New pelt setup underway! **Pics Inside**

[Revivalist]

Ok, I got everything I need for a new pelt setup. Now it's time to start operating. :D

http://img299.imageshack.us/img299/2779/01items6wq.th.jpg (http://img299.imageshack.us/my.php?image=01items6wq.jpg)

The setup includes the following...

• Maze4-1 CPU waterblock: (dangerden (http://www.dangerdenstore.com/product.php?productid=133&cat=58&page=1))
• 320W (50mmx50mm) Peltier: (15V max, 26A max, potted) (ebay (http://cgi.ebay.com/Awesome-320-Watt-Thermoelectric-Peltier-Cooler_W0QQitemZ7557309657QQcategoryZ4660QQssPageN ameZWDVWQQrdZ1QQcmdZViewItem))
• Meanwell 480-15: (480W, 13.5-18V, 35A)(trcelectronics (http://www.trcelectronics.com/Meanwell/pdf/sp480.pdf))
• Power relay: (bestbyteinc (http://www.bestbyteinc.com/prodinfo.asp?number=LIQ-SWI-RSK))
• Full Insulation Kit for Maze4-1: (includes neoprene pads, conformal coating, and thermal grease) (dangerden (http://www.dangerdenstore.com/product.php?productid=93&cat=31&page=1))
• 5.25” bay triple fan cooler: for cooling the Meanwell power supply (ebay (http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=6818795018&rd=1&sspagename=STRK%3AMEWN%3AIT&rd=1))
• Power chord, 12 gauge wire, connectors, & rubber stand-offs: Lowe's hardware

I'll keep posting more pics as progress comes along. . . .

The only thing I'm sorta of wondering about right now is how to know the amount of torque to use when screwing the cold plate onto the block with the pelt in between. I read that there should be 200-300 lbs per square inch. But I can't find a torque wrench that will fit the holes of the Maze4. Also, how would I compare the amount of torque from a torque wrench to the lbs per square inch on the cold plate? (Just add up the torque on each screw and divide by the area of the plate?)

[Holst]

Tigten it up as tight as you can without damaging the bolts.

You cant overtighten the coldplate.

I used to clamp mine in a vice and then tighten the screws to get it as tight as possible.

[KILLorBE]

You should gradually tighten each screw.

This is how I do it:
Put everything together but don't tighten the screws yet, clamp everything together either with a strong clamp or your fingers (Most pressure should be on the center of the TEC so pressure is evenly spread), now tighten the screws by turning the shaft of the screwdriver between 2 fingers (Do NOT use the handle yet, as you'll easily tip over the coldplate slightly, thus ending up with uneven pressure=bad contact=bad performance).
Depending on the thread of the screws you *should* tighten each screw by 1/8 - 1/4 turn crosswise (IE: Left Bottom, Right Top, Left Top, Right Bottom, Right Top, Left Bottom, Right Bottom, Left Top....etc, You'll find a way, just skip one screw ((Counter)clockwise) after tighentening 4 screws.

*Should* = You should only have fun ;)

If you use a vice like Holst it's not really important to tighten the screws crosswise, but personally I would still tighten them crosswise, only not by 1/8 - 1/4 turn, but more likely by 1/2 turn.

Also, as long as you use a normal screwdriver it's pretty much impossible to overtighten the screws....unless you don't use a vice and tighten every screw at once.

[n00b 0f l337]

I use a set of clamps. Get them insanely tight and then tighten down. :)

[Revivalist]

Thanks guys! . . .

I don't actually have any vice grips, so I went ahead and just tightened them all up by with the L-wrench as hard as I could, turning each screw about a 1/4 turn at at time, as suggested. . . And of course havin' fun all the while! . . . Here are my very first frosty pics (mom would be proud. :p: ) . . .

http://img516.imageshack.us/img516/6214/03frost7fh.th.jpg (http://img516.imageshack.us/my.php?image=03frost7fh.jpg)

http://img345.imageshack.us/img345/4753/06frost8og.th.jpg (http://img345.imageshack.us/my.php?image=06frost8og.jpg)

[Banzai]

Nice. What temps you getting underload ?

Ryan

[Revivalist]

Nice. What temps you getting underload ?

RyanThanks. Hopefully I'll be able to report on temps tonight. . . Right now I'm working on a volt-mod on my video card before I put it all back together again.

[Revivalist]

Well, here are the results so far. I hit -7C idle! Here's some pics . . .

