• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Eric's Ultimate Guide to TECs

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
How about a guide or something about protecting your CPU/mobo from condensation. I recall reading something somewhere but I can't remember, one that involves insulating foam on the back of the mobo, around the CPU socket and around the heat sink or water block plus dielectric grease inside the socket to keep moisture out.
 
Premmer said:
The recommended compression for a TEC assembly is 150 to 300 pounds per square inch of module surface area. Using the following formula, you can calculate the torque setting per screw:

T = (C x D x F x in^2) / (# of screws)

T = torque per screw (in-lbs)
C = torque coefficient; generalized values for copper/mild steel (0.36 dry, 0.18 lubricated)
D = nominal screw size (8 = 0.164, 10 = 0.190, 1/4 = 0.250)
F = Force (lbs / in^2)
in^2 = module surface area (length x width)

Check the torque setting after one hour and retighten if necessary.

Example: Assuming (4) # 8 screws are used to secure a coldplate used with a 62mm (2.44 in) square module surface area and a surface force of 300 lbs per in^2 is required, what is the torque setting required per screw?

T = (0.36 x 0.164 x 300 x 5.95) / 4 = 26.35 in-lbs per screw



If a torque wrench/screwdriver is not available, a properly-ranged pull spring scale attached to the end of an L-shaped hex wrench can be used to establish the screw torque setting. To calculate the required spring scale pull force, apply this formula:

F = T / D

F = Force (lbs)
T = Torque (in-lbs)
D = Distance (in)

Example: What is the spring scale pull force required at the end of a 3.75 inch long L-shaped hex wrench to produce a 26.35 in-lbs screw torque setting?

F = 26.35 in-lbs / 3.75 in = 7.02 lbs pull force
 
Last edited:
Hey Premmer! It's great to run into you over here at OCForums as well. . .

As for compression, I always just torqued the screws as far as humanly possible with a wrench until they were just about to start burrowing through the cold plate. Does that seem good enough or is it really necessary to use such an exact calculation?

By the way, I almost have my rig setup to do that experiment we talked about. I'll post back in our original thread when I manage to do it. (Probably about a week more while I wait for my motherboard to be RMAed.)
 
Revivalist said:
Hey Premmer! It's great to run into you over here at OCForums as well. . .

As for compression, I always just torqued the screws as far as humanly possible with a wrench until they were just about to start burrowing through the cold plate. Does that seem good enough or is it really necessary to use such an exact calculation?

By the way, I almost have my rig setup to do that experiment we talked about. I'll post back in our original thread when I manage to do it. (Probably about a week more while I wait for my motherboard to be RMAed.)
Nice OC on the Opteron. I can't say for sure if it's excessive force. I'm just adhering to what more than one TEC manufacturer has stated, 150 to 300 psi max. Based on what I've read, some individuals seem to really go overboard.

Periodically the torque question has surfaced and I never read a response that had any detailed information other than the force range noted above.

So for those individuals who prefer not to just wing it, they have the data and a simple method of setting the torque.
 
There is some really imortant information I didn't see covered: the relationships of Imax, Qmax, and Vmax when a voltage lower than the TEC's specified Vmax is used as the input. I haven't seen a TEC with a Vmax above 12V but yet *less* than 14.7. Thus, they will all be running *at least* 18% lower voltage than their manufacturer specified Vmax when the PC's ATX PSU is employed. In this example, 24% lower voltage:

Example TEC, IMB Products model# inb127-1.4-1.05

Imax 8.6A
Qmax 84W
Vmax 15.7V
DTmax 69C
ATX PSU 12V


So...

Question #1
If one runs the TEC at 12V instead of 15.7V, does the current (Imax) go up or down as a result? Standard electrical theory would lead me to believe Imax would increase to maintain Qmax, given a decrease in Vmax (V * A = W). However, I don't know enough about the design and behavior or TECs, thus the reason for the quesiton. I.e. will a 12v input in this example make:

Imax = (12/15.7) * 8.6 = 6.57 Amps?
Or will it increase to 11.25 Amps? (since the TEC's total draw is apparently 15.7V * 8.6A = 135W, due to the TEC's inefficiency)



Question #2
With the 12V ATX PSU input will:

Qmax = (12/15.7) * 84 = 64.2 Watts?
Or will it stay at 84 Watts?



Question #3
If the GPU thermal output at full boogie is 50W (this is probably way too low an estimate for the top end nVidia and ATI GPUs--68xx/78xx, x18xx/x19xx, but probably about right for 66xx/FX59xx/9800/x800/etc), then will:

Delta T = (1-(50/64)) * 69 = 15.09 degrees C?
Or will it be 27.92C, if Qmax stays at 84W?


Assuming Qmax does decrease when using a 12V input, at an ambient case temp of 25C, the cold side of this TEC, assuming a 100% efficient water or air cooling system, would be 10C (or 50F). Anyone know a ballbark figure as to what % higher the GPU core temp typically is compared to the TEC cold side plate?

