Peltiers on GTX 480 SLI and Core i7

[xkon]

So I've read up a little bit on the info available here, but decided to start a thread of my own anyhow. One day I hope you can all point to this page and say, "That's how you DON'T do it!" I'm planning a build and wanted to add TEC into the mix as they've always interested me, but I never had the funds to play with them. All parts are selected except for the PSU and some of the cooling stuff. The computer will start with these specs:

SX58H7 Mobo
i7 920 C0
6GB DDR3
2x GTX480's /w FC blocks
3x OCZ Vertex or 1x Intel X25M SSD

Options for PSU:
Antec Signature SG-850
Thermaltake Toughpower 1200w
(looking into Meanwell PSUs, but don't know how much power i need)
Add a PCP&C 750 silencer?

What are the recommended peltiers for the CPU and GPUs?
I was thinking about using 2x ~320w for the GPUs and a ~220w (Fermi is 250w TDP and i7 920 is 130w right?) Or does TDP have nothing to do with watt temps? Any suggestions?

Heat will be dissipated by 2x 240 rads. Will add 2x 360's or a 480 soon (hopefully before I figure this out and buy the peltiers). Will this be sufficient cooling?

What kind of electricity bill should I be expecting? Sounds like a durr to me, but I'll ask anyways, can I run the system without the peltiers on without removing them (keep them sandwiched between the processors and waterblocks).

Should I have the fire department on speeddial? Will the city come a calling when I cause a brown out? Will I travel back in time when I hit the on switch?

Oh yah. I suck at modding. I have few to no tools. So the option with the least tools is a win for me.

[duron]

...I guess there's something about Pelts that made them notorious that's why they did not flourish in the PC world
but you could give it a shot
will be watching this

[xkon]

...I guess there's something about Pelts that made them notorious that's why they did not flourish in the PC world
but you could give it a shot
will be watching this

Having a lot of second thoughts now... Got other ideas brewin.

Thanks to the Scifikg and his success with the TRUE TEC Module, I'm thinking of the same thing, but with water instead. Say I get 2 copper waterblocks to sandwich a couple peltiers between them, and run them in watercooling setups parallel to each other.
One is comprised of cpu > gpu > mobo > res > coldplate > pump
The other is radiator/res > pump > coldplate.
Would this no longer be considered TEC cooling? Is it now chilled liquid? Do I need to move my thread? I'm not as hardcore as those chilled liquid people. Honestly, their dedication scares me a little.

[[XC] gomeler]

Honestly.. you are better off with either straight water or chilled liquid. TECs throw a handful of monkey wrenches into any plan and it is very difficult to get a clean build going. You don't see much activity in this section because it is sort of a dead technology. I'm all for people pushing boundaries and I'd love to see such a rig running but it just isn't the best way to tackle this problem. A chiller with a PID controller to keep water a few degrees above the dew point is ideal

[zipdogso]

I'm planning a build and wanted to add TEC into the mix as they've always interested me

This can be dangerous thinking....

but I never had the funds to play with them.

Any project involving TECs will always ultimately cost more than you budgeted...

What are the recommended peltiers for the CPU and GPUs?

There aren't really any. It is at the mo' a bit of a case of going with what you think will cover your heatload. Increasingly just knowing the TDP's is not sufficient.

I was thinking about using 2x ~320w for the GPUs and a ~220w (Fermi is 250w TDP and i7 920 is 130w right?) Or does TDP have nothing to do with watt temps? Any suggestions?

I would forget the GPU it is notoriously difficult to get a TEC of sufficient power but small enough to use with GPUs - where exactly do you think you will have enough room to put two ? Or are you thinking of doing a chiller ? that are definately a DIY job.

