MCW 452 chiller build

[RichieGP]

Hi there,

I have bought a swiftech MCW452 chiller to stand at the centre of my cooling setup and i was wondering if you could help me with some power issues.

Current system:
Asus AN8 32SLI deluxe
AMD X2 4800+ CPU
2GB Corsair 3500LL pro RAM
2 X raptor 74GB HD
1 x Samsung Spinpoint 250GB HD
2 x XFX 7800GTX 512mb XXX GPU's in SLI
Dell 30" screen

Case:
Silverstone Temjin TJ07

PSU:
2 x Seasonic S12 600W PSU

Cooling:

Hot loop
MCW 452 chiller - single loop including both hot sides and 2 x radiators
Water pumped by laing DDC pump with integral reservoir
2 x Alphacool NexXxoS Xtreme II radiators in series (ie 240mm rads x 2)
10mm tubing


Cold loop:
2 x alphacool nexus GPU coolers for 7800's
Swiftech apogee CPU block
10mm tubing (insulated)
XSPC X2O Pump with integral reservoir


One PSU powers the computer components and one PSU powers the chiller. Seasonic claims +12V1: 18A / +12V2: 18A on its dual rails so attaching the pelts to the PSU should be ok, should it not? Swiftech claims they need 12V and 17a supply. I have cut off the molex connectors and attached the pelts using connectors that are rated to 20a each, to prevent melting of the molex

However, I am only getting CPU temps of 31-39c and GPU temps of 31c at all times (i think this is a problem with nvidia bios which i will sort out later). Even so, these are still higher than i would expect from a chiller with >400W cooling capacity and in reviews that i have read, some people have got sub-zero temps.

I was wondering if the problem lay with my power supply. Does anyone have any experience with this chiller and powering it? Can anyone recommend an alternative PSU for use with the chiller? Most meanwells are only single output so i was wondering what would work to power both pelts?

Are there any other reasons anyone can think of for such poor temps (almost ambient). I know the 10mm tubing may not be adequate for the hotside loop and that i should probably run each hot side block in parallel with its own rad.

thanks for your help in advance

[Brians256]

The best way to solve your problem is to characterize the system. Find out where the heat is and so on.

What's are the water temps in the hot side (pre-MCW452 and post-MCW452)? What are the water temps in the cold side (pre-MCW452 and post-MCW-452)? What are the voltages actually being fed into the TECs (look for voltage sag)?

Once you do that, you can see whether the hot side is being kept cold enough and whether the problem is just too much heat or whether it is insufficient voltage or if it is insufficient flow, etc...

[ricecrispi]

looping that beast must be mind boggling.

======
To check psu I would get a multimeter and check how many amps and what voltage you are getting at the pelt.

I would check water temps like brain suggested but I'm pretty sure what the main problem is.

The loop will have have Qt of 275-300 watts and not 400 watts. You're system at idle is running close to Qt and higher at load so you are getting bad results. Plus inefficiency. If you can't get the chiller to perform right I beat the two rads will be enough.


1) tubing size is too small. Bigger tubing size would improve performance alot. Doesn't help you have alphacool flow killers in sli config either and you need to insulate those puppies.

2) IF psu is underperforming. Getting meanwell psu will help a little if you can run at 13.5-14 V. Meanwell PSU SE-600-12 power supply on ebay is up for bid.

[RichieGP]

thank you both for replying back so fast.

I am in the process of getting the inline water temps set up, but taking some time (work being very busy).

As for the tubing, i appreciate the 10mm tubing is too small for the hot side loops and i will definitely try and sort that out, but for the cold side loop, i thought that low flow would be appropriate (i think that this is suggested by Swiftech) and therefore a restrictive circuit wouldnt be too bad.

Everything is insulated and has dieclectric grease around the components.