In my personal setup, my current loaded GPU temp on my FX5900 is in the high 50C range (maybe low 60s), and mid 40s idle. So, I could expect what, 20-25C under load now, or will core temp not be that much higher than the cold plate? That's a pretty decent drop, but still not anywhere near freezing, which I would think would be an absolute requirement for major overclocking, which is my goal in building my TEC.

Even though it would cost me a new higher amp 12V rail ATX PSU, I'm really hoping the answer to Qs 1 & 2 are that higher amps will be drawn and an 84 watt Qmax maintained, because this would yield a Delta T of 27.92, an additional 13C drop, which will give me a cold side temp of -3C assuming 100% cooling efficiency (not likely, but maybe close).

If lowering the voltage lowers all these other parameters, I guess I'll just have to drop some cash on a dedicated variable voltage PSU, although I'd really really like to avoid it, as I'm designing an air cooled, *compact* (compared to liquid that is) solution to mount directly to the card. Once my Qs are answered (hopefully accurately, and swiftly), I can order my parts and get to work. :burn:

It should be a fun process, probably time consuming, potentially fruitless. However, I've not read of *anyone* doing an air cooled pelt for GPUs. I like being unique, and "doing what *they* say can't be done". :cool:

Besides, quickly evacuating a paltry 185-220 watts with air over fins shouldn't be too difficult. :bang head:
 
The difference between using 12v versus 15v is negligible. Years ago, I knew many people who experimented at the different voltages and found similar CPU temps. We found that the relationship between V and Q isn't linear, at least not near Vmax. The relationship between V and I may not be linear either, but I can't remember (the TEC may change resistance).

I'm trying to comb the archives of EOCF to find the specific information. I will post again when I find it.
 
Could you use a TEC to cool a heatsink for CPU?

Or is the bottome line that water is a must when using a TEC?

When you use a TEC and water, do you use the water to cool the TEC or the TEC to cool the water?

I want to go with a better cooling setup than my stock crap, but damn, I'm afraid to push water through my system.
 
jeepguy_1980 said:
Could you use a TEC to cool a heatsink for CPU?

Or is the bottome line that water is a must when using a TEC?

Water mostly, switech made a HSF TEC for a while but I recommend water. Learn a bit with water and when you have the knowledge you can try some tricky stuff like TEC HSF setup.


jeepguy_1980 said:
When you use a TEC and water, do you use the water to cool the TEC or the TEC to cool the water?
Either, if you cool the water it is a TEC chiller. I would not run a chiller setup, try a regular setup for now.

jeepguy_1980 said:
I want to go with a better cooling setup than my stock crap, but damn, I'm afraid to push water through my system.
You probably wont be happy with a HSF TEC unless your going with a fairly small heat load. Read up alot, with any luck I will be getting some test data soon for TEC users to work with to make things easier.
 
How much thermal wattage does a core 2 duo at 1.55vcore and 3.6ghz give off? It says the thermal spec of conroes are 65watts, but thats probably at stock speeds, stock voltage...
 
Let’s say Bob wants to cool his XP 2000+ which generates 60-watts of heat. Bob just so happens to have a ‘Magic Heatsink’ that magically cools a heat load of 60-watts to room temperature (in this case room temp is 25C, remember that). Not a bad heatsink (must be a Thermalright). So Bob figures, “If I attach this bad boy to the hot-side of my 226-watt TEC then I can maintain room temperature on the hot-side of 25C, then the cold-side will be -26C assuming a Delta T = 51C.” Bob is about to get a rude awaking.

What’s Bob’s problem? The heat load generated on the hot-side of the 226-watt TEC is not just 60-watts anymore (the original heat output of his CPU). It is the original heat output of the CPU (60-watts) plus the Cooling Power of the TEC (226-watts). 60watts+226watts = 286-watts of heat generated on the hot-side of the TEC. That’s enough watts to heat 1 liter of water by 4.1C in one minute. See Bob’s problem? Well, it’s about to become yours. :D


THL = Heat Load + Max Cooling Power

-THL (Total Heat Load) = The heat load generated on the hot-side of the TEC in watts.
-Heat Load = The heat generated (in this case) by the processor measured in watts.
-Max Cooling Power = Maximum TEC rating in watts. (e.g. a 226-watt TEC)

[Originally written 2/9/03]

Now i've created an account to correct this peace of information. As it'is incorrect

The above suggests that the total heat load rejected to the hot side is equal to the load being cooled by the TEC (CPU) + the rated cooling power of a TEC
In the above case that is given as 226watts (Qmax)

This is not the case however. If we presume the TEC is being run at max volts (Umax) then the TEC will be consuming FAR more than Qmax.
Qmax is an indication of the amount of heat transferred from one side of the TEC to the other, to a delta of 0 NOT the amount of electricity required to do so.

In fact a 226watt TEC at Umax will consume approximately 474watts of electricity to move the 226wats of heat to a delta of 0

So the equation for how much heat will need to be cooled is actually CPU wattage + Electricity consumed by the TEC
 
So if you have a 95W cpu + 200W Tec ... you're actually dumping 495W of heat into the waterblock?