The figures your quoting for the TECs are probably Qmax (unless they are Ebay ones.) the problem is once powered the actual cooling you get is significantly less...assuming your planning 12v power the 220w could easily be as low as 150w once running. Under normal circumstances (i don't have an i7.) I would say it is possibly not enough if you plan OC or run stress tests however it has been shown recently this is not the expected outcome.
Of course if you dont use multithreaded programs or processor intensive software you probably wont ever use more than 2 cores so this TEC might be sufficient but I don't feel sufficiently happy to stick my neck right out and say it will work.

Should I have the fire department on speeddial? Will the city come a calling when I cause a brown out? Will I travel back in time when I hit the on switch?

Oh yah. I suck at modding. I have few to no tools. So the option with the least tools is a win for me.

Well I will be perfectly honest I don't think TEC is something you sound particularly capable of completing successfully. It will need to be completed successfully to work and I am sure you have gathered it is not quite just a case of poking a TEC between the CPU and a waterblock. No tools isn't helpful particularly as there isn't actually many premade parts for TEc usage, and the fire department on speed dial and the other comments, even if made in jest, leads me to believe you have no confidence in yourself and your not happy messing with electricity.
Not being sure what size meanwell you need for the TECs shows your lacking fairly basic electrical knowledge and I certainly would not recommend using a PC PSU they are not built to run high current devices like TECs. It is possible to use one if you know what your doing otherwise you might possibly need the fire department on the speed dial....

[Ultrasonic2]

Honestly.. you are better off with either straight water or chilled liquid. TECs throw a handful of monkey wrenches into any plan and it is very difficult to get a clean build going. You don't see much activity in this section because it is sort of a dead technology. I'm all for people pushing boundaries and I'd love to see such a rig running but it just isn't the best way to tackle this problem. A chiller with a PID controller to keep water a few degrees above the dew point is ideal

#*#*#*#*#*#*#*#**###***

TEC's are not dead they are actually making a come back. TEC's only stopped being used because people dont understand them and therefore are unable to make something that actually works. and unfortunately the OP falls into this. Using TEC's successfully isn't just a case of getting 2 CPU water blocks and sticking a TEC in between.

A TEC chiller when made CORRECTLY can be very successful and efficient however 99.999999% of people believe that with just a 2$ TEC they can get good results which is not the case. To get the job done right you'd need to spend big money to get a chiller machined for you.

[xkon]

Thanks for all the info. I've been truly humbled by this thread. Those were wisecracks at the expense of other noobs who've come before me and have failed. I know I don't sound very confident in this venture, but I think a lot of that stems from not knowing what I'm really getting into. I've done research and have read up on some of this stuff, but it's a learning process. And one of the best ways to learn is hands on knowledge. I know many of you are experts in this field and are truly knowledgeable. I simply ask for your assistance in helping me find out more if this is something I would like to get into. If I don't have the tools I'll make due or find a way to get my hands on them. Simple hand tools are available, but I was talking more along the lines of CNC and other shop machines that so many of the other extreme members have access to. This will likely rub some people the wrong way, but judging by the abrasive responses on threads in this section, this area seems almost elitist and those who want to venture in are very suddenly and often humiliatingly shot down by the handful of pros. This easily thins the ranks of people who would like to give TEC cooling a shot. I willingly put myself out there to take that beating because I will know more after and one way or another I WILL have an extreme cooled machine whether it be with TEC, Phase, or Chilled liquid.


Ok let's add another factor. I will be submerging the above system into mineral oil. I've come up with some options and would like to know if any are feasable. If none are, feel welcome to chime in.
1. Cooling the oil using a pump and rad /w heatsinks on the CPU and GPU's
2. TEC's on the CPU and GPUs /w heatsinks and cool the oil /w rad
3. Heatsinks on CPU and GPUs /w oil cooled by TECs sandwiched between two copper waterblocks running in a parallel loop.
oil loop= pump > heatsink > waterblock.
water loop= large waterblock > rad/res > pump > waterblock.
4. Heatsinks on CPU and GPUs /w TECs sandwiched between a large copper coldplate and a large waterblock cooled by a waterloop.
I'm thinking maybe situated above the system submerged in oil also, or under it /w mobo mounted on top of the plate.
Not sure if submerging it all in the oil will cancel out the effect of the TEC since the hot and cold side are in the oil. (Maybe insulate the hot side so that heat goes to rad instead of back into oil?)
Perhaps just submerge the copper plate and keep the waterblock above the oil?
5. Scrap all ideas and start fresh. You're suggestions are very welcome.

