Mindchill vs Krow: All-In-One Liquid Chiller Build Off!

[KrowniK]

Mindchill vs Krow: All-in-One Liquid Chiller!

Sponsored by:

http://www.arqtik.com/

Mindchill and I have been working together for a while to bring you awesome cooling solutions for you computers. We figured we wanted to design an All-In-One Chiller, but we just couldn't seem to come up with anything cheap enough, with enough power; until now. This thread will host the "Head-to-Head TEC Liquid Chiller" and contain all costs, specifications, cooler benchmarks, build logs, and whatever else is needed.

In the end, testing will be performed to compare the analytics of this contest to determine the winner. Criteria such as:

1. Performance
2. Noise
3. Power Used

The processor being tested will be the Intel Core i5 2500K (4.5GHz). The overclock isn't too demanding to ensure that we're both able to reach it (as we're going to be testing in two different locations), but whoever scores the lowest temperatures will win the performance category. We may go for a higher overclock to load the chillers but we won't be comparing results of overclocks if one cannot obtain the overclock. To cancel out differences in ambient temperatures, we're going be comparing deltas instead of actual chiller temperatures. But all the information will be contained in the posts for your own reference.

Rules:

1. Must be All-In-One:
   
   • Contain Inlet/Outlet to and from CPU
   • Must contain all the necessary cooling equipment (Heat Exchangers, Pumps, Power Supply, etc...) and only require a waterblock on the CPU.
2. Must be able to handle the load of an overclocked Intel Core i5 2500K
3. Must be under $500 to build.
4. Must be built with parts from retailers or wholesalers to ensure continuing production if a winner is chosen.

A special point system has been created to fairly gauge the winner, and will be unbiased.

When all is said and done, a poll will be taken to see which one you would personally prefer, but it will have no sway on the winner.

We hope you enjoy the show, and will hopefully be done in a few weeks at the most. May the best man win!

[KrowniK]

Design (Google SketchUp):
July 25, 2012

Color Code:

• Bright Blue: Bottom Plate Screw Holes
• Bright Green: LGA1366 Mounting Screw Holes
• Bright Red: Gasket Engraving
• Dark Red: Warm Water Flow
• Dark Blue: Cold Water Flow

This design is going to use the 4x Intel STS100C to cool 4x 225Qmax 40mm TECs. The Dimensions are 180mm x 180mm x 97mm.

I think this gives it a nice compact cooler, but it remains to be seen if the heatsinks will exhaust properly when pushing against each other. So far it seems to still work well if I block the back completely, it still exhausts out the top. That's just other testing I'm gonna have to do, when I get all of the heatsinks.

Testing:
July 26, 2012

Ambient was ~24.5°C
Not sure if I'm right, but the 50°C Delta implies on the chart that the surface of the TEC is loaded by about 12W from the surrounding environment.

Low Fan Speed (Barely audible) - Heatsink Thermal Resistance = (40.6° - 24.5°C) / 147W = 0.10952°C/W
Medium Fan Speed (Faint hum) - Thermal Resistance = (35.2°C - 24.5°C) / 147W = 0.072789°C/W
High Fan Speed (Consistant high pitched noise, but not overly loud) - Thermal Resistance = ( 33.2°C - 24.5°C) / 147 = 0.059184°C/W

[mindchill]

Reserved Build Log

[KrowniK]

Reserved for Results

[mindchill]

Reserved Results

[bds71]

looking forward to seeing how this plays out - best of luck to you both!!

[defect9]

We need more threads like this. Good luck to both of you

[Ultrasonic2]

KrowniK design does not meet the challenge requirements ?

And that has pitifully small fans

But that doesn't mean it wont win :-)

one would think Mindchill can't loose this since he has more experience

[Elloquin]

KrowniK design does not meet the challenge requirements ?

And that has pitifully small fans

But that doesn't mean it wont win :-)

one would think Mindchill can't loose this since he has more experience

Those are actually server sinks from the looks of them so 60mmx38ishmm DELTA's are going to be on them. So quiet? hell no lol but they should work. TBH I don't see how an air v.s. water solution can be a fair and level playing field. At any rate since it is going to be a 180mm x 180mm square a 180mm fan on top of the sinks can't hurt.

[KrowniK]

Testing temperatures of the heatsink while running a 135W load. TEC is a 12709 run at 15V, 9A. This isn't to test the capability of the TEC, seeing as this is just a fake load to attempt to figure out the thermal resistance of the Heatsink. The real TEC will result in lower temperatures due it it's increased efficiency at 12V.

Ambient was ~24.5°C
Not sure if I'm right, but the 50°C Delta implies on the chart that the surface of the TEC is loaded by about 12W from the surrounding environment.

