certainly you need in excess of 4. It certainly helps if your general ambient is not too high. On reflection perhaps Subzero is not really realistic but quite a way below dew point (single figures.) is easily possible...you would need a chiled ambient for subzero.

OK i was going a bit too far with the hotside of -15ºC but I only quoted that to show what happens. It shows you just lose a little bit of COP..a litttle bit..not like 1.5.

This is the theory Uncle jimbo was bandying about.

[Scarlet Infidel]

My example was for a single TEC handling 20w or so, for example 10 would be needed for 200w.

My chiller has 10 TECs in it, I'd still prefer more. I still think you lose a lot of COP for high dTs. Could you give an example to show what you mean (mention type and number of TECs, voltages, heatload etc)

[zipdogso]

My example was for a single TEC handling 20w or so, for example 10 would be needed for 200w.

My chiller has 10 TECs in it, I'd still prefer more. I still think you lose a lot of COP for high dTs. Could you give an example to show what you mean (mention type and number of TECs, voltages, heatload etc)

Well I can give an example but it's not actually a working at the moment...I am waiting on some temp sensors then a bit of wiring and soldering and finally some copper parts which I have been forced to make myself, cause I cant find anyone to do it for me !!
My TECs are 26316's (31.6v - 16 amps) with a qcmax of in excess of 300w at 27ºC (The manufacturers words NOT mine...couldn't even give me a proper performance chart !!!)
Well the nearest thing on ferrotec is here:- http://www.ferrotec.co.uk/index.php?...ail&mod_id=118
UJ said the figures need adjusting by 263/242 = 1.0867 but i am happy to take the 8v figures but actually run at 7v or just over - shouldn't be far wrong.
After discussion with Uncle Jimbo he reckoned with watercooling a good ballpark figure would be hotside = ambient +10ºC and dT 15-20ºC with the dT being conservative.
The supply I'm using is variable 4-16v with a 22amp max.
Average room ambient at the mo is about 22º but it is dead easy to get it lower so for this take 22º+10 =32º hotside.
So adjust the ferrotec calc hotside to 32º

8v gives about 4amps for pump of 70w and 90-100w rejected (bare in mind 100w can be cooled quite well with quality heatsink /fan arrangement.) and a COP of 1.5 with a coldside just over 10º reducing the ambient a bit could swing it to single figures.
I only 4 ( 2series/2 parallel - 7v 4amps each.) of these so total pump = just under 300w for just over 110w in. I could use more to improve heatpump and reduce input power by another volt or so.

In my final project of which this is a small part the ambient/hotsides and dT will be lower and obviously the COP higher.

[Scarlet Infidel]

Im impressed by those numbers and they all look right to me. The only TECs I can get hold of dont have anywhere near as good properties as that. With a COP of 1.5 and dT of 15-20 my units can only pump about 10w so even though I have 10 of them this isnt going to be very useful.

Of course it is important to bear in mind that the chiller I was building was designed for acceptable performance at very low cost (not much more than a straight watercooling system). I have clearly lost sight of what is achievable with higher end parts. Where did you get them from? How much were they and how many do you have?

[zipdogso]

Im impressed by those numbers and they all look right to me. The only TECs I can get hold of dont have anywhere near as good properties as that. With a COP of 1.5 and dT of 15-20 my units can only pump about 10w so even though I have 10 of them this isnt going to be very useful.

Of course it is important to bear in mind that the chiller I was building was designed for acceptable performance at very low cost (not much more than a straight watercooling system). I have clearly lost sight of what is achievable with higher end parts. Where did you get them from? How much were they and how many do you have?

I got them from Hebei I.T.
http://www.hebeiltd.com.cn/?p=z.peltier.pricelist
They are the last ones at the bottom of the list....
In my view they are unfortunately about the only ones worth buying.

Luckily...or not as the case may be ....They don't have a distributer in UK so I could place an order for just $99 (the normal limit is $250.) , so I ordered 4 there was a premium for a small order so in the end I paid $107 and $33 shipping, they came in fast only about 5 days...then 2-3 weeks later I got an invoice from the shippers for another GBP 15 for import/VAT !!! This was last year when the rate was a bit better so the 4 ended up costing about GBP 95. More than I initially hoped but far better than Kryotherma prices !!! I'd like to get some more but the rates not so favourable now...need a bit more available cash.

