MCWChill Review Progress
 

Ok my plans to just solder up another CPU and use it for some system testing are all falling to pieces. Over the weekend I killed a 1700+ TBredB, my Epox 8K7A+ that I repaired a few months ago seems to be dead now, and a 2000+ XP I borrowed from KnightElite to solder up and test with was DOA when I tried it this evening. Ever have a bad run of luck?

Sooooo I really want to test out a Swiftech water chiller and complete water cooling kit that Swiftech was kind enough to send me. I happen to have a motherboard that reads from the AMD diode (Epox 8K3A), and what I feel is a reasonable approach. My plan is to epoxy one of the DOW sensors underneath the center of my remaining 1700+ Tbred, make DOW probes that can go into water, and then to test by logging in MBM the CPU diode, CPU under, ambient air, radiator exhaust, and waterblock inlet temperatures to evaluate the water cooling kit. For the water chiller I think that measuring the chiller water in and out temps would be useful. This isn't ideal guys but it's the best I can do with budget constraints. I still haven't purchased the DMMs needed for measuring current and voltage and the insulation for my die simulator so that project isn't really even close to done.

The DOW sensors aren't bad; +/- 0.5C guaranteed and 0.125C resolution. With my current luck I'll end up not sealing them properly and short out something that way.

When I get the money I'll pick up another old AMD motherboard and a CPU I can solder on again.

Waterchiller.
Testing stand-alone? (Pump,tubing/fitting,measuring equip)
Closed loop or water to heaven?
Inlet and Outlet Temps. Additionally Flow-Rate would be interesting.
Varying TEC Voltage? ( Possibly the most "User Important" investigation - Peltier Efficieny, as defined by COP, increases with decreasing Voltage http://www.jr001b4751.pwp.blueyonder.co.uk/COPa.jpg )

Could compare with a Kryotherm produced Family of curves.Family cos Kryotherm does not compute for 220W Peltiers.
Any numbers, with which to play, would be appreciated.

Before I discovered Kryotherm, I setup a spreadsheet to calculate performance of any TEC based on the commonly used specs:

Qmax
dTmax
Vmax
Imax

It's very crude in comparison to Kryotherm and comparing my spreadsheet's results to Kryotherm's for a 172 W TEC generated errors as high as 5C iirc. (Although that error was at an operating point far away from typical.) It seems to give fairly accurate results under the operating conditions typically used in CPU cooling.

If you are interested in playing with it, let me know. (I had attached it to a post at O/C'ers, but they have gotten rid of attachments since.)

Sean
Also cannot find the attachment.
That is if it was in this thread http://forum.oc-forums.com/vb/showth...hreadid=137660.
Please send .
All
The above thread is high quality and well worth a read.

I'll attach the spreadsheet to a post later tonight so anyone interested can download it.

I'd also be interested in the low voltage operation of the Swiftech chiller. (Perhaps connect the two TEC's in series to a 12V supply to get 6V per TEC.)

The problem is that between the two of us, we have one 12V power supply and one 13.8V power supply... both of which are not particularly variable in their voltage (my 13.8V is not variable at all).

I think pH mentionned using ATX PSUs and connecting them up in parallel, but I'm not sure if we will do that.

Obviously though, 6V is easy to achieve, because all we would have to do is connect the TECs in series.

Connecting the TEC's in series to the 13.8 Volt supply would be good too. Off the top of my head, I'd guess you'd see a loaded CPU temp near ambient, with water temps near dewpoint for a typical climate controlled home.

This is an operating region I'm particularly interested in.

"you'd see a loaded CPU temp near ambient"
"with water temps near dewpoint for a typical climate controlled home"
all depends on the rads (hot side cooling)
both probably a bit (~10°C) below ambient if the rads are effective and the cold side lines short and well insulated

I was at 21°C with a coolant temp of 35°C and an actual load of 80W, and NO insulation
- since NO (present) cpu is putting 80W into the wb, expect your temps to be lower if you power the TECs at 12V

Vmax on the TECs is 15.4V, should be ok at 13.5 (87%)
more efficient at 12V though

"Both" about 10C below ambient?

I'm lost. Coldside coolant and...?

Is this 21C at the die simulator with 6V to each TEC?

EDIT - disregard the 'both' comment
the cpu should be ~ 15°C, the cold side coolant ~ +1°C
yes, there will be lots of condensation

my tests were running the TECs at 12V

Is this review just for swiftech, or can it turn into a TEC block comparison. I have always wanted to know if the DDen4 was a good TEC cooler as I have one.... Once you test this, could this be your test bench, then you just swap out the block using the same pelts, power, chiller etc.. I would like to see how this and other TEC blocks handle the setup. I have yet to see any TEC block roundup on the net, and I think this sounds like a GOLDEN opportunity for me to find some really good information.
Perhaps others can donate a TEC block?

I may take you up on that winewood WHEN I get my die simulator insulated and running. Probably have to sink a few dollars into that yet though.