Low Fan Speed (Barely audible) - Heatsink Thermal Resistance = (40.6° - 24.5°C) / 147W = 0.10952°C/W


Medium Fan Speed (Faint hum) - Thermal Resistance = (35.2°C - 24.5°C) / 147W = 0.072789°C/W


High Fan Speed (Consistant high pitched noise, but not overly loud) - Thermal Resistance = ( 33.2°C - 24.5°C) / 147 = 0.059184°C/W


Already ordered 3 more of these bad boys to complete the set.

[Ultrasonic2]

can you put an ammeter inline so you call what it's really pulling wattage wise

[KrowniK]

When I get the multi-meter I will. I gotta stop spending money I don't have lol.

[KrowniK]

Ok, so my multimeter finally showed up in the mail today, which has opened the doors to actually testing the heatsinks entirely. Here are my results! I added 10W to the load on the Heatsink for ambient loads.

Low:
Ambient: 23.8°C
Hotside: 46.8°C
Coldside: -6.4°C
Voltage: 14.89V
Current: ~7.26A
Thermal Resistance: 0.19244°C/W
Total Assembly TR: 0.04811°C/W
Thermal Loss of Assembly (200W Load): +9.6°C

Medium:
Ambient: 23.8°C
Hotside: 42.4°C
Coldside: -10°C
Voltage: 14.88V
Current: ~7.36A
Thermal Resistance: 0.15563°C/W
Total Assembly TR: 0.0389075°C/W
Thermal Loss of Assembly (200W Load): +7.8°C

High:
Ambient: 23.8°C
Hotside: 38.4°C
Coldside: -12°C
Voltage: 14.88V
Current: ~7.35A
Thermal Resistance: 0.12216°C/W
Total Assembly TR: 0.03054°C/W
Thermal Loss of Assembly (200W Load): +6.1°C

So the numbers were definitely off from my previous predictions, due to buying the PSU as 15V 350W, it was set to 11.8V, it definitely wasn't up to specifications. It could only reach 14.89V and that was with the PSU cracked to the highest voltage it could muster. So my predictions took a hit but I'm still committed and it will definitely do sub-ambient.

[bds71]

2 weeks - any updates?

[KrowniK]

My preliminary testing has been finished on an "unloaded" source. Due to problems encountered by rushing to get the cut pieces to me, we had to use premade 62mm plates, which are 22mm wider than the TEC. And since they're only about 3mm away from the surface of the heatsink, there is probably a feedback effect where they're absorbing some heat which is loading the TECs slightly. That will subtract from my delta a bit, so I think some insulating is needed on the plates to allow for as minimal interference as I can from the outside world.

All that being said, assuming it is able to move the amount of heat it needs to, this chiller seems to be exactly what I want.

Entire test assembly:




Yes, my reservoir is a Coke Zero bottle because the EK Res is as leaky as toilet paper under a waterfall... I cut holes for an inlet, outlet, and a spot for the Temperature sensor. The water in the bottle started at 37°C. I just ran hot water from the tap. Target temperature means nothing as the controller is not controlling the TECs. I didn't wanna go to all that trouble of wiring when I'm just testing full power settings anyway.

I ran it for 10 minutes (using my iPhone as a stop watch) and the water temperatures were dropped by 20°C. For some reason I didn't grab a picture of the 10 minute mark but it's not unbelievable; the water isn't a constant load and this thing is suppose to have a 30°C for about 200W.

After one hour, the final temperatures for the fans on medium (which are no louder than a graphics card half loaded, or a computer case that isn't designed to be quiet) [Heatsink/Water/Ambient]:



Before I switched the fans on high, just to see what maximum delta I could achieve, the display was changing between 10 and 11. Ambient temperatures did increase slightly because I started with 23.4°C.



I only let the fans run at maximum for 20mins but from previous testing on hot days, it wasn't going to get much better than this. Ambient increased slightly but I think the sensor is just finicky.

The next step is to go get some neoprene to put in between the coldplates and the heatsinks, and then to insulate the pop bottle and possibly the fittings, seeing as I'm using compression fittings that I will never use again. From here on out, its barbs, not only because they're easier to use, there's less metal exposed to the atmosphere and they won't be destroyed by pliers.

Once the tweaking is done, I'll drop by with another update. Enjoy!

[defect9]

is Mindchill gonna pop in with a design concept or do we have to wait until everything is built?

[mindchill]

is Mindchill gonna pop in with a design concept or do we have to wait until everything is built?

He defect I see your close to my old neck of the woods.. Reno Nv.

I decided that I'm going big, because I can really use a 6 tec chiller on my testing bench.. I had krow mock it up in sketchup so I'll have him post it.. My concept can still be downsize to a 2-4-6 tec chiller depending on the load and still use the same acrylic box(s) which wouldn't add that much more cost to the build.