Testing the setup . . .

http://img164.imageshack.us/img164/2843/11peltsetuptesting5bb.th.jpg (http://img164.imageshack.us/my.php?image=11peltsetuptesting5bb.jpg)


Running steady at -7C idle.

http://img164.imageshack.us/img164/7392/12negative7clowestever8ox.th.jpg (http://img164.imageshack.us/my.php?image=12negative7clowestever8ox.jpg)

[KILLorBE]

I can't really comment on your results as I never ran a WC TEC setup (I'm one of those few idiots that ran an aircooled TEC ;) ).
Also the kinda CPU you're running matters a lot...I mean 1.36V could be several CPU's, but at least it seems that your TEC is mounted properly.
Give us some more specs...please :fact: (CPU, MOBO, RAM)
And don't let those bums scare you, their sigs may say FX69, triple 8200GTX XXXL, 40GB 1-1-1-3 Moshzy memory, DoYouFIt Better than Ultra Expert MOBO, it's all lies, they borrowed a couple of my P1's and PII's ;)

Specs, we want specs!!

[sin0822]

ok so the lucite top is ok for those temps?

[Revivalist]

Ok, I'm up and running . . . Motherboard Monitor reads the following:

-9C at idle (underclocked to 1.06GHz and undervolted to 1.07V)

28C at full load for 20 min (overclocked to 2.60GHz and overvolted to 1.70V)

Here's a screenshot at idle . . .

http://img382.imageshack.us/img382/3549/idletempsnegative8c7gx.th.jpg (http://img382.imageshack.us/my.php?image=idletempsnegative8c7gx.jpg)

I was surprised about a few things . . .

1. The CPU is coldest when the Meanwell power supply is set to 12.3V (which is the minimum) rather than 19.2V which is the maximum. I thought that the more power I sent to the peltier, the cooler it would make the CPU.

2. I couldn't get higher than 2.65GHz which I could already do just on water (although temps would get up to 45C). I even tried 1.8V Vcore but it still didn't pass occt. I think the problem is that it's still hitting 28C at load. Does this mean I have to have a better way of cooling the water block? Would a dual heatercore do the trick?

P.S. added my specs in my sig as requested. :)

[n00b 0f l337]

Wait you were putting 19.2volts into the 15volt peltier? That would be why. You won't get any more power out of it, and if it survives the treatment, the efficiency would be terrible.

If you were to lower voltage more than the minimum your load temps would most likely get better to a point were the tec cant hold the cpu.

[Revivalist]

Wait you were putting 19.2volts into the 15volt peltier? That would be why. You won't get any more power out of it, and if it survives the treatment, the efficiency would be terrible.You're right. I was just experimenting through the range of voltages the meanwell can do to see how temps were affected. Now that I think about it, I probably shouldn't have risked putting so much voltage across the pelt.

If you were to lower voltage more than the minimum your load temps would most likely get better to a point were the tec cant hold the cpu.I'm don't think I follow what you're saying here.

[n00b 0f l337]

Running a pelt at full voltage will give u a lower idle temp, but the efficiency is so poor the load will suck.

[Revivalist]

Running a pelt at full voltage will give u a lower idle temp, but the efficiency is so poor the load will suck.Ok, so right now I found that my lowest idle temp is achieved when the Meanwell is at it's lowest voltage, 12.3V. . . So I guess that would be the best setting at load also, right? . . . I'll try testing it.

[Revivalist]

In the meantime, here's a few more pics, just for fun :) . . .


The side -- I put the Meanwell at the very top there.
http://img518.imageshack.us/img518/3199/16finishedside8ep.th.jpg (http://img518.imageshack.us/my.php?image=16finishedside8ep.jpg)


The back -- I bore out holes for the Meanwell fans. Not the cleanest holes but they'll do the job.
http://img518.imageshack.us/img518/2833/15finishedback5al.th.jpg (http://img518.imageshack.us/my.php?image=15finishedback5al.jpg)


The front -- I installed the 5.25" bay intake fans to help keep the Meanwell cool behind them.
http://img518.imageshack.us/img518/7528/14finishedfront6pq.th.jpg (http://img518.imageshack.us/my.php?image=14finishedfront6pq.jpg)


The inside -- I took the UV acrylic plate from my old Storm and put it over the Maze 4-1. (I was stoked that it actually fit!)
http://img518.imageshack.us/img518/3770/17finishedinside9yl.th.jpg (http://img518.imageshack.us/my.php?image=17finishedinside9yl.jpg)


Some UV action.
http://img518.imageshack.us/img518/4299/18finishedinside7yf.th.jpg (http://img518.imageshack.us/my.php?image=18finishedinside7yf.jpg)


A little more UV action.
http://img518.imageshack.us/img518/7130/19finishedinside7ym.th.jpg (http://img518.imageshack.us/my.php?image=19finishedinside7ym.jpg)

[Premmer]

Ok, I'm up and running . . . Motherboard Monitor reads the following:

-9C at idle (underclocked to 1.06GHz and undervolted to 1.07V)

28C at full load for 20 min (overclocked to 2.60GHz and overvolted to 1.70V)

Here's a screenshot at idle . . .

http://img382.imageshack.us/img382/3549/idletempsnegative8c7gx.th.jpg (http://img382.imageshack.us/my.php?image=idletempsnegative8c7gx.jpg)

I was surprised about a few things . . .

