Quick idea with TEC.

[Boulard83]

I already have a loop with a TFC360, GTZ, Komodo, 2xmcp655 and an MCres.

Cooling an I7 @ 4ghz and an OCed HD5870 (950/1300).

My temp are already fine and my delta water/air is ~7°c after several hour of gaming.


Here is my TEC idea !

Using this :
• Noctua NH-C12P
• EK Supreme LT
• Asetek VapoChill Compound
• 245W Potted Peltier

My idea is to mound the Noctuah upside down in the bottom of my case ( or another cooler if you think this one cant do it ! ). Placing the TEC hot side on the Noctua, and the cold side on the EK block.

Loop order :
Res - 2x MCP655 - TFC360 - EK/TEC - CPU - GPU - Res

My goal isnt subzero temp but its to reach just something under my room temp

I also just want to try it .. play with it !

Edit : just remembered that i have a brend new AC accellero 8800GTX laying around. Spec capable of 240W dissipation. I could fit the HX block ok the accellero

[zipdogso]

My goal isnt subzero temp but its to reach just something under my room temp

I also just want to try it .. play with it !

So....

From the sound of this you will do it regardless of what anyone says, so why ask ?

Do it....play with it !

[Boulard83]

yeah, ill do it.

But if anyone can tell me anything that can help, ill take it

I know that i cant really reach subambiant cause my radiator is most likely going to ReHeat the water but by placing it in front of the CPU the 2 blocks gona heat it before the rad. If i can have 1-2c under ambiant at the block

[Ultrasonic2]

It's pretty much isn't going to do much

Whats this gonna cost.

Why not spend some real money and buy a TEC water block that actually does something

it will be interesting any way

[NKrader]

i had this same idea.. but figured out to be way too much money for the bang.. :/ also i dont have the money.. lol..

Good Luck

[Boulard83]

Ill try it and post my result.

I already have the cooling needed ( Accellero 8800gtx capable of 240w of cooling ). I just need to order a TEC + coldplate + Block.

ill be back with result as soon as possible.

[okydoke]

Let me know how you go - I attempted this with an older 120mm Thermalright Cooler (not a true, can't quite remember now), and the cooler would be able to handle the hot side, but only for a couple of minutes. With that said, my mounting was not ideal, so if you get some results, I may have to reevaluate

[Boulard83]

My order is shipped. Waiting for it and ill be back with my testing result.

[Mescalamba]

TEC generates 2x ammount of heat taken.. you would need really awesome water loop to cool that.. And even than, not worth it.. just pure water would be better, or get phase instead.

[Ultrasonic2]

TEC generates 2x ammount of heat taken.. you would need really awesome water loop to cool that.. And even than, not worth it.. just pure water would be better, or get phase instead.

This is such a load of rubbish . and can't be backed up by any facts

[flak-spammer]

Indeed, in the future could people please not post their opinion unless asked. There are several cases where TECs can easily double the efficiency of a phase system, albeit it's not cheap but it's doable. Pure water cannot possible get below ambient and therefore cannot yield at best better temperatures.....facts people facts.

[zipdogso]

TEC generates 2x ammount of heat taken.. .

No problems there.....well not much. The heat output from the hotside (take from Manufacturers graphs.) is roughly twice the cooling. Not all of this heat is generated by the TEC. A portion of it just passes through from the heat source.

EDITED - After looking at my notes I realized this is a little off - the ballpark figure is actually roughly 2.5 times the cooling as Ultrasonic reminded me it is a rough ballpark calc without the need to get accurate.

you would need really awesome water loop to cool that...

But remember it is just the i7 and a GPU. Granted though...they are both OC'd.
The Jury is out regarding the output of an i7 but recently it has been shown that it MIGHT not be as much as one thinks especially at normal usage of non-multithreaded programs and someone recently calculated the STOCK heat output of a 5870 GPU (not the whole card !) as 60w.

The possible problem would occur as the TEC is a 245w, assuming he is going to run this at 12v he might find the Q is only in the region of 160-170w. So I would hanker it is not quite big enough ( Sorry I didn't do the calcs earlier. Didn't really read the OP properly.)

Whatever.....the heat output is unlikely to be more than 350-400w an average loop can do it easily.

EDITED - Oh shoot......I really must read posts properly....he is cooling the hotside with a noctua...I think your gonna have a problem Bud...

And even than, not worth it.. just pure water would be better, or get phase instead.

Flakspammer has answered that....nuff said.


Just for my pennies worth....
That noctua will raise your case temp into the stratosphere which in turn will hurt your TEC cooling and everything else.....I know people hate things outside the case but I think you might need to consider it....

Indeed, in the future could people please not post their opinion unless asked.

OUCH....now that's not in the spirit of a forum is it.

When do people request the opinion of a specific person ??????
Come on now....get real. I wouldn't be surprised if some extra touchy individual goes and reports that post.

[flak-spammer]

When do people request the opinion of a specific person ??????
Come on now....get real. I wouldn't be surprised if some extra touchy individual goes and reports that post.

Okay I'll agree that came out very poorly, what I meant is that people should not try and pass their opinion as fact. It's quite easy to get caught up in what we have experienced.

[yngndrw]

There are several cases where TECs can easily double the efficiency of a phase system

Can you post some more information about this, how to and so on. Specifically, concise information about it as everything I usually read about TECs are massive pages of information - More that is required to get the point across.

[Ultrasonic2]

oh zipdogso .. stop playing the devils advocate. WE both (including flak) know you don't run TEC's at their VMAX

[Ultrasonic2]

Can you post some more information about this, how to and so on. Specifically, concise information about it as everything I usually read about TECs are massive pages of information - More that is required to get the point across.

