Will This Work???

[s7e9h3n]

The guys @ Petra's Tech Shop (www.shoppts.com) were kind enough to "lend" me 2 Silverprop blocks to try them before I buy them :toast: Originally, I was looking for just a TEC block, but the only TEC version that Silverprop currently makes is a Low-profile w/b. I've heard good things about the "normal" height gpu cooler, but not too much about the lp version. So naturally, I got both the blocks, took them home, and then disassembled them :p: in order to "make" my own Silverpro High Profile TEC GPU block. I've run into a couple of issues though. First I need to know if the base of the block is large enough and has enough capacity to cool a 50x50mm 226W TEC. Second question pertains to mounting the coldplate on the block. Here are some pics so you can see what I'm talking about:

Here are the two blocks disassembled. The top block is the regular-height w/b, and the bottom one is the L/P TEC block:
http://img229.imageshack.us/img229/8872/sptop1pu.jpg

View from the side:
http://img229.imageshack.us/img229/5589/spsides3gk.jpg

Here's what the 226W tec looks like sandwiched between the cold-plate and the w/b:
http://img231.imageshack.us/img231/9585/tece5uv.jpg

Now here's one of the problems. The TEC extends that tiny bit into the mounting holes. Any suggestions on how I'd get around this prob?
http://img8.imageshack.us/img8/8923/techole9kx.jpg

I'm gonna mount this on my monster x850xtpe and try to take some records down with it (although I may just slap a phase unit on if this doesn't get cold enough). Since it's ATI, I only need two of the holes for mounting purposes. Should I even bother trying to use this 226W tec? Or should I just use a 126W 40x40 TEC I have lying around? Or maybe even stack 2x126w? The block will be in it's own cooling loop so shouldn't be a problem dumping the heat produced.......ANY AND ALL HELP IS APPRECIATED.....thanks.......

[s7e9h3n]

Help?

[MaxxxRacer]

If you skew the 226watt TEC slightly it should work. Just need to make sure not much of the TEC is exposed or you will probably burn it out.

[s7e9h3n]

If you skew the 226watt TEC slightly it should work. Just need to make sure not much of the TEC is exposed or you will probably burn it out.
I think that it may not be possible to skew the TEC since it extends out on both sides......Is it possible to "shave off" or file down a TEC without screwing it totally up? Of course I'd make sure to re-seal anything that may get exposed......

[Mr. Tinker]

You can hold it up to a bright light and see if there are elements in the ceramic where you want to chop.

[s7e9h3n]

You can hold it up to a bright light and see if there are elements in the ceramic where you want to chop.
How close can I get to those elements without fear of anything shorting out? Or is it possible to operate a TEC without all the elements in place -> possibly remove the element around the area?

[Johnny Bravo]

no the tec is wired up in a series format so that would knocked out the whole element, unless you got fanct and make a wire jumper - http://www.noahprecision.com/images/peltier.gif

[Mr. Tinker]

I've suspected that if the p-n semiconductor pairs are wired in parallel, removing a couple pairs wouldn't hurt as long as you can manage to keep the circuit in tact. But I'm pretty sure for the pelt effect to take place, the wiring of the circuit needs to be series. That may be an oversimplification on my part.

http://www.quick-ohm.de/peltier/kpics/bild6_gen.jpg
http://www.quick-ohm.de/peltier/kpics/bild5_gen.jpg

http://www.quick-ohm.de/peltier/erlauterung_generator.html

I think what maxxx is suggesting is that you can turn the tec just a hair, so that two screw holes are more obstructed, but at least two other screw holes are completely open.

The thing is, actual tec elements are cheap, so it might not be too dangerous (financially) to get with the intrepid experimentation.

[s7e9h3n]

Thanks for the help guys...the pics are very useful :toast:

Well, decided to see just how plausible it would be to file away from the edge of the TEC so the screws would be able to go through without damaging/touching any of the elements. I'm wondering, though, if I may be cutting it a little close as it looks like the clearance for the screws would be less than 1mm:

Here's the placement of the TEC's elements:
http://img8.imageshack.us/img8/3843/elements3ek.jpg

It looks as I have enough clearance to go through, but just how safe is it to cut that close?
http://img230.imageshack.us/img230/4482/clearwhole8yd.jpg

Of course, there's always Max's alternative by skewing the TEC, but doing that would require the cold plate be secured to the w/b using the same holes to mount. Would I be able to apply enough uniform pressure to the TEC to get it to function optimally?:
http://img230.imageshack.us/img230/2356/skew6po.jpg

Thanks again for the help....I'm being a pest since I don't want to blow this card up on first boot :p:

[afireinside]

TEC will need way more pressure than the GPU die. You won't get good temps with the skew and single bolt through all method.