1. The CPU is coldest when the Meanwell power supply is set to 12.3V (which is the minimum) rather than 19.2V which is the maximum. I thought that the more power I sent to the peltier, the cooler it would make the CPU.

2. I couldn't get higher than 2.65GHz which I could already do just on water (although temps would get up to 45C). I even tried 1.8V Vcore but it still didn't pass occt. I think the problem is that it's still hitting 28C at load. Does this mean I have to have a better way of cooling the water block? Would a dual heatercore do the trick?

P.S. added my specs in my sig as requested. :)

1. That makes sense. 12V would have been my setpoint using the SP-480-12. Suggest you read this review which contains some important background information on TEC. http://systemcooling.com/mcw5002-64t-03.html

2. Running at 2.65G with a Vcore of 1.7V will yield a thermal heat load of ~135w. I suspect your limited overclock potential is more a function of other Newcastle CPU characteristics rather its operating temperature.

The specs on a CP1-12726 peltier are as follows:
Umax = 15.4V
Imax = 26A
Qcmax = 243.5W
Tmax = 68
R = 0.45+/-0.05 ohms

If you did subject your peltier to a voltage greater than its maximum rating (15.4V), it may be damaged. Even though it appears to be operating to some unknown degree, it may be a walking wounded. If you had the ability to monitor the peltier current draw, you could confirm or reject this possibility.

[n00b 0f l337]

Or you can go read my guide at HWA, its a bit newer...

[Revivalist]

1. That makes sense. 12V would have been my setpoint using the SP-480-12. Suggest you read this review which contains some important background information on TEC. http://systemcooling.com/mcw5002-64t-03.htmlYes, I read that guide before when I was first starting this project. By the way, I’m actually using the SP480-15, not the SP480-12. That’s why I can’t go below 12.3V . . .

2. Running at 2.65G with a Vcore of 1.7V will yield a thermal heat load of ~135w. I suspect your limited overclock potential is more a function of other Newcastle CPU characteristics rather its operating temperature.So are you suspecting that the overclock on this chip is limited by something other than the temperature at this point? What else could limit it?

If you did subject your peltier to a voltage greater than its maximum rating (15.4V), it may be damaged. Even though it appears to be operating to some unknown degree, it may be a walking wounded. If you had the ability to monitor the peltier current draw, you could confirm or reject this possibility.I really hope the pelt isn't a "walking wounded." Is there a way to monitor the current to it? I do have a volt meter.

A. Have you ruled out a poor block mount?
B. You are using a BIX III, correct?
C. What is the CFM rating of the fans?
D. Do you know what the water temperature is?
E. What is ambient temperature?A. I’m pretty sure that the block is mounted well since I’m able to run at -7C idle.
B. Yes, I am using a BIX III. And there is nothing else on the water loop at the moment.
C. The fans are rated at 80cfm. But they are slightly undervolted because the rheostat they’re on gives them 11.7V max. But I would think they’re still pumping out about 75cfm.
D. The water temperature is between 30-34C.
E. The ambient temperature is 26C.

Thanks for any help.

@ n00b 0f l337 – Do you have a link to your guide?

[n00b 0f l337]

Yes as a matter of fact I do!
http://www.hardwareasylum.com/forums/showthread.php?t=891

[Revivalist]

Yes as a matter of fact I do!
http://www.hardwareasylum.com/forums/showthread.php?t=891Thanks for the link. . . I actually checked out that guide as well. Very comprehensive and helpful. Thanks!

[Premmer]

Revivalist:

If you’re willing to collect some data, suggest you do the following.

Set the CPU operating frequency to 2.2G and Vcore at 1.50V. These are the default settings for a Newcastle 3400. At these settings, the processor thermal load is rated at 89W.

Run the fans at maximum voltage. Set the TEC power supply output voltage to minimum.

Run Prime95 using the “in-place large FFTs” selection for ~30 minutes.

Just prior to the end of Prime95 run duration, record the CPU, water and room temperatures along with the TEC supply voltage.

Let your system idle for ~ 30 minutes then record the CPU, water and room temperatures along with the TEC supply voltage.

My reasoning for this test is to simulate a low thermal load (89W) to see how well your system can handle it. The most important measurement would be your CPU load temperature.

Post your results.

As for my reference to the SP-480-12, this was not an error. Your selection of the SP-480-15 was not the best. Ideally you want to operate the TEC at 70% to 80% of its maximum voltage rating.

As for your 320W peltier rating, its quite misleading. Based on Thermal Enterprises documentation for the CP1-12726, at 12V your current draw is ~ 21 to 22A and your cooling power is ~ 160 to 180W, not much more than a 226W peltier operating at 12V.