This should help you

[yngndrw]

Thank you that was a nice guide.

When hunting for C.O.P., is it best to go for larger TECs or smaller TECs ? (Having lots, under-volted of course.) Also is your Watts Vs Input Voltage graph always going to be the same shape (And have the same proportions.) for other sized TECs ?

Sorry for it being a bit off-topic, but I think the OP would also be interested in the information.

[Ultrasonic2]

Thank you that was a nice guide.

When hunting for C.O.P., is it best to go for larger TECs or smaller TECs ? (Having lots, under-volted of course.) Also is your Watts Vs Input Voltage graph always going to be the same shape (And have the same proportions.) for other sized TECs ?

Sorry for it being a bit off-topic, but I think the OP would also be interested in the information.

When talking about high COP you want a very large Qmax it doesn't theoretically matter if it's from one massive TEC or lots of small ones as long as the surface area and Qmax is the same.

However a lager-er surface area is always more desirable because it's easier to cool and therefore you get a better delta or more COP.

As for the graph changing depending on TEC used. TEC's are pretty much all the same these days. But it is just a simulation. For your info the graph does change depending on the load your cooling relative to blah blah . as the load increases the TEC's will uses a bit more power.

[yngndrw]

Thank you.

[zipdogso]

oh zipdogso .. stop playing the devils advocate. WE both (including flak) know you don't run TEC's at their VMAX

I wasn't suggesting he did...i said 12v...Qmax at the operating point of 12v allowing a fair Dt will be about 160-170w...He has an OC'd 5870 when a STOCK one is 60w...doesn't leave much for an OC'd i7 but results recently suggest it might be borderline enough.

times that by 2 that is a heat output of 340w possibly a bit more so i said 350-400w.
Even at 340w I think the noctua will struggle as the TEC is possibly underpowered (both i7 and 5870 OC'd.) and without good cooling on the hotside, cold side temps will suffer as you well know...

devil's advocate....not really...there is a POTENTIAL problem a.) with his TEC size and b.) with his hot side cooling especially if he plans to keep it all in the case.

[zipdogso]

Thank you that was a nice guide.

When hunting for C.O.P., is it best to go for larger TECs or smaller TECs ? (Having lots, under-volted of course.) Also is your Watts Vs Input Voltage graph always going to be the same shape (And have the same proportions.) for other sized TECs ?

Sorry for it being a bit off-topic, but I think the OP would also be interested in the information.

To be honest to go for COP you best with multiple TECs undervolted
Whether they are big or small really depends on how easy they are to power but as high current PSU's are expensive if you can find TEC's with a high volt/low current relationship it makes life easier...these are generally those with higher numbers of thermocouples...certainly not 127's.
On the COP graph from full power to 50% the COP rises slowly from around 0.5 to 1 maybe a bit more it then rises relatively steeply to about 20% power and then drops like a stone.
The big problem is Dt....the higher the Dt the very much less of a peak you get with a Dt of 20ºC that peak becomes a mere bump. At 20ºC Dt you don't get good cooling unless your hotside is practically frozen !!!
Yes....TEC's can get a better COP then Phase but due to the small Dt a Phase system at the same COP would yield much lower temps.


I don't actually look at Q-input voltage charts because all the manufacturers actually use Q-input current or compound charts with Q-input current and voltage. But if Ultrasonics graph is right (I am 100% sure it is.) then yes it should be applicable to all TECs for ballpark figures anyway.

[yngndrw]

Thanks, a Dt of about 20C would give fridge-like temperatures (From an ambient hot-side, which is unlikely.) - Ideal for a chiller if you didn't want to use anti-freeze or other additives like that. Having said that, for cooling computers I'm not sure it's really worth the trouble unless you were to stack them in parallel stacks. (Which of course then makes finding the sizing and input power "sweet spot" all the more harder.)

[zipdogso]

The problem with stacking (What i call stacking ! I don't think it is quite what you mean.) is it only increases the Dtmax so a 2 layer stack will get you in the region of 80-85ºC Dtmax this is degraded by your input power, the load on the coldside etc so the gains are not phenomenal it really works on the tiny TECs e.g Ferrotec has some 3 stage with a Dtmax of 111ºC but I am sure you don't need telling about the difficulties of trying to do that with TEcs of a reasonable worthwhile size Ferrotecs 3 stage have a Qmax varying from 9.7 - 22w !

So far my little experiments have gone, they have just highlighted the necessity for good hotside cooling.
The easy way to cool the hotside especially from the PC angle is to use a rad cooled water loop but as far as TECs go that is not good hotside cooling - you will never get below ambient and even an impressive 50ºC Dt will only get you headed on for -20ºC ( your loop will be warmer than ambient and the hotside warmer still.)
It was this problem that spawned that stupid idea some years ago of using the the same cooled coolant to cool the hotside on the same TEC (a circular loop.) of course it doesn't work because TECs aren't coolers they are merely heatpumps.
It would seem from my research that TEC's will work down to a hotside temp of -50ºc but it begs the question that if you have got to super chill the hotside would'nt it be better just to go phase. Although I aam trying to find a way round it 'cause phase is an arm and a leg over here in UK, if , and it's a big if, you can find a builder.

[yngndrw]

If I get the time I will try and make an efficient 50Watt chiller with a decent Dt. (By stacking with differently powered TECs and spreading the load out.) If it works then it would just be a case of multiplying it by however many are needed for a more realistic heat load. I believe that if you oversize / undervolt each stage enough that it is feasible, although for computer loads I'm not sure it would be even slightly cost-effective.

[Ultrasonic2]

i hope you've got a pile of cash for this because it can't be done on the cheap.