[Syzygy1290]

Hey all,

So I've been a little confused about my computer temperatures, so I decided to dig into some Martin and Skinnee graphs to try and figure it out. I thought I might as well post my thought process.

Martin and Skinnee data taken from these places...
http://daemonkin.hubpages.com/hub/Choosing-a-radiator-for-pc-watercooling#
http://martinsliquidlab.org/2012/04/09/hardware-labs-gtx-360-radiator/4/
http://martinsliquidlab.org/2012/12/22/xspc-ex360-radiator/4/

I have:
1 Xinruilian 140mm fan on an ex140 rad (as intake)
2 Bitfenix 120mm LED spectres on a ex240 rad (as intake)
2 Bitfenix 140mm LED spectres on a ex 280 rad (as exhaust)

The Xinruilian runs at 1350 rpm max
the spectres do 1200 rpm max

A 360mm rad has an area of 120X120X3 = 43200
My setup has a radiator area of 120X120X2 + 140X140X2 = 68000 at 1200 rpm
and 140X140X1 = 19600 at 1350 rpm

So at max capacity at a delta T of 10C and using Martin's graphs (assuming fans are essentially the same at a given rpm for simplicity):

A ex360 dissipates approximately 165 watts at 1200 rpm
and about 180 watts at 1350 rpm

So my heat dissipation to achieve Delta T 10C should be...
165W * (68000/43200) = 260W
180W * (19600/43200) = 82W
or 342 Watts

According to Skinnee with a delta T of 10C I should get (assuming thermochill PA120 rads are about the same as ex ones)...

An PA120 dissipates somewhere between 360 and 490 watts. So using 425 for the 1200s and 475 for the 1350 I get...
425 * 1.57 = 669 Watts
475 * 0.454 = 216 Watts
or 885 Watts

A little confusing...

So in reality I am cooling a 2x7950 overclocked to 1050/1450 and a 3770k at 4.4ghz which according to my Kill-a-watt gives me 650ish watts at full load on the GPUs (litecoin mining). I added some pics below. The 280 at the top is exhaust which might be throwing stuff off a little bit.




So my delta Ts are well over 10C...currently about 25C = water temp - air temp. So not to knock on Skinnee at all, but it seems my temperatures are more in line with Martin's data. I can't get to Skinnee Labs original articles so it is hard to say if any other differences in testing that are causing the discrepancy. According to that graph though, I should be able to cool 1000 or so watts with a dT of 25C and a 360mm rad.

According to Martin, In order to cool to 10C delta I need about 250 watts more cooling or double the amount of radiator space so my 25C delta is actually not too bad (at least relative to what I thought when I first saw Skinnee's graph)

 

[Jakusonfire]

There is a key difference in Skinnee's testing and Martins later testing. Skinnee and Martin originally used open bench rigs for rad testing, but Martins more recent testing was built into a special chamber to more closely simulate a PC case like environment.
External testing is affected by the surrounding air and will give higher numbers. There is also much less resistance for fans to overcome so actual airflow per RPM is much higher.

When approximating cooling you can't just multiply the square surface area, it is the airflow and air volume of the fans that is important and they are not square and rarely equal in pressure.

In your particular rig you will get better temps if you change the top rad to intake, at the moment it is breathing warm air. Also the rads should be completely sealed around their edges, so air can only enter the case through the rads, at the moment the bottom rad looks like it could re-breathing air from the case through the holes around its edge. Relatively small adjustments to airflow can make quite good improvements, at the very least free ones.

 

[Syzygy1290]

Makes sense. I was just confused because I've often seen "120W = 120mm rad" and "10C delta is normal." Which made me think something was horribly wrong in my loop (bad block seats or something...). I'm actually pretty surprised how much of a difference open vs closed makes. Over 500 extra watts dissipated is pretty impressive...makes me think that external rads are really the way to get good performance.

I tried to estimate from the graphs the watts dissipated at a particular rpm. Different fans will have an effect due to better design. What really is important for cooling purposes is the speed that air can move through the rads. I don't think it will make the results too bad though:

http://martinsliquidlab.org/2013/04/24/fan-testing-round-11/

This shows that at a given rpm range..say 1000 to 1200, the flow speed is pretty similar across those 5 fans. At least within 10 to 20%. Same goes for radiators...if you look at the earlier Martin link I posted, it shows pretty similar performance (within 25 watts at 1000rpm) between rads. So yes, better fans/better rad combos will give better performance, but for loop planning and/or sanity checks I think radiator area calculation are pretty handy.

As for my build. It is still in the tweaking stage. I just added in some new fittings. I'm still waiting for a replacement res (mine had a small crack that caused a slow leak). Once that gets here I plan to try and optimize things. The top rad should be flipped probably, but I have to unscrew the mobo, graphics cards, and back 140 rad to do it...so I've been a little lazy

I kind of wish Martin had a fan cfm thru rad vs rpm per fan graph (maybe he already does?). Since he already has some rad performance numbers, that would make it pretty easy to get closer approximations of heat dissipation.

 

[givmedew]

I would recommend that you do what jack said and reverse the one set of fans and seal everywhere a fan touches something even in between each of the fans on the radiator.