An S-300-12 at a cost of $86 + shipping would have been adequate and less costly.

The only way for you to monitor the current is to buy an amp meter. They are reasonably cheap. Link to source for amp meter w/shunt:
http://cgi.ebay.com/NEW-DC-0-30A-Digital-LED-Amp-Current-meter-Shunt_W0QQitemZ7564884112QQcategoryZ25411QQrdZ1QQc mdZViewItem

[Revivalist]

Thanks Premmer, I'll try what you mentioned. . . By the way, I didn't really understand what we'll be deteremining from this test. . . .

I got the 480W because I was hoping to power a GPU pelt from it as well later on. And I got the 15V because I didn't know that it's better to run the pelt so far under its max. Right now at 12.3V it's at about 82% of it max, so I guess it's fairly close to the range you mentioned. . . I wish I had your advice before I purchased the Meanwell since it seems that it wasn't the best choice. . .

Do you think there would is a way to limit the voltage that that pelt is getting, such as by having some kind of resistor load between the pelt and the power supply?

By the way, how did you get those specs on the pelt? That's very interesting.

[Premmer]

Thanks Premmer, I'll try what you mentioned. . . By the way, I didn't really understand what we'll be deteremining from this test. . . .

I got the 480W because I was hoping to power a GPU pelt from it as well later on. And I got the 15V because I didn't know that it's better to run the pelt so far under its max. Right now at 12.3V it's at about 82% of it max, so I guess it's fairly close to the range you mentioned. . . I wish I had your advice before I purchased the Meanwell since it seems that it wasn't the best choice. . .

Do you think there would is a way to limit the voltage that that pelt is getting, such as by having some kind of resistor load between the pelt and the power supply?

By the way, how did you get those specs on the pelt? That's very interesting.

If your system performs as well as a Swiftech 226W TEC. Their performance graph is attached below. Note, with a coolant temperature of 32C and a thermal load of 90W, the corresponding coldplate temperature is ~ 16C. Your CPU load temperature most likely would be 10 to 20C higher than your coldplate. On my system with an estimated CPU/pump heat load of 143W under maximum load, coldplate is at -14.6C and CPU is at 5C.

12.3V/15.4V = ~80%; close enough.

It's not practical to use a series voltage dropping resistor.

Got spec package from Thermal Enterprise.

[Revivalist]

Ok, I finished running Prime95 for about 45min with the fans on full blast and the Meanwell at its minimum of 12.3V. The temps I got were CPU: 9C, Water: 33C, Room: 25C. . . Matching up with that graph, it looks like I’m doing ok. . . After sitting idle for about 30min., I got CPU: -6C, Water: 31.5C, Room: about 25C. . . . Sound reasonable?

Thank you for all of your help thus far. It is much appreciated. . . . I hope you wouldn’t mind if I ask for your advice/input on a few more questions? . . .

1. Do you think it would be worthwhile to sell/exchange this Meanwell and get the 480-12? I’m wondering if I would get significantly better performance if I could drop the voltage/current lower. I was just reminded of this post (http://pcpitstop.invisionzone.com/index.php?showtopic=103252&view=findpost&p=1063432) where it was recommended that I use around 9.2V for a 226W pelt. . . . But at the same time I’m rather confused as I read over again the guide you posted earlier. In this page (http://systemcooling.com/mcw5002-64t-10.html) it seems that the reviewer is getting better temps the higher his voltage is until he is unable to go past 14V. That was the reason I bought the 15V Meanwell. But this seems contrary to the idea of using only 70-80% of the max power. . . How do we reconcile this? Is it because he is using two heater cores and thus is dissipating the heat much better?

2. Do you think that adding my video card with a 110W pelt would drastically harm the temps on my CPU? If so, would switching to this dual heater core (http://www.dangerdenstore.com/product.php?productid=48&cat=14&page=1) be better than my BIX III (or perhaps having both on the loop if the pump can handle it)?

3. Regarding the ammeter, does it require disconnecting the wires to the pelt to use it, or can it be used like the voltmeter by just touching the terminals of the power supply?

4. Is this (http://new-electronics.gr/peltier/12726-12730.pdf) the spec sheet that you were referencing for the peltier?

Thanks again for all your help!

[Premmer]

As compared to the Swiftech documentation, your loaded CPU and water temperature in a 25C ambient appear to be quite reasonable to me along with your idle to load delta.

1. No I would not switch to a SP-480-12 at this point. I doubt that there is much, if any, performance gain in the form of a lower CPU load temperature by decreasing the voltage. There may be some gain in lowering your coolant temperature with additional cooling by changing your (3) fans to ~104CFM or lowering your room temperature if that’s practical or increasing the voltage somewhat if your coolant system is capable of dissipating the additional heat generated by a higher peltier voltage. You could run a simple experiment by increasing the voltage in, let say, 0.5V increments up to15V while holding all other variables as constant as possible. You may experience a lower CPU temperature before the peltier flat lines or heads south in performance.