These type of things always go over budget. I had $300 set aside for my first watercooling setup and ended up spending $500 so I know how these things go.
What TEC's do you guys recommend for this job?
What Power supply do you recommend?
What will be the most efficient way to set this up?

I need to run to work. Thanks again for your input.

-KX
-Life's short, play naked

[Ultrasonic2]

Im sorry if your feeling got at.
This this now this forum goes
A noob comes in there and "tells" what he's gonna do and how wonderful it's gonna be, but is opinion is not base on any fact
The people that do know what their talking about try with every bone in their body to convince the OP that the idea isn't a good one
The OP knowing better than the people that have actually done it before goes ahead with their idea anyway
The performance of the block or what ever is then predictably poor
We then never hear from the op again

noobs just dont seem to understand that they haven't come up with a new and revolutionary idea that is going to work just because they say so. TEC's haven't changed for a ever, so the results a responsibly predictable.

Noob see some of our comments as attacks but REALLY we are trying to help but they wont listen and the proof's in the results.

The above has just happened with what ever his name was who decided that his 226 TEC would be able to cool his over clocked 9650 just great ... Yeah Right !

But back to your question .. if you want to submerge the PC in oil that's fine But you should put normal water blocks on what ever you want to chill then pump the water to the cold side of the TEC then water cool the hot side of the TEC's and pump the water to be cooled to a large stack of radiators.

If you dont do this results will be poor

Thanks. We really are trying to help

[yngndrw]

xkon:

Elitism - This is something you see in not just any technical discussion area (Not just computer related, but anything technical.), but also any other area where knowledge is involved. (I've seen it on plenty of pet forums as well.) Unfortunately it is a fact of life with forums and you have to learn to ignore it, or course there are also cases where users will give good advice but will still come across as being elitist, in which case you should learn to not just ignore it, but to see through any elitism.

"Noobs" - This actually works both ways, on one hand the experienced people should understand that they too started knowing nothing, but new people should also understand that there is generally a cycle of the same questions.

Beliefs - There are some areas in which the best you can do is take an educated guess as nobody really knows. This is commonly a reason for heated debates both on forums and in real life. An example would be religion, I personally don't believe anything exists but others would put their lives on it. Opinions work both ways - Everyone has one.

With that said, let's get onto the subject. I do not pretend to have any experience in this area. My "views" are based upon observation alone.

Let's compare TECs to Phase Change. (General observations, YMMV.)

TECs:
Possible to get the best COP.
Generally a worse temperature differential.
Expensive !
Not the best for temperatures.

Phase change:
Still a good COP, but not the best.
Can have silly temperature differentials. (In a good way.)
Cheaper, but the tools can be costly if you make your own.
The best temperatures.

COP is something which is raised a lot in this area. I believe that while TECs can give a better COP, this is offset by the cost. If you need such a high COP, then you probably can't afford the extra price.

If you're just wanting to play, then by all means experiment. (I love playing with stuff, personally.) Just keep in mind that you should be starting with something easier, such as a lower power heatload or some much older hardware. Keep in mind that these experiments may, financially, be of no gain to you. (If you find that it's useless for your needs.)

Finally, on the submersion part. Keep in mind that when submersing your computer, the heat still has to go somewhere.

[[XC] gomeler]

Isn't the COP on TECs typically around 1 or worse? A chiller to achieve slightly sub-ambient temps(what a TEC would accomplish) would have a COP close to 2.