It isn't my rig but for a litecoin rig why does sound matter? I would have gone with server grade fans even if you game on it sometimes because even a 4000RPM 120 x 38MM Delta can be ran down to ~500-700RPM (PWM) and is nearly silent at that speed.

 

[Syzygy1290]

This isn't a litecoin rig. Its my gaming/home pc, but I have it mine when I'm at work.

This is a litecoin rig...

 

[givmedew]

Yeh so when it is mining whats wrong with 4000RPM you won't even be home

You can snag 4000RPM PWM Server fans with dual ball bearings for $8-15/ea on the bay. AFC1212DE I buy the 1.6a ones personally... they work very well and make very little noise at the lowest setting! I think the ones that say 1.6a have what is called soft start but I am not certain. I know the PFC1212 that is the hp proliant version has soft start but the PWM is reversed so you have to control the speed manually with a $10 PWM signal generator from frozencpu.com or build a PWM signal inverter with parts from radioshack. You can't control the HP PFC1212 with a motherboard right out the box.

But that is the extreme side of things there are fans that are half way in-between... just make sure they are PWM and you will have silence and power.

 

[Syzygy1290]

Good point. I should look into it actually. I used the fans that are there because I had them already originally. The XSPC fan came with the radiator, and I was already using the bitfenix ones before I had the second gpu or the watercooling setup. I actually have a 8 way pwm fan header to 12 volt laying around somewhere. Do you have any 140mm suggestions? Quiet with half decent performance at low rpm would a plus. How's the cooling on those AFC1212DE at sub 1200 rpm? Clearance is another issue...I just so happened to pick a case that is a pain to watercool. I've already done a fair bit of modding to fit any of the rads in. Any suggestions for thinner fans if I need to stick with that?

Thanks for the advice! +rep

 

[givmedew]

Where is the ex240? Is that the one in the bottom? If so I don't see why it couldn't handle 38mm fans... Maybe just 2 afc1212de fans would be enough of a change that while you are lite coin mining it wouldn't get too hot. They should cool fine. I wouldn't want you to blow a ton of money on any brand of dual ball bearing fans and then find out you hate them. Buy 2 afc1212de fans put them on the 240. Be prepared to re-wire them so go to performance-pcs.com and buy a some fan header fins and a few 4 pin headers of whatever color you like. You don't need a special crimp tool you can use needle nose pliers.

That's if you want to save yourself $30 a fan. Those fans don't come cheap at retail pricing and all the cheap ones are going to have dell connectors on them. There is an adapter you can buy but putting new plugs on them is easy as cake.

Having 2 fans the equal 300CFM total should make a dent on your temps and without investing more than $25-40

 

[givmedew]

The ones I have are AFC1212DE p/n MJ989 1.6a there are many variations. All 4 wire ones that are from dell computers should be 4 wire with a normal PWM control and a normal tach signal. Be aware though I can't be certain of that... My wires are red black yellow blue with the yellow being tachometer and blue is PWM. I see ones that have a white wire instead of yellow... My HP deltas do not have a normal tachometer and the PWM is inverse. So just be aware. Don't go buying 10 of a different P/N number and then get upset if they are non standard. Also there are different amperage ratings... I have no clue exactly why this is. I know on the pfc1212de that the low and high amp models are the same speed but the low amperage one has special circuitry that starts the fan slowly so no surge so the high amp one is with the surge. I do not know if this is true with the afc line.

$10 ea shipped
http://www.ebay.com/itm/Dell-XPS-700-710-720-CPU-Processor-Case-Fan-Delta-AFC1212DE-MJ989-XM060-/190881350268?pt=US_CPU_Fans_Heatsinks&hash=item2c716a427c

That is the version I have... Buy em don't like em and I will buy them from you (no more than 2 please).

 

[Syzygy1290]

Cool! How will deltas perform compared to say the GT AP30? The 240 rad at the bottom definitely has room. the 140mm and 280...probably not. I'm not so worried about high rpm noise, but when taken down to 1000 or 1500 rpm, will the GTs dominate?

http://www.performance-pcs.com/catalog/index.php?main_page=advanced_search_result&keyword=Typhoon&categories_id=696&inc_subcat=1&manufacturers_id=&pfrom=&pto=&dfrom=&dto=&x=32&y=10

 

[Syzygy1290]

Whats the difference between PFC and AFC series? Pressure vs Airflow?

 

[givmedew]

Quote:

Originally Posted by Syzygy1290

Whats the difference between PFC and AFC series? Pressure vs Airflow?

PFC is specially designs for higher static pressure and airflow. It is capable of 200CFM. It can not spool down as low as the AFC and it will cost quite a little more since you have to make sure you get the model that has standard PWM control.

I really don't know how they compare to the AP30 at part throttle. You might as well grab the AP31 if you are going to buy GTs. It is probably overall a better fan to have than the AFC1212DE.

 

[Syzygy1290]

I've been looking at some of the delta fans...and am rather overwhelmed with the diversity

Do you know what some of the product codes mean? PFC is pressure optomized. AFC is airflow. FFB? QFR? AFB? PFB? I take it 1212 are the 120mm fans. What about the ending prefixes? SHE and UHE? D and DE? FOO?

Also any opinion on San Aces?

Thanks!

 

[Syzygy1290]

I kinda wish I had the cash to buy a couple pairs and try them out