As for the PSU, I am based in the UK and ebay.co.uk doesnt seem to be selling switching PSU's, so I would be having to buy at full price, which is a bit steep

Do you think my setup is pushing past the limits of the chiller? If so, i guess what ricecrispi is saying is that just using the WC system without the chiller would give comparable results.

sorry for all the questions, but also the only reason i stuck with TEC chilling and not phase change was noise. Is there a phase change kit out there that would be as quiet as a WC setup? Vapochill used to sell a setup that could fit inside a large ATX case, but i cannot find that for sale anymore

Thanks for all your help

[ricecrispi]

I think the main problem is running the chiller at full capacity or past it. Might be a psu problem so check that first. If you have psu running at full capacity there are minor tweaks you can make. I would then setup a water temp monitor to help troubleshoot. If nothing works, with two dual 120mm rads you should get great temps.

PH did a review on the chiller. He got very cold coolant temps at higher flow rates but didn't change cpu temps that much.
http://www.procooling.com/index.php?...s&disp=84&pg=2

The review shows flow rates on the cold side is important and higher flow rates meant lower temps. I think with the sli setup, 10mm tubing and 450lph 1.8M x2o pump means you are having very low flow rates. You can get better performance with bigger tubing on cold and hot side and a bigger pump on coldside loop.


He concluded insulatation was a possible propblem since he didn't insulate the cpu block. Did you insulate the apogee and gfx cards in dieletric grease then enclosed in foam? The foam is a key component and the grease alone isn't enough. Due to their large surface area, if you don't insulate the two cards properly you will have a second pathway that will transfer the heat.

[jaydee]

http://www.swiftnets.com/products/MCWCHILL-452.asp

Take a look at their Cooling capability chart.

You can extrapolate/guesstimate how well it will perform at whatever wattage you are dumping into the system.
According to this article http://www.extremetech.com/article2/...1885978,00.asp your cards use 120watts max each so 240watts max. Your CPU stock is either 65watt or 89watts depending on the version http://www.short-media.com/review.php?r=316. We will go with the lower and say your system is dumping 305watts max. So using the data available we can add 3C to each 10W more added to the system. 305watts would run about 55.5C over ambient...... Now we all should know your computer is not going to be running max wattage and more than likely 1/2 of the wattage under normal loads. So let's say 150 watts. That would be 19C over ambient. If your ambient is 25C then that would be 44C.

Those numbers are not perfect but I would say you are lucky to be getting what you are. This all could be crap though if the wattage info is not close.

[ricecrispi]

thanks for the verification jaydee and some solid evidence. I pinned him at 300 max wattage as well. The guy is gettting 31 C idle to 39 C load for CPU and 31 C at gpu. If that is accurate temps I thinks thats pretty good. I think his 2x dual rad would get about the same temps, mayber lower idle and higher load temps.

[jaydee]

If had that setup I would get a high quality CPU air cooler and just run the GPU's on the chiller.

[ricecrispi]

That would be the cheapest and best alternative.

Otherwise I would buy another pump and use one of the dual rads to create a seperate loop for the cpu.

[jaydee]

Originally Posted by ricecrispi

That would be the cheapest and best alternative.

Otherwise I would buy another pump and use one of the dual rads to create a seperate loop for the cpu.

Yeah would be ashame not to use that nice Apogee block. Could sell it and buy the CPU air cooler though I guess.

[RichieGP]

Thanks everyone for replying.

I have been waiting a long time to be given a Freezone from coolit systems. I was going to use this for cooling the CPU and then just use the chiller for the GPU's. I waited too long and got a bit impatient, therefore got the apogee to act as an intermediate product.

If it does arrive, then I will experiment with using this for the GPU at high power and then also post some results just for the GPU's. (They never move from 31c so i think that they are actually misreading --> a known problem i believe --> will apply the nvidia bios tool to get round this problem and post results also)

Otherwise, think i might try a separate loop using one of the dual rads to just cool the CPU with the apogee and see what i get with that.

As for my previous post, anyone know of a phase change product that would fit in the base of a silverstone temjin TJ07?

Cheers

[RichieGP]

"If it does arrive, then I will experiment with using this for the GPU at high power "

oops, i meant CPU here not GPU, sorry