[s7e9h3n]

TEC will need way more pressure than the GPU die. You won't get good temps with the skew and single bolt through all method.
It's possible to put enough pressure on the TEC by using nuts to first mount the cold plate and then bolting the block down to the card. The thing I'm worried about is if the pressure on the TEC will be uniform enough since it would be skewed and using only two bolts for pressure......

[msimax]

<a href='http://www.smileycentral.com/?partner=ZSzeb008_ZNfox000' target='_blank'><img src='http://smileys.smileycentral.com/cat/23/23_11_59.gif' alt='' border=0></a> do it <a href='http://www.smileycentral.com/?partner=ZSzeb008_ZNfox000' target='_blank'><img src='http://smileys.smileycentral.com/cat/36/36_1_1.gif' alt='Smile' border=0></a>

[msimax]

from the pic above you should be good id get small round file use a marker to make the half circle on the tec and file it a little a time and if your successful
mount everything up and run the tec by its self to see if it get nice and frosty
then turn it off let it thaw wipe the water off prepare your gf card and slap it on
btw.. i made mines from a swifttec tec unit made to cool a a64 cpu but the coldplate is so large i drilled and tap it i like ur setup cause if u getit going u can use that in a cf/sli setup which would be killer

iwas also thinking u should use the 4 hole setup

[s7e9h3n]

from the pic above you should be good id get small round file use a marker to make the half circle on the tec and file it a little a time and if your successful
mount everything up and run the tec by its self to see if it get nice and frosty
then turn it off let it thaw wipe the water off prepare your gf card and slap it on
btw.. i made mines from a swifttec tec unit made to cool a a64 cpu but the coldplate is so large i drilled and tap it i like ur setup cause if u getit going u can use that in a cf/sli setup which would be killer

iwas also thinking u should use the 4 hole setup
I had thoughts of using my old swiftech cooler as well (that's where the 226w tec came from), but that thing was so gigantic, it would have snapped my card off in the pci-e slot. I do, however, hear that there's a mod for the swiftech mcw6002 cpu cooler -> gpu pelt cooler. Anybody got any links?

[afireinside]

Not sure... How do you snap a card off a PCIE slot when it's on a table? :D Just stick some foam under it. I assume you're talking about an MCW5002? I bought an MCW5002-PT with a 320 watt TEC from FCG and was going to C clamp that bad boy to my x800pro but couldn't find a suitable PSU :(

[s7e9h3n]

Not sure... How do you snap a card off a PCIE slot when it's on a table? :D Just stick some foam under it. I assume you're talking about an MCW5002? I bought an MCW5002-PT with a 320 watt TEC from FCG and was going to C clamp that bad boy to my x800pro but couldn't find a suitable PSU :(
I wuz exaggerating :p: , but I do believe it's a mod with a 6002 though.......

[msimax]

http://img394.imageshack.us/img394/9149/imga04525bp.jpg
http://img394.imageshack.us/img394/2488/imga04474up.jpg
http://img394.imageshack.us/img394/7715/imga04481ej.jpg

[JasonDTM]

I think Viperjohn does the mod, might want to contact him on what to do :)

I wuz exaggerating :p: , but I do believe it's a mod with a 6002 though.......

[n00b 0f l337]

I've done it as well, quite easy, just gotta build a coldplate clamp setup.

[msimax]

I've done it as well, quite easy, just gotta build a coldplate clamp setup.
any pics on a how to??

[n00b 0f l337]

Nah but i got 2x 6000' blocks here, i may do another for a 320watt tec. I'll post it up if i do.

[afireinside]

any pics on a how to??

nol never takes pics

[Holst]

You will struggle to file the TEC down unless you have a good file. Diamond file might do it.
The ceramic TEC are made from is very hard.

I dont think there is an easy solution to this problem.
Replacing the coldplate with a larger peice of 5mm copper would probably be the easyest method. Then you can make a custom clamp.

scewing the TEC over and using the card mounting holes to clamp it wont work anywhere near as well as clamping the TEC seperately.. The coldplate needs allot of pressure against the TEC to get a good thermal contact.
But you can try it this way for testing, it wont cost you anything.