[KrowniK]

So this is Mindchill's latest thought experiment for a benchmark chiller. It will house 6x Modular Liquid Chillers, and contain 3x Triple Radiators and 1x Double Radiator; aswell as be entirely plug-and-play. Fill up the 2 reservoirs after connecting your PC, and it should be able to cool about 900W to 20°C below hot water temps, and 510W to 30°C.

It should be capable of cooling an entire SLi/Crossfire system and a 200W CPU. Theory is one thing, but practice is another. He plans to build this, aswell and in the process test against my chiller with 2 modules and the proper amount of rads.

[KrowniK]

Just want to bump this thread, just so it doesn't fall off the face of the earth like all the other threads in this forum.

[batlecryawesome]

I wish I could build something like that. I guess I ll have to buy one.

[Nidor]

Any updates? Love to see TEC still being used and improved.

[KrowniK]

Ok, so here is the update I promised, however it is the pre-final stages; as I can't test it on a CPU until Mindchill sends me one of the waterblocks we designed (more on that to come later?).



Anyway, so this was the entire rig setup. It consisted of a 24v PSU, with the 50mm fan on the Meanwell PSU (it was pretty loud by itself). The TECs were wired up to receive 12V each, which is what I intend to run them at with an ATX PSU.

These are 4 Intel STS100C intended to be used on Xeon Processors at 130W TDPs. I will say now that because they're server heatsinks, their fans are in the ridiculously loud range when set to high speed; LUDICROUS SPEED!!! I didn't think they'ed be that loud but, ya full speed is too much for even me; and I run my GFX cards at 70% to 100% without flinching. If the fans became out of sync, it would create a fluttering sound, but that was due to me not buying all the heatsinks at the same time. The first one I bought came with a Delta fan that was rated for 12V @ 1.4A, the other 3 came with Foxxcon fans which are 12V @ 0.5A.

Another problem with the fans and my ghetto fan controller, is that the fan controller is based on resistive voltage limiting. So the controller only has 3 "channels", while I have 4 fans, 1 of which will use different power settings. This leaves me with the problem that the fans aren't being limited the same. So what I'm going to need to do is get an Arqtik Fan Controller to ensure that the fans are being limited to the same voltages. This however did not stop me from testing the High and Medium fans speed settings, even if they are not what they will be with a quality controller. When I tested low fan speeds, some of the fans would not even spin, so I negated that part of the test until I can do a fair benchmark with them working. Moving on from the fans...

I insulated the TECs this time, and after many leaks, it netted me about 4 - 5?C from my previous results, which definitely proved that the heat from the TEC was feeding back into my cooling loop just from heat radiation (kind of obvious but I didn't think it would be that much).

Next we move onto the coolant (1.5 - 2L of tap water). I know that this isn't a constant load, but the rate at which heat was removed from the water was pretty amazing. The hotter the water, the faster it cooled down; I don't have any pictures to back it up though. After all, what matters is the ability to chill the CPU, and I can't test that yet.




This is with the fans set to high speed (Ambient reads 22?C). I'm getting about 10?C delta between Ambient and Heatsink, which is a couple degrees more when I did my first tests. The STS100C contains heatpipes and something I've noticed with heatpipes is that, they work better when temperatures are higher. Which bodes well for having a cpu load on there. Unloaded the TECs are a load of about 65W, loaded with a 200W CPU, they are about 115W which can make a difference of 2 - 4?C with heatpipes. Doubt it will negate the temperature gained from the load though. Can't really predict their ?C/W as it changes with load.

Here is the results of Medium fans speed; I just took these pics tonight so it is a different ambient temperature.



15.2?C Delta from Ambient to Heatsink to deliver a delta of 15?C lower than ambient, at its coldest. In all honesty this means nothing with no ability to test it against a heat source. I wish this could be more conclusive but it just isn't in the cards at the moment. Those who want more of a review, will have to get on mindchill's case to send me one of the waterblocks but in all honesty don't. He's busy and I know he's got lots of work ahead of him. It'll happen when it happens. Testing is not a quick and easy process. So just keep updated with this thread and it will eventually be finalized on my end.

Sound clips recorded with my iphone at 2 feet will follow shortly, stay tuned....

[Puck]

Keep it up guys.

Can't wait to see both results.

[iNF3rN0]

why dont you use an electric hot water kettle, cut the cord on it and use an inline reostat/light dimmer switch and use and ammeter to accuratly create your heatload. It wont be 100% accurate but it should be pretty damn close.

[Ivanlabrie]

This is shaping up nicely...terribly interested in this.