[Ultrasonic2]

A 345 Qmax TEC running at 16ish% of Umax can achieve a theoretical cop of 10ish while achieving a delta of 10c theoretically

The chiller i made had a COP greater than one and the coolit bores has a COP greater than one

[xkon]

so then. if i wanted to cool just the oil with the TECs, what's the best way to go about it? I had this crazy idea about 8 CPU waterblocks on a copper plate cooling the hot side of the TECs, while the cold side was stuck onto a copper plate with heatsinks screwed on and a fan pulling the oil through the heatsink fins. Is that at all feasible?
or
pumping oil out of the tank to a waterblock hooked up to the cold side of the TECs?
kinda like the one that Naekuh showed on a thread a bit ago. http://www.xtremesystems.org/forums/...40&postcount=5
It's actually more similar to the Dual TRUE setup.
I think I'll have to illustrate these ideas because I sure can't explain them well.

ULTRA!!! you've already done what i was thinking of. haha! had to go searching through your old threads to find it.
http://www.xtremesystems.org/forums/...61&postcount=1
so one loop for the oil (cools Computer), the other loop for the water (cools TECs)
How'd you build those giant sinks? I have a basic idea, but just wondering what tools were required.

[zipdogso]

The problem with cooling oils is their viscosity change.

Nicely runny at room temp, distinctly treacle at low temps you would require an enormously powerful pump to move it especially through small channels of a heat exchange like the one Ultrasonic made in the picture above.

I am not sure that the temp achieved at the point were the oil is sufficiently runny to pass the heat exchange would interest you at all.

[zipdogso]

A 345 Qmax TEC running at 16ish% of Umax can achieve a theoretical cop of 10ish while achieving a delta of 10c theoretically

The chiller i made had a COP greater than one and the coolit bores has a COP greater than one

The cooling power of this single TEC would be pretty low and not have any use on its own.....

BUT that is the theory of the multi-TEC chiller. Ten of these TECs built into a chiller would provide a very nice chilller for one of the latest CPUs.
Indeed if your thinking down the line of one of 6 core offerings and you have sufficient computing needs to utilise the power of it, then a chiller will be your only recourse with TECs.

But it goes without saying that without....
1.) a basic toolbox at least. 2.) being able to use the tools proficiently. 3.) Have a GOOD understanding of TECs and basic requirements for installaion etc. 4.) a GOOD understanding of basic electric circuits and formulas.
...you are not going to succeed in completing the project in the correct manner and the end result will not work how you originally proposed.

[Ultrasonic2]

Both of ZIPDOGSO post are 100% on the money. To recap, cool your PC with a water to water TEC chiller

I made that chiller on a MANUAL milling machine. and it's currently in my copper recycling bin.


my overclocked q6600 at full load for some time ( hot and cold plate temps )




im making a MUCH improved one at the moment when the money's available
http://www.overclock.net/peltiers-te...r-chiller.html

[xkon]

The problem with cooling oils is their viscosity change.

The cooling power of this single TEC would be pretty low and not have any use on its own.....

BUT that is the theory of the multi-TEC chiller. Ten of these TECs built into a chiller would provide a very nice chilller for one of the latest CPUs.
Indeed if your thinking down the line of one of 6 core offerings and you have sufficient computing needs to utilise the power of it, then a chiller will be your only recourse with TECs.

But it goes without saying that without....
1.) a basic toolbox at least. 2.) being able to use the tools proficiently. 3.) Have a GOOD understanding of TECs and basic requirements for installaion etc. 4.) a GOOD understanding of basic electric circuits and formulas.
...you are not going to succeed in completing the project in the correct manner and the end result will not work how you originally proposed.

yes. multi TEC is what is needed.
okie dokie. i'll do some more research, work on my tool proficiency, and get more familiar with the math involved before i really jump into this project. '

Both of ZIPDOGSO post are 100% on the money. To recap, cool your PC with a water to water TEC chiller

I made that chiller on a MANUAL milling machine. and it's currently in my copper recycling bin.