[msimax]

he could use a dremel too. i got a whole bunch of mini bits that would work

[s7e9h3n]

he could use a dremel too. i got a whole bunch of mini bits that would work
...and that's just what I did.......

http://img21.imageshack.us/img21/1686/spbottom9lt.jpg

http://img9.imageshack.us/img9/272/spmountf7op.jpg

[msimax]

good job man<a href='http://www.smileycentral.com/?partner=ZSzeb008_ZNfox000' target='_blank'><img src='http://smileys.smileycentral.com/cat/36/36_1_55.gif' alt='Clapping Hands' border=0></a> <a href='http://www.smileycentral.com/?partner=ZSzeb008_ZNfox000' target='_blank'><img src='http://smileys.smileycentral.com/cat/36/36_1_11.gif' alt='Thumbs Up' border=0></a>
<a href='http://www.smileycentral.com/?partner=ZSzeb008_ZNfox000' target='_blank'><img src='http://smileys.smileycentral.com/cat/36/36_11_21.gif' alt='Na-na-na-na' border=0></a>

[Holst]

Well done, thats excelent :D

Please post your results as soon as you can :)

[n00b 0f l337]

nol never takes pics
AFI if your implying that I havent done what i've said i've done, you are sourly mistaken.

[Overconfidence]

AFI if your implying that I havent done what i've said i've done, you are sourly mistaken.
Oo; I've seen NoL and his 6 foot bong. ;)

*cough*

...*cough*

[Aphex_Tom_9]

nice work, now insulate that beeotch and lets see some temps/clocks

[s7e9h3n]

Ok....I finally got around to insulating my card and mounting my my ultra-d up on my autocascade. No pics atm, but my x850xtpe @ ~1.6-1.65Vcore was idleing around -12C, but what's odd is that the TEC doesn't seem to hold a load worth $hit. I installed a thermocouple that touches the cold-plate and it's read temps up to 28C - which isn't too much better than using straight water and no TEC. For comparison, take a look at this which was done with a swiftech mcw50 and just regular water:

http://img241.imageshack.us/img241/5061/40k7au.jpg

Note the 32C high reading......A 28C reading on the cold-plate would equate to a much higher core temp imo. I'm using a meanwell 320 @ 13.8V to power the 226W TEC which is rated @ 15.4V. I would think that my temps should be able to hold MUCH better than they have....any ideas? :confused:

[Mr. Tinker]

Be sure that clamping pressure is very strong. Use AS ceramic on the cold side of the tec and cold plate, and AS5 on the hot side. What is the rad situation?

[n00b 0f l337]

The 15.4 is probably the pop and burn voltage.
13.8v is normally the max for these tec's. Running it at high voltages decreases efficiency. Load it up to 100% then lower the voltage coming from meanwell and watch the temps.

[s7e9h3n]

Thx for the suggestion guys :toast: BUT :D ....I've run TEC's before and I'm pretty sure I'm familiar with the basics. Clamping pressure should be very tight due to the 4 bolts securing it to the wb. And, isn't 13.8V optimal voltage for a 15.4V TEC? It's been awhile, so I may be a little rusty :p: Oh yeah...I lied....I did take one pic when I first installed the TEC in the loop. It pulled pretty hard down to this temp - no load. Doesn't a deltaT of ~50C mean that it's running OK? TIA

http://img38.imageshack.us/img38/5887/tec8rj.jpg

Oh yeah - the loop is actually from a bunch of older parts I had lying around, but they still are decent performers - Eheim 1250 - Dtek Customs Heatercore - SilverpropTEC - bay res - Eheim -

[n00b 0f l337]

Your other problem is probably that the hotplate you are using is not very peltier optimized. The surface area isn't that high and its really only cooling the center. This leads to bad loads.

[s7e9h3n]

Your other problem is probably that the hotplate you are using is not very peltier optimized. The surface area isn't that high and its really only cooling the center. This leads to bad loads.
Hmmmm....that's true...how much area on a pelt it required to cool so it functions at optimal levels? The only part that's not being directly cooled on the pelt is it's outside edges.

[Holst]

Thats a good point NOL.

Really you need to cool the entire hot side to get the best temps, as this will give you the best deltaT onto the coldplate.

You should also run a coldplate thick enough to "spread" the load so that the the TEC gets as even as possible heat load (not easily acheived with CPU/GPU dies)

Can you probe the edge of the TEC and see what the temps are at the edges please.

Due to your waterblock design probably only 50% of the TEC hot side has the cooling bit of the waterblock close to it, the 50% around the edge is only cooled by conduction through the GPU coolers base (which is way too thin for this application)

It might help to add copper plates either side of the TEC to spread the heat better, but this might not work as the extra TIM interfaces might loose you any gains.

Manufacturing a thicker coldplate might be more use, but I think the hot side is the problem.

[s7e9h3n]

Thats a good point NOL.

Really you need to cool the entire hot side to get the best temps, as this will give you the best deltaT onto the coldplate.