[Scarlet Infidel]

They are very impressive. My cheap TECs were £17.50 for 10 including postage etc. Bit of a bargain but now about double that I think (I was thinking about getting another ten).

[ilkkahy]

but keep in mind .. watercool isnt bout 1 pass of water .. it requires .. lotsa loops

If we assume you have flow rate of 200 Liters per hour:

200/3600=0,055L/sec

water weights about 1kg per liter so and it has thermal capacity of 4,19kj/kg*k so...

0,055*4190=232W/K

So you need 232 watts of power to warm up the water one kelvin. Water doesnt get _that_ hot after all does it?

[demonkevy666]

If we assume you have flow rate of 200 Liters per hour:

200/3600=0,055L/sec

water weights about 1kg per liter so and it has thermal capacity of 4,19kj/kg*k so...

0,055*4190=232W/K

So you need 232 watts of power to warm up the water one kelvin. Water doesnt get _that_ hot after all does it?

heh I was just looking for something like this today, but for an Air heat sink
what's the formula for all this that's what I'd like to know
it's like you need to an actuary to figure these tec's out completely

[zipdogso]

heh I was just looking for something like this today, but for an Air heat sink
what's the formula for all this that's what I'd like to know
it's like you need to an actuary to figure these tec's out completely

What are asking here?

The calculation you highlighted is merely a calculation for the power required to heat or cool water by 1 degree.

It has nothing specifically to do with TEC's or air heat sinks.

[ilkkahy]

Those are very basic and easy calculations of heat removal, which i think people should do at the beginning to get the idea whats possible and whats not.

For air heatsink:

Air weights 1,29kg/m^3, it has thermal capacity of 1kj/kg*K

If we have airflow of 100m^3/h trough the heatsink... = 100/3600=0,0277m^3/s

Air weights 1,29kg/m^3 so 1,29*0,0277=0,0358kg/s =mass flow

0,0358*1000=35W/K

So you need 35 watts to heat air for 1K. With 350 watts heatload the air would heat 10K while passing heatsink.

[flak-spammer]

What does this have to do with what is being discussed here. We're just trying to make a TEC waterblock not an air heatsink. I realize the same principles follow but it's not particularly relevant.

[ilkkahy]

Well someone said that certain waterblock model requires lots of loops due high heat output of the system (i assume?). I think that its bit wrong unless you have very very low flowrate. Also demonkevy wanted to know how it goes with air heatsink so i answered to him.

Anyway..

I have thought about this tec assisted cpu waterblock sometime myself too. The idea is not to try to move all cpu heatload with the peltier. It would be possible to get slightly lower temps than with just a normal waterblock but i dont think that its eventually worth the wasted electricity and complexity of the system. Its intresting concept though.

[ilkkahy]

I made quickly picture of the version i have thought about testing if i build water cooling rig some day.



I would propably use Thermaltake 120 copper version as heatsink and Drift-08 peltier at 12v as i happen have to some of those at my junkpiles.

So the peltier+heatsink would move heat from the waterblock and the water. The trick would be to optimize the ratio which to cool more (water or the waterblock/cpu). It might seem stupid at first but having thought this more i think there is kinda fun/intresting possibility to optimizing and tinkering with this kind of "hybrid" system.

[Ultrasonic2]

i've thought of these too. it would be interesting to see how much of a difference they could make ... i hink construction is the real issue you need to make it out of a single block of copper. NO covers

[zipdogso]

I made quickly picture of the version i have thought about testing if i build water cooling rig some day.

just for my 2 cents....ilkkahy..I think your on the nod.

I have been working on a similar idea for TEC cooling. This is the principal I have been thinking about for the last 7-8 months.

I have, for a long time, felt water chiller designers were trying to put too much volume past a TEC in too few passes in many cases only one pass. I think there is also some merit in guiding the coolant past in a tube/slot/hole mainly because I think it negates many of the laminer flow problems. There are some issues like oval tubing will be better than round etc but generally I think it's a better idea than putting a large flow past a copper plate with slots in and basically hoping it works.
True, as one person said sometime ago, you only cool less than a "coke" can full of coolant at a time but at least it is being cooled instead of shimmying past the slots on the easiest path and avoiding most of the cooling.