More radiator surface with adequate coolant circulation and air flow is how I would reconcile it.

2. If your determined to TEC your GPU, I would setup a separate cooling loop to be on the safe side.

3. The amp meter w/ shunt that I suggested is hard-wired in series with the peltier. It is quite straight forward to install. The in-series connection is implemented at the PS end of the circuit.

4. Yes

Let me through out a few thoughts on your lower CPU temperature quest. Lower temps will not result in a larger OC if the CPU is a dude. I have no knowledge of the Newcastle relative to how well they OC. The Venice 3000 and 3200 (S939) do reasonably well and the Opteron (S939) is quite remarkable on air and water unless you get a dude. From recent threads I’ve read on the Opteron, a ~50% OC on air or water is achievable.

On the upside of your power supply purchase, you could power a 437W peltier at 19V if you decided you needed more cooling power. The Vmax is 26.7V.

[Revivalist]

As compared to the Swiftech documentation, your loaded CPU and water temperature in a 25C ambient appear to be quite reasonable to me along with your idle to load delta.Sweet!

1. No I would not switch to a SP-480-12 at this point. I doubt that there is much, if any, performance gain in the form of a lower CPU load temperature by decreasing the voltage. There may be some gain in lowering your coolant temperature with additional cooling by changing your (3) fans to ~104CFM or lowering your room temperature if that’s practical or increasing the voltage somewhat if your coolant system is capable of dissipating the additional heat generated by a higher peltier voltage. You could run a simple experiment by increasing the voltage in, let say, 0.5V increments up to15V while holding all other variables as constant as possible. You may experience a lower CPU temperature before the peltier flat lines or heads south in performance.Ok, I'll try that. So far, I tried this with idle temps and found that 12.3V (the minimum voltage was best). I figured that with a load it would be even harder to dissipate the heat so more voltage would be worse. . . But I should try it just to make sure.

2. If your determined to TEC your GPU, I would setup a separate cooling loop to be on the safe side.The problem is I don't really have anymore room in my case (for another pump) or outside my case (for another radiator). So I was hoping I might be able to pull it off with the current water loop. I know it's hard to predict, but do you think that the GPU pelt would drastically harm the CPU temps? Is there a way to calculate that?

3. The amp meter w/ shunt that I suggested is hard-wired in series with the peltier. It is quite straight forward to install. The in-series connection is implemented at the PS end of the circuit.Interesting. I might try that. Speaking of a shunt, do you think it would be worth using a shunt resistor just to check if a lower voltage would indeed let this pelt run better. I was thinking that it may not be a permanent solution, but it could be a good test.

On the upside of your power supply purchase, you could power a 437W peltier at 19V if you decided you needed more cooling power. The Vmax is 26.7V.Great idea! . . . But would 19V be too low for a 437W pelt? Also, where could I get one from? . . . O wait, one more thing, if I run my Meanwell at 19V, then I suppose that wouldn't leave any juice left for a GPU pelt. (And there probably wouldn't be a GPU pelt that I could run at 19V anyway since this PSU can only put out one voltage).

Thanks again for all the wonderful help!

Edit: I just found this 437W peltier (http://www.pcextreme.com/catalog/sku/exp-04.html) but it's 62mm x 62mm. How can that be properly cooled inside a Maze 4-1?

[Premmer]

If your quite sure that increasing the voltage will drive the CPU temp higher then there may be some gain in lowering your coolant temperature with additional cooling by changing your (3) fans to ~104CFM or lowering your room temperature if that’s practical. After reviewing the cooling power vs. current draw characteristic curves for this peltier, once you exceed ~ 12V, the cooling power increases at a rate ~ 2.75 greater than it does between the 8V to 12V range. In summary, if you do decide to increase the voltage above your 12.3V minimum, it probably should be done in ~ 0.25V increments.

The series dropping resistor required to drop lets say 1.0V so the peltier was operating on 11.3V would be quite small < 0.1 ohms and the wattage rating would be quite high. If you could source it, my guess is that it would be quite expensive for an experiment; again, not practical at these ohmic values and wattage ratings.

You should be able to create ~1v drop by inserting a 20ft. length of No.14 gauge stranded wire in series with one of the peltier power leads. Remove one of the wires powering the peltier at the PS terminal strip. Fasten one end of the 20ft. length to that terminal. Solder the other end of the 20ft. length to the peltier power wire removed. When you power up the system, use your DMM and measure the voltage between the ends of the 20ft. length of wire. Subtract this value from 12.3V and result will be your peltier voltage. Add an additional 10ft. or 20ft. length if previous temp results indicate a positive gain, lower CPU temp. A voltage drop calculator program is available via the link noted below. Suggest you use No. 14 gauge and 20 to 22A as your input parameters.
http://www.stealth316.com/2-wire-resistance.htm

As for the TECed GPU on the same cooling loop, I’m going to have to defer a response to other forum individuals that can hopefully reply based on their own experiences.