So, using the TEC cooler you made, you cool the hot side with chilled water (by a multi TEC chiller) to achieve the best results? and then you cool the multi TEC chiller with regular watercooling?

ok. so i think i'll put off cooling the hefty i7 /w dual fermi's in oil until i dabble a little and learn more.

Thanks guys! I'll still be on this thread bugging you guys about it though.

[Ultrasonic2]

So, using the TEC cooleryou made, you cool the hot side with chilled water (by a multi TEC chiller) to achieve the best results? and then you cool the multi TEC chiller with regular watercooling?

no no no Dont stack TEC's (it's possible but miles out 99.999% of peoples league)

ether uses a water chillier or a direct die system

[zipdogso]

Not wishing to correct you unneccessarily Ultra but I think he meant making 2 chillers...

The first with normal watercooling on hotside to produce chilled coolant to feed to the 2nd chiller hotsides to cool his coolant to eventually go to the PC...

One hell of a lot of work (not to mention $7-800 probably more.) but to give him credit he is thinking straight...

The second chiller with a Dt of only 10ºC needs chilled coolant on the hotside.

If he seriously was thinking this route as a USA citizen, opps... he is isn't he, (phase builders fairly easy to find, units reasonable cost.) he might find it better in get someone to build him phase.
It won't be as good COP-wise, he might still be only cooling his CPU, but he will get lower temps with it and possibly cheaper.

[Ultrasonic2]

Not wishing to correct you unneccessarily Ultra but I think he meant making 2 chillers...

The first with normal watercooling on hotside to produce chilled coolant to feed to the 2nd chiller hotsides to cool his coolant to eventually go to the PC...

.

umm yeah so hes's talking about TEC stacking.

[zipdogso]

umm yeah so hes's talking about TEC stacking.

LOL...I can see we are never going to see eye to eye on this....

I call stacking....literally stacking one TEC straight on top of another....The carp way to run TECs...

But I can see where the naming problems arise... 2 chillers like this is effectively a cascade but I can't see any problems running this NOT like the problems you get stacking, like I call stacking.

In my own tidy little world:-

stacking - a TEC or TECs directly on another.

Cascade - a series of TECs all chilling the same coolant run one after another or chillers with the output of one feeding the input of another.

Those premade "Stacked" TECs (often called cascades.) are to me 2, 3 or 4 stage TECs

Then I know in myself exactly what I am talking about.

[Ultrasonic2]

i dont think it's about seeying eye to eye cos we do. i think you didn't understand what he siad

if the cold side of one TEC is cooling the the hot side of another TEC .. ether by direct contact ( in one chiller ) or indirectly using 2 chillers . its still effectively the same thing.

He isn't suggesting getting 2 chillers and connecting hot side 1 then hot side 2 then taking this water to be cooled by a rad. then with the cold side connecting cold side 1 and then cold side 2 then sending it off to the load to be cooled

his quote said this
So, using the TEC cooleryou made, you cool the hot side with chilled water (by a multi TEC chiller) to achieve the best results? and then you cool the multi TEC chiller with regular watercooling?

That is stacking:- cooling the hot side of one TEC with the clod side of another TEC

[flak-spammer]

Ah, yes I see what he meant, and no it definitely isn't the most efficient way of doing it because it means you need to deal with probably 1200 or so watts of heat by the end. However it could be made to work but it isn't what I'd suggest to begin with. Honestly people think too big with TECs sometimes and you have to start small and get bigger as you understand the intricacies of these little but powerful devices. That being said, I would suggest just start off with cooling the chip and using regular water cooling for the hot-side.

[bamtan2]

I'm planning a build and wanted to add TEC into the mix

Oh yah. I suck at modding. I have few to no tools.

this thread went way further than it should have. the only answer to his questions are "no. stop."

[flak-spammer]

You do know it's very easy and possible to make TECs work? People who say "no. stop" should in fact be questioned. Saying "no" to someone is the easy way out, how about explain why not.

[Ultrasonic2]

i might point out NO one said

NO Stop