No it wouldn't (though close), it would use less than that because the electricity used is also dependent on the load applied to the cold side.

A 226watt Qmax TEC cooling a 95 watt load while the hot side is kept at 300kelvin would result in a theoretical 414 watts used. so 509 to be cooled

I might point out the original poster neglected the most important factor when dealing with TEC's and that's COP. As the input voltage is reduced relative to Umax (max volts) the COP increases. This means a TEC running a 20% of Umax cad move far more heat than electricity applied

This is why you under volt them
 
Hi.
i'd like to ask if you can make your own tec.
i'm a newbie at this so i don't really understand technical stuff so pls
try to explain in layman's terms.
by the way since this post is a bit old and aircoolers are much better now like thermalright's IFX-14.
can you now use those to cool a TEC's hotside?
 
This may be Late, but if anyone has questions regarding TEC's I'd be happy to Answer them (Im an Engineer in the Business).

Hey, :welcome: to the forums Smittys!

It will be good to have an expert on TECs here. I've run them before, but wound up killing hardware (due to my mistakes, not due to hardware failure with the TEC).

Once was due to overheating due to the pump not coming back on when the power came back after an outage. The other was due to condensation from a pinhole in my conformal coating.
 
Thanks for the Welcome! Eweast's guide covered all the basics.

Ouch on your failures, generally for the Computer Enthusiast its better to get the cheapest TECs you can find (Sized appropriately) because you will definitely fry some.

I haven't seen TECs become real popular for CPU cooling, as they are not efficient and require a large power supply (often a separate PSU).
 
I did some benching on one a while ago, cooled the tec with icewater, which then cooled the cpu. Sucker got nice and cold. Did warm up on high load high voltage testing though, apparently a 136w (at 12.5v) tec isn't enough when it's being run at 11.5(power supply i ran it off didn't appreciate it much).
 
Thanks for the Welcome! Eweast's guide covered all the basics.

Ouch on your failures, generally for the Computer Enthusiast its better to get the cheapest TECs you can find (Sized appropriately) because you will definitely fry some.

I haven't seen TECs become real popular for CPU cooling, as they are not efficient and require a large power supply (often a separate PSU).

Actually, the TEC was fine both times. The first time I cooked the CPU, the second time, the pins corroded off the CPU. Heck, the first time, in addition to cooking the CPU, it warped the top of the waterblock.

TECs are ok, but it is about the same power cost to run a phase change system, and you get a lot colder.
 
Thanks for the Welcome! Eweast's guide covered all the basics.

Ouch on your failures, generally for the Computer Enthusiast its better to get the cheapest TECs you can find (Sized appropriately) because you will definitely fry some.

I haven't seen TECs become real popular for CPU cooling, as they are not efficient and require a large power supply (often a separate PSU).


This isn't the full picture ..i wrote this in the past



Q. I’ve heard that TEC’s are very inefficient is this true. And what about phase change ?

A.
To begin with the Term “Efficient” is NOT the correct term to use when comparing the amount of electricity used to what It’s able to move. As this can exceed 100% by a long way. When talking about moving heat we use the term C.O.P (Coefficient of Performance)

The C.O.P number is ascertained by taking the Qc (Which may NOT be QMax) and dividing it by the input power. Higher the number equals higher the “Efficiency”

Now there is NO blanket rule to whether or not a TEC is “Efficient” (high COP). This is the case because the COP of a TEC changes depending on the input power relative to it’s max. As you increase the input power the COP is reduced (Less efficient).

If we compare the COP of common phase change systems to a TEC system including pumps and fans. The COP can be very similar PROVIDING you have a relatively low input power relative to the max of the TEC. However to achieve a TEC system with such high COP you’ll need to use many under voted TEC’s. Making the unit far to large to direct die cool your PC. You’ll have to make a water chillier for that kind of COP. A Single TEC direct die cooling a PC will require a high input power relative to it’s max giving it a comparably MUCH lower COP than phase change.

For a 360Qmax TEC running at 100% of Umax it would have a COP of about 0.6 (360Q/600)
For a 360Qmax TEC running at 50% of Umax it would have a COP of about 2.3 (269Q/114)


129330d1256895419-everything-youve-wanted-know-about-tecs-untitled-1.jpg

This graph shows you the relation ship between input power and watts moved in Q (orange) and Electricity used (purple) and total watts your cooling system will be cooling.
As long as the orange line is above the purple one the TEC has a COP greater than 1


NOTE: As you decrease the input power relative to the Max of the TEC, the maximum achievable delta is also decreased. So if you have a phase change and TEC system running at the SAME COP the phase change system will always result in lower temps because the phase change unit will have a far greater delta.


129329d1258277801-everything-youve-wanted-know-about-tecs-untitled2.jpg

This graph shows you the relationship between input voltage V delta

Please note these aren't exact numbers, just used to demonstrate my point
 
Back