I run my 437W peltier at 19.9V. At higher voltages my coolant temp starts to rise which drives my CPU temp higher. As of last week, Wintsch Labs had the Artic Web in stock. I believe there are people running a GPU peltier off of a shared meaty PC power supply.

When I mentioned the 437W, I should have just referred to Wintsch Labs Artic Web TEC CPU Assembly. Unless you go custom build, it’s the only choice as far as I know.

[n00b 0f l337]

I geuss following that last sentence, anyone who would like to work out some custom tec blocks, send me a PM if your interested in just generally desiging and thinking up crap. Ive got nothing to do. But please, have a clue to what your thinking bout.

And yes. The 437's (rev2 dominator blocks) at 19.9v are very nice as its closer to there sweet spot. Try lowering the voltage more if you can.

[Bloody_Sorcerer]

actually, there are no choices in the 62x62mm block arena now; wintsch labs closed its doors recently.
Using large coils of wire like that is a very bad way to drop voltage; it could generate electromagnetic fields (think electromagnet) which could royally eff your stuff.
basically your best solution is either more fannage or more radiatorage or both.

[Premmer]

actually, there are no choices in the 62x62mm block arena now; wintsch labs closed its doors recently.
Using large coils of wire like that is a very bad way to drop voltage; it could generate electromagnetic fields (think electromagnet) which could royally eff your stuff.
basically your best solution is either more fannage or more radiatorage or both.

The wire length is for experimental purposes only in a DC power circuit. Please explain means of generating EMF in this application.

As of last week, Devin still had some stock of the Artic Web CPU Assembly.

[Bloody_Sorcerer]

ah, if its for testing purposes only and you have that much wire laying around... i guess it'd work.

[n00b 0f l337]

Yes but keep it uncoiled. EMF's are a :banana::banana::banana::banana::banana:. And Bloody. There are no blocks being made....

For now.... You know that ;)

[Revivalist]

Ok, I've confirmed that I can't get better results by running the TEC at higher voltages. . . .

The following comparisons where done with a room temperature of 25C and the radiator fans set to maximum. Prime 95 was set to “in-place large FFTs” and was running at for at least 15min. before temps were recorded.

Voltage: 12.3V
• CPU: 19-20C
• Water: 33.0C

Voltage: 13.0V
• CPU: 20C
• Water: 33.3C

Voltage: 13.5V
• CPU: 20C
• Water: 33.9C

Voltage: 14.0V
• CPU: 20C-21C
• Water: 34.9C

I didn't bother to go any further. It seems that the best possible voltage for the Meanwell 480-15 running a 320W peltier is 12.3V (which is the lowest it can go.) Too bad I can't run it lower to see how much better it can do. . .

By the way, I still don't understand how to reconcile this with the data from this review (http://systemcooling.com/mcw5002-64t-10.html). How come the reviewer was getting better results at 14V than at 12V?

[n00b 0f l337]

Who knows. I dont trust on desk reviews. I trust in a case reviews were most people use there actual comps.

Premmer, you have a PM.

[Revivalist]

Ok, so I was doing the calculations for the variable resistor that I would need to put in series with the TEC in order to drop the voltage across it. . .

First off, using the max voltage and max current of the TEC, I know the resistance will be about 15V/26A = 0.577 ohms. (I know that the TEC isn't perfectly linear, but I'll use this as a rough estimate.)

I also know that the total resistance in the circiut (which is the resistor R plus the TEC) multiplied by the current (I) must equal the voltage of the Meanwell. At the lowest setting this would be 12.3V. So I have

(R + 0.577ohms) * I = 12.3V

Solving for R gives R = (12.3V / I) - 0.577ohms.

The lowest reasonable voltage to test on the TEC is probably 9V so the current would be I = (9V / 15V) * 26A = 15.6A.

Then plugging in the current I=15.6A into the equation above for R will give R=0.211ohms. And the power it has to dissipate is P=(I^2)*R = (15.6A)^2 * 0.211ohms = 51W.

This is in perfect agreement with Premmer. So I need a variable resistor capable of dissipating 50W with a resistance of < 1ohm. . . .

So far I only found this one (http://www.digikey.com/scripts/dksearch/dksus.dll?Filter) and this one (http://www.newarkinone.com/NewarkWebCommerce/newark/en_US/endecaSearch/partDetail.jsp;jsessionid=0KRJP0MVTMENGCXFEOFCFGIK 2URYWIV1?SKU=93B8546&N=0), but they're about $30! :( That's just not realistic for a single rheostat for testing.