Commercial industrial TEC water chillers rely on multiple passes on the TECs and irrespective of high/low flow pumps I think this is the best way froward.

There have a couple of designs on here recently "ultrasonic2" is the best to date and actually finished but he also had laminer flow problems. "coity" and "snake" have been on here recently but have "gone cold" for one reason or another and I too am working on the principal.

Unfortunately I have spent several months hunting for a "friendly" engineer, metalworker to help me but despite offering good money it appears there are no "friendly" engineers in the UK...I wouldn't be surprised if they are all best buddies with the HVAC engineers in the UK...that's why there aren't many "friendly" ones of those either !!!

Anyhow I have managed to get the engineer inside of me to get his overalls on and despite not having done any metalwork/engineering of any sort (I am a printer/machine op) for over 20 years I am hoping to get things moving shortly. Just hoping I am not biting off more than I can chew.

[Ultrasonic2]

There have a couple of designs on here recently "ultrasonic2" is the best to date and actually finished but he also had laminer flow problems. "coity" and "snake" have been on here recently but have "gone cold" for one reason or another and I too am working on the principal.

LOLOLOLOLOLOLOLOLOLOLOLOL

Actually it's not finished but it is working. It never will be "Finished". i really want to make another one now with what i've learnt

It is my most successful TEC Project so far though

[zipdogso]

LOLOLOLOLOLOLOLOLOLOLOLOL

Actually it's not finished but it is working. It never will be "Finished". i really want to make another one now with what i've learnt

It is my most successful TEC Project so far though

Yeah Ok...LOL it was the certainly the MOST finished.... and pretty good results by all accounts.

[Ultrasonic2]

I made quickly picture of the version i have thought about testing if i build water cooling rig some day.



I would propably use Thermaltake 120 copper version as heatsink and Drift-08 peltier at 12v as i happen have to some of those at my junkpiles.

So the peltier+heatsink would move heat from the waterblock and the water. The trick would be to optimize the ratio which to cool more (water or the waterblock/cpu). It might seem stupid at first but having thought this more i think there is kinda fun/intresting possibility to optimizing and tinkering with this kind of "hybrid" system.

A block like this will never perform that great or cool better than a decent chiller. but it should reduce CPU temps a little.

One of the issues with a chiller is that it cools the water not the CPU this means the water has to go through a Water block . This water block as a C/W and therefore CPU temps end up being much higher than water temps.


A design like yours reduces this issue as it's cooling the same block of copper that is touching the CPU

The problem is that it can't go much below ambient because then the water flowing through the block will take the cold water away to be heated back up to ambient.

So the idea is limited in it's maximum performance. But should be able to reduce your CPU a few C

[Ultrasonic2]

Yeah Ok...LOL it was the certainly the MOST finished.... and pretty good results by all accounts.

Thanks man. i should be happy with it but im not. I've run out of money for now

These are hot and cold side temps (c) after running 100% for some time. Ambients are 21c (Q6600 @3600)

[zipdogso]

A block like this will never perform that great or cool better than a decent chiller. but it should reduce CPU temps a little.

One of the issues with a chiller is that it cools the water not the CPU this means the water has to go through a Water block . This water block as a C/W and therefore CPU temps end up being much higher than water temps.


A design like yours reduces this issue as it's cooling the same block of copper that is touching the CPU

The problem is that it can't go much below ambient because then the water flowing through the block will take the cold water away to be heated back up to ambient.

So the idea is limited in it's maximum performance. But should be able to reduce your CPU a few C

I have pretty abandoned the idea of direct die cooling but I think this principal will work well in a TEC water chiller...and as TDP's go ever higher chillers might be the only option soon.

[Ultrasonic2]

I have pretty abandoned the idea of direct die cooling but I think this principal will work well in a TEC water chiller...and as TDP's go ever higher chillers might be the only option soon.

i actually think chillers have been the only option for quads for sometime now.

So what are you working on then ?