You should also run a coldplate thick enough to "spread" the load so that the the TEC gets as even as possible heat load (not easily acheived with CPU/GPU dies)

Can you probe the edge of the TEC and see what the temps are at the edges please.

Due to your waterblock design probably only 50% of the TEC hot side has the cooling bit of the waterblock close to it, the 50% around the edge is only cooled by conduction through the GPU coolers base (which is way too thin for this application)

It might help to add copper plates either side of the TEC to spread the heat better, but this might not work as the extra TIM interfaces might loose you any gains.

Manufacturing a thicker coldplate might be more use, but I think the hot side is the problem.

I've got this I could try to throw in the sandwich :D
http://img211.imageshack.us/img211/7760/coldplateamd2ka.jpg

[n00b 0f l337]

See I dont beleive we should be looking to spread the load more, we should be looking to just get the heat out of the entire thing faster. A THINNER hot plate with a design made to cover all of the 50mm surface, (therefore entire hotplate beingaround 65 mm), is what we really need. Not really impingement but spread jeteson is what would really help.

[s7e9h3n]

See I dont beleive we should be looking to spread the load more, we should be looking to just get the heat out of the entire thing faster. A THINNER hot plate with a design made to cover all of the 50mm surface, (therefore entire hotplate beingaround 65 mm), is what we really need. Not really impingement but spread jeteson is what would really help.
Ok...this looks as if it's getting a little TOO complicated :p: I think it's time I fly out the Chilly1's and finiish off my mach2 as a gpu cooler :D

[Holst]

See I dont beleive we should be looking to spread the load more, we should be looking to just get the heat out of the entire thing faster. A THINNER hot plate with a design made to cover all of the 50mm surface, (therefore entire hotplate beingaround 65 mm), is what we really need. Not really impingement but spread jeteson is what would really help.

Absolutely.. this block isnt designed for a TEC this big, and its not working as well as it might were you to redesign it

But I was trying to come up with an easy practical solution... that might work.
Although this idea of adding a copper plate might make things worse, but its easy enough to test it.

[s7e9h3n]

Absolutely.. this block isnt designed for a TEC this big, and its not working as well as it might were you to redesign it

But I was trying to come up with an easy practical solution... that might work.
Although this idea of adding a copper plate might make things worse, but its easy enough to test it.
Or I can always just toss my 40mm 126w TEC under the block and compare results....

[Holst]

Try both if you can, might be interesting.

The copper plate might help, but its a bit hard to tell why the temps are bad, and exaclty what effect it will have (if any)
Interesting to find out if it does.

[Piper]

That was a good analysis of what is limiting the cooling. I also agree that the big hinderance is the entire surface area of the hot side of the TEC is not being properly cooled. I don't think adding the larger copper plate will do much, but if that 126w TEC is up to the job then it's the better looking solution. How many watts does that GPU kick out ?

Also, I think you can find some 172W TECs out there that are 40x40 and still 12-15V ? I know there are ome 24V ones but unless you can supply the voltage you'll be loosing ground.

[n00b 0f l337]

The main problem is running things at their max voltages. So many people do it to get lower idle's then complain that TEC"s cant hold a load. :(

It's like running your cpu at max possible safe voltage with stock cooling. You just don't do it. So why should you do it to a part of your cooling? Think of the peltier as your cpu, give it only what it needs.

[Piper]

Also good info, look at your heat burden from the GPU and give the TEC enough power to pump that burden + a little headroom, like another 25%.

But you are still going to find yourself limited by that block design unless you can get your hands on a 40x40 that will handle the load.

[n00b 0f l337]

Even then, the pattern of the block isnt optimized for peltiers. Optimized means a thin base in my opinion with a high surface area, medium restriction design. pins and diamonds are good, but I think we may need to look into larger forms of jets and maybe switching peltiers to a higher tubing ID for flow.

[s7e9h3n]

So many people do it to get lower idle's then complain that TEC"s cant hold a load. :(

I thought it was exactly the other way around - Run less voltage, you'll get lower idle but higher loads. Run more voltage, you'll get higher idle with lower loads. At least that's how it works in phase (I know it's not the same thing, just thought the principals may be related)

[n00b 0f l337]

Nope, increasing the voltage will produce a higher net cooling effect. But the efficiency is worse.

[s7e9h3n]

Nope, increasing the voltage will produce a higher net cooling effect. But the efficiency is worse.
Hmmm...OK...well maybe before I start tearing the loop down, I should play with the voltage supplied to the pelt as you said earlier...:toast:

[n00b 0f l337]

Sounds good post results