Trying to "make do" for these system tests so that I can get reviews turned around in reasonable amount of time. Eager to actually see some numbers from the chiller...

How will you know if that chiller is good compared to others? It may be a great chiller, but do you have anything as a real world benchmark? Or perhaps you can get that data off another sites specs? I was wondering, as I read numbers, but since I know nothing about chillers... what is really good for the money?

Im ashamed to ask since I know SO little about them. :|

I can't wait for this one pH.. the numbers and the actual review will open up some excellent doors on some numbers that aren't out there. Hopefully you can answer some of those after the review.

pH, what kind of power supplies do you have available? If we can get some 24V action going, we can get some really beefy cooling with my new 48V fans ;).

On the topic of testing TEC blocks, would they be measured with a negative C/W? I'm sure this is a dumb question, but I was just curious. I would figure that the C/W value might not be good for anything but what you can achieve at that wattage the CPU dissipates, as the TECs effeciency changes based on temps, and as you inched closer to the maximum amount of heat the TEC can pump.

Did any of that make any sense? :shrug:

Thanks,

BO(V)BZ

KE:

I have a 25V 12V PSU and several 18-20A 12v rail ATX psus. I have 4 of the Papst 105CFM 120mm fans so I think we should be "ok" as far as cooling goes.

I haven't tested TEC blocks before, but my instinct would be to drill a hole into the coldplate for a temperature probe and then measure delta T water to cold plate and deltaT coldplate to die.

Here's the spreadsheet I mentioned. Jeesh it's crude. Shows me how much I've learned in the last year. Still it's somewhat useful for doing things Kryotherm won't.

Hopefully anyone interested will be able to use that URL. IE opens the spreadsheet inside a browser window for me I don't know if other browsers will. For some reason I can't post attachments anymore, otherwise I would have attached a 'zip' that anyone could download.

Just adding a thought,
You can right click on your link and "save target as" if you want to DL it. at least in IE you can.
Aardil

I am going to pick up a 2500+ Barton from the local store if they are in stock for my main box. This will free up the 1700+ Tbred for testing with the chiller and water cooling setups tomorrow. I will have to wait a bit for the thermal epoxy to dry but should otherwise be ready for action tomorrow afternoon KE. I didn't get ANYTHING done (was grading papers) so it will be a long evening/night of setup and tests. Should be interesting though.

Indeed. I'll drop by your office or call you tomorrow morning sometime to determine specifics as to when we're starting, etc...

Ph,

Would a ATX power supply that has adjustable pots be any help?

And it has a 34 amp 12v rail.

BE

We should be good as to power supplies I think, and since we're doing testing tonight.... ;).

Opps, didn't relize the time frame involved.

:)

Some results
 

KnightElite and I spent all evening setting up the MCWChill. Two hotside loops cooled by large single pass heatercores with 2 120mm Papst 100CFM fans each. With peltiers at 12V and 13.8V (these were the PSUs we had) water temperatures quickly (within a minute) start to dramatically drop. For testing we used a MCW5000-a waterblock, an Eheim 1048, and 3/8" ID tubing and a ball valve to throttle pump.

CPU: TBredB 1700+ @ 2084MHz and 2.00V

Diode temp from epox 8k3a+ motherboard: 12 ºC
Room temperature 23 ºC (but rising)
water temps measured on inlet and outlet side of chiller were 1.90 ºC and 3.45 ºC (with Digitecs)

Some pics of the monstrosity:

http://www.procooling.com/~phaestus/.../mcwchill1.jpg
http://www.procooling.com/~phaestus/.../mcwchill2.jpg
http://www.procooling.com/~phaestus/.../mcwchill3.jpg

My plan was to use the pressure gauge to estimate flow rates, but the battery is dead. Will get a new 9V tomorrow and then update this.

I will play around with the unit for a week or two more as long as Bill doesn't mind and keep this thread updated as more useful (temp vs flow/temp vs voltage) numbers are produced.

Re: Some results
 

Hot-side water Temps would be nice.Shame to miss if(?) only involves a minimal effort.
Otherwise have to estimate Heater -core performance to get an estimate of Chiller performance.
My guess for perfomance of Heater-cores would be 0.02 - 0.08c/w per Heater Core giving Hot-side Water Temp 6 - 24 c(~300w per Heater-Core*) above Room Temp - not a very satisfactorily estimate.

* Used Since87 Spreadsheet for 220w TEC power dissipated( ~250w per TEC) then added added "CPU + Insulation Gains" guess(50w per TEC).
In limited play found Since87's Spreedsheet to:
1) Give reasoable agreement with my own crude VA measurements of an in service 220w Peltier.
2) Give "heat pumping" results similar to "Pefectly Insulated" Kryotherm sums for a 172W TEC.

Sean, I think it is both a neat and useful tool.. Guesses at the enigmatic "Insulation Losses/Gains" can be added using Kryotherm.