[Premmer]

Ok, I've confirmed that I can't get better results by running the TEC at higher voltages. . . .

The following comparisons where done with a room temperature of 25C and the radiator fans set to maximum. Prime 95 was set to “in-place large FFTs” and was running at for at least 15min. before temps were recorded.

Voltage: 12.3V
• CPU: 19-20C
• Water: 33.0C

Voltage: 13.0V
• CPU: 20C
• Water: 33.3C

Voltage: 13.5V
• CPU: 20C
• Water: 33.9C

Voltage: 14.0V
• CPU: 20C-21C
• Water: 34.9C

I didn't bother to go any further. It seems that the best possible voltage for the Meanwell 480-15 running a 320W peltier is 12.3V (which is the lowest it can go.) Too bad I can't run it lower to see how much better it can do. . .

By the way, I still don't understand how to reconcile this with the data from this review (http://systemcooling.com/mcw5002-64t-10.html). How come the reviewer was getting better results at 14V than at 12V?
The results you noted above have little value unless you qualify them with:
- CPU Frequency
- Vcore Setting
- Estimated CPU Thermal Load http://benchtest.com/calc.html

Paste in the review statement(s)/data that need to be reconciled.

Also, refer to Page 2 of the Thermal Enterprises documentation on the CP1-12726. The graph in the upper left hand corner plots Qc(W) vs dT(C). Look at the intersection point of the I=26A plot with the y-axis scaling of Qc. The value indicated validates the Qcmax noted on Page 1 of 243.5W. Why do people continue to refer to this peltier as a 320W device? It's hype. The Qcmax rating is as stated, 244W in round numbers.

[Banzai]

I've only read a bit more since my last post. So I hope this hasn't already been said.

But to me it looks like you need some more compression on the waterblock/pelt.

Ryan

[Revivalist]

The results you noted above have little value unless you qualify them with:
- CPU Frequency
- Vcore Setting
- Estimated CPU Thermal Load http://benchtest.com/calc.html

Paste in the review statement(s)/data that need to be reconciled.

Also, refer to Page 2 of the Thermal Enterprises documentation on the CP1-12726. The graph in the upper left hand corner plots Qc(W) vs dT(C). Look at the intersection point of the I=26A plot with the y-axis scaling of Qc. The value indicated validates the Qcmax noted on Page 1 of 243.5W. Why do people continue to refer to this peltier as a 320W device? It's hype. The Qcmax rating is as stated, 244W in round numbers.Sorry about that. Those value are . . .

CPU Frequency: 2.60 GHz
Vcore Setting: 1.70V
Estimated CPU Thermal Load: 135W (thanks for the link)


The data that needs to be reconciled is the following statement:
"Doubling the input power (151 watts at 10.0 VDC to 294 watts at 14.0 VDC) only resulted in a 3.3ºC drop in cold plate and thermal die temperatures."

The reason I was using the name "320W pelt" is because that is how Thermal Enterprises refers to it. I noticed that they refer to it as a "245/320W pelt." I see that the 245W matches up with the rating from the documentation as you mentioned. So where does the 320W come from? How do they justify the hype?

[Premmer]

Sorry about that. Those value are . . .

CPU Frequency: 2.60 GHz
Vcore Setting: 1.70V
Estimated CPU Thermal Load: 95W (thanks for the link)


The data that needs to be reconciled is the following statement:
"Doubling the input power (151 watts at 10.0 VDC to 294 watts at 14.0 VDC) only resulted in a 3.3ºC drop in cold plate and thermal die temperatures."

The reason I was using the name "320W pelt" is because that is how Thermal Enterprises refers to it. I noticed that they refer to it as a "245/320W pelt." I see that the 245W matches up with the rating from the documentation as you mentioned. So where does the 320W come from? How do they justify the hype?
For the Newcastle I used the following:
Stock CPU Frequency: 2.20 GHz
Stock Vcore Setting: 1.50V
Stock CPU Thermal Load: 89W

For the overclocked parameters I used:
CPU Frequency: 2.60 GHz
Vcore Setting: 1.70V
My results were 135W. Am I in error?

A decrease of 3.3C in coldplate temp for a doubling of input power from a 10V level is pretty small.

I was hoping that you would have dropped the peltier voltage by now using the wire length I suggested. 50ft. of No.14 probably would'nt cost more than $5. It would be an interesting experiment. 40ft. should produce ~ a 2.3V drop.

[Revivalist]

@ Banzai - Thanks for the input. . . . Do you really think it's the compression on the pelt? . . . I mean I turned those screws until they were either about to strip or I started to see the copper of the waterblock bulging out. I think any tighter and I might have harmed something.

@ Premmer - Oops, yeah your right about the thermal load calculation. It should be 135W. (I changed it above.)

About that article, I realize that it's certainly not efficient to double the input power in order to drop 3.3C more on the coldplate. But my question is why that even worked. I thought that he would be far out of the "sweet spot" of the TEC by doing that?

About the wire experiment, I see that 40ft. of 14 gauge wire would be about 0.103ohms of resistance. The pelt is about 0.577ohms. Together that's 0.68ohms. If the Meanwell is set at 12.3V then the current will be 12.3V/0.68ohms = 18.1A. So the wire will give a voltage drop of 18.1A * 0.103ohms = 1.86V and the pelt will be at 12.3V-1.86V = 10.44V.

Wouldn't it be better to be able to go lower than that on the pelt, say to 9V? If so, that would require a resistance of 0.22ohms (which I calculated earlier). That would give a current of 15.45A and the voltage across the pelt would be about 9V. And I guess I would need 85ft. of 14 gauge wire for that. . .

Sound right?

I'll try the actual experiment as soon as I have more time.

Thanks again for all the excellent advise.

[Premmer]

Referring to the sysytemcooling results, it was a die simulation using a 66.6w thermal load with the following hardware:
Iwaki MD-15R pump
(2) 6”x 6”x 2” heatercores with fan shrouds on both sides
(4) Delta 120mm SHE, 152 CFM fans in push/pull @ 8 VDC

Given the cooling/circulation capacity along with the sizeable thermal load differences between their system and yours, not to mention that they used lab grade measurement instrumentation, is how I reconcile the results you experienced. They were able to dissipate the additional heat created by the increases in power and still realize a meaningful gain up to ~13.5V. At 14.0V the rate of change decreased significantly.

Yes, go down to 9V. I’ll put myself out on a limb and project the following ROM results. This assumes an internal case temp of 25C, CPU frequency of 2.60GHz, Vcore at 1.70 and Prime95 (max heat) stressed for at least 30 minutes.

Tabulations below staring at the extreme left are as follows:
peltier voltage, peltier current, water temp(C), coldplate temp(C) and CPU temp(C).

12.3, 22.2, 33.0, 4.6, 19.5
11.0, 20.2, 32.0, 6.0, 21.2
10.0, 18.6, 31.3, 7.8, 23.2
9.0, 16.9, 30.7, 10.4, 26.4

Now it’s up to you to buy that wire, schedule some time and be as meticulous as possible in tabulating your results and holding all other variables constant.

[Revivalist]

Wow Premmer! You never seize to amaze me . . . How in the world did you come up with those detailed predictions?

So are you essentially predicting that I’m already at the “sweet spot” at 12.3V since the temps would be higher at lower voltages and I already know that they are also higher at higher voltages?

By the way, thanks for the response about the systemcooling results. So they basically had a lower CPU thermal load to deal with and a better heat dissipation system. That make sense.

[RussC]

Hi,
Just a comment, your results are good for a TEC, just be carfuls running real low temps for extended periods. Condensation will be a real problem if you don't properly insulate for the lower than ambient temps.

RussC

[Revivalist]

Hi,
Just a comment, your results are good for a TEC, just be carfuls running real low temps for extended periods. Condensation will be a real problem if you don't properly insulate for the lower than ambient temps.

RussCThanks Russ. I used this entire insulation kit from DangerDen. The CPU socket was oozing over with dielectric grease, the CPU and socket were completely surrounded by conformal coating on the top and bottom of the board, and 4 neoprene pads were used (around cold plate, socket, CPU, and behind board). I figure that should be enough, right? :)

[Holst]

Thanks Russ. I used this entire insulation kit from DangerDen. The CPU socket was oozing over with dielectric grease, the CPU and socket were completely surrounded by conformal coating on the top and bottom of the board, and 4 neoprene pads were used (around cold plate, socket, CPU, and behind board). I figure that should be enough, right? :)

Have you checked that your getting a good contact with your CPU.

Insulation often messes up your waterblock/coldplate mount on the CPU.

Ive had this problem several times myself when I try and overinsulate and it gets in the way.

[Holst]

I also recon that your problem may your maze4 waterblock.
Its just not got enough surface area for TEC usage IMO.. I think thats why your temps are worse. Although your water is not too hot I think the hot side of the TEC gets warmer as the block isnt able to cope with the extra wattage ... this increase in hot side temp counteracts any gains from the higher voltage.

They (maze4) are poor for TEC use IMO and I recomend changing to a winchlabs block (if you can get one) or a swiftech block (which should perform better but has some aluminium annodised components)

[n00b 0f l337]

Also adding onto Holst's statement. Ive found the maze4 to be terrible with pelts as it also cannot provide even cooling nor a good clamp.

[FireDragon]

i didn't think the maz 4-1 wouldfit a 50mmx50mm tec?!?!?! if that is the case screw biulding one this is much easer to buy one...btw great job

Dragon