My cooling problem....need help?

[danaltick]

Here is an email I sent to Swiftech Technical support:

Hi,
My name is Dan Altick. I purchased an MCX462 with the Delta 68.5CFM fan
(which by the way is outstanding). I have it installed on my EPoX 8K3A
motherboard using an XP1800+(1553MHz) processor. I installed it using
Arctic Silver 3 very presicely and accurately. I have fairly good
airflow through my case using a 50CFM intake fan in the bottom front and
a 50CFM exhaust fan in the top rear. With the cover on the case, at 20C
ambient temperature, my system temperature (which is being read from the
EPoX BIOS) reads 24C when fully warmed up. My CPU temperature reads
37C, a difference of 13C above system temperature. I have tested and
confirmed all temperature readings with an accurate temperature probe.
The processor is running non-overclocked with a VCore of 1.75V in this
test. I used the probe to measure the temperature of the air at the
intake to the Delta Fan. It reads 24C. The question I would like to ask
you is have I reached the limits of the conduction/convection of the
heatsink when combined with this processor? If not, then what would be
the best approach to closing this 13C gap, outside of using a more
powerful and noiser fan? What I would like to know is if I could expect
to effectively reduce this gap just by increasing the airflow around the
base of the heatsink? This, I could probably do just by adding tubing
from the exhaust fan down to the base of the heatsink. This would not
add any additional noise, but should remove the warm air from around the
base more effectively. But before I embark on this endeavor, I would
like to ask you if it would be worth trying, or do you think that I have
reached the convection/conduction limit of the MCX462? I hope I have
given you enough information. If not, please let me know. Any help you
could give would really be appreciated. Thank you.

Here was his response:

I wouldn't worry at all about the system temperature. Your CPU temp is
perfect, that's what counts.

Here was my response:

Yes I am pleased with the CPU temperature. However, this is running in
a non-overclocked state. I purchased this heatsink solely for the
purpose of overclocking the CPU; therefore, I would like to maximize its
potential. I ran an overclock test by upping the VCore to 1.8V and the
CPU to 1666MHz from 1553MHz. My system temperature stayed at about 24C,
and my CPU temperature rose to 43C. Again, this was at 20C ambient. It
appears as though my CPU temperature is rising even though the air going
into the Delta Fan is not. So I would like to know if I could minimize
this by somehow sucking air away from the heatsink base and pins using
dryer-type tube from the exhaust fan down to the base of the heatsink to
draw away the hot air that must be accumulating at the base and around
the pins; as opposed to getting the 80CFM Delta which would increase the
noise level. Thanks.

Here was his response:

Temp rise is normal in O/C conditions. Heatsink is designed to work best
in impingment mode (blowing down).

In other words; basically he was a waste of my time.
Can anyone offer me some suggestions here? Thanks.

[Kaitain]

It looks very much like you're reaching the limits of that heatsink/fan combo's capabilities for convective heat transfer (you can ignore conduction - the rate determining process is the metal/air transfer).

Now, there are two ways to design a heatsink... one requires a human to physically manipulate 3D partial differential equations for n discrete elements to come up with a design for a heatsink.

The other uses tested and accepted correlations developed for the chemicals industry.

Now guess what Swiftech will be using...

The correlations allow you to manipulate a pinned or finned surface design, given a known approach temperature. This is the assumed temperature difference the block will be capable of at a given heat flux. A reasonable assumption is that the temperature difference between the fin base and the air is 10 C at 80W

Therefore, factoring in the resistance of the metal (Fourier's law) you'll reasonably get a delta-T of 15C at stock speed. As you can see, your heatsink performs rather better than that, reflecting a closer design.

Now, when you overclock your CPU, it's heat dissipation will go up. Therefore, your temperature difference will increase as a linear function of that heat output.

So really, if you leave your heatsink stock, then it's doing rather well and needs no better answers than Swiftech gave you.

Now, to improve the effectiveness of the heatsink, you can do one of two things:

Reduce its internal resistance
Increase the air speed.

To do the first, your only realistic option is to lap the base, clean it up and keep it free from dust.

To do the second, you need a bigger fan. nb, axial fans (such as computers use) are better at blowing than sucking, so always have it blowing air into the heatsink.

[chazz469]

Kaitain....That's a lot of pontification for such little output...

Dan,

Here are a couple of other things. It seems as if you have your fan blowing away from the heat sink instead of blowing onto it. If this is the case, switch it. That's what the Swiftech guy was saying. (maybe I didn't follow you completely, if it is blowing onto the HS)

Also, with the cover of your case off, your intake/exhaust fans are virtually ineffective. Are the fans in your case blowing through the stock grill? Or have you modded your case? Either way, I would suggest cutting holes where the fans currently are to fit 120mm fans and arrange them according to the diagram below. Leaving your cover on should help to increase the effectiveness of the fans. If you're concerned about noise, the Enermax 120mm fans are really quiet. I have two in my case at the moment, and the pump from my water rig is louder. :-D

http://members.cox.net/hard10/images/flow1.jpg

BTW your name sounds really familiar....do you post on [H] as well??

[danaltick]

My 68.5CFM Delta is blowing on the heatsink. I have 2 50CFM Sunon fans for intake and exhaust just like your diagram. These fans are pretty powerful. I don't think the 120mm will buy me much and I just purchased these 50CFM's. What I'm thinking is maybe I could cut a hole in the side cover opposite the HSF and install PVC 4inch to 3inch reduction pipe directly over the HSF. Then I would turn around the front intake fan and make it an exhaust fan along with the rear exhaust fan. I really think this would probably be my best approach. This would draw ambient air directly on to the heatsink from outside the case. What do you think. If I felt this would give me an extra 4C drop, then it would be worth it. I already have the PVC pipe and sheet metal cutters. I have also already cut out the perforations for the rear and front fans and used foam insulation tape between the front cover and bay areas to prevent the front fan from recirculating inside air from the bay areas.

[Random Nonsense]

well, a 13c deltaT is AMAZING for a heatsink, the waterblock me and kaitain put togeather gave me an 11c delta! I dont think you will be able to improve that! But, by using a duct, you will be getting the 20c ambient air directly to the Heatsink, rather than your 24C air in the case, this will definately drop the temps!

With an 80mm delta, the airflow is so ferocious that you dont have any need to assist it away from the heatsink base.

[chazz469]

The 120mm fans that I referred to can move some air when they are cranked up. Like I said they are still quieter than my pump, and both of them are @ about 1700RPM.....which I think give the same or more cfm than the Sunon's you talked about. They are supposed to go up to 2300 rpm, but I've got them cranked all the way up and MBM is reading 1700 or so. I have a Sunon in the side of my case, and I think it's making the most noise. But it's just there to give fresh are to my RAD....(I'm working on a redesign).

Again.....make sure that you're case-cover is on, and that you're cables aren't inhibiting airflow. Then again, some procs just run hot. I had a 1.4 T-Bird that just wouldn't go below 45°C air-cooled. Load temp was around 62°C!!!!!! So, if you have the ability, try another proc with the exact same setup.

Venting like you suggested might make the exhaust design difficult...because heat rises and it would be better to go through the top while leaving enough room for air intake.

Have you got any pics? Maybe your case could be cleaned up with some round cables and/or electrical tape that would make all the difference.

Those HS are pretty *****-hot...I've seen people make water coolers out of 'em. So maybe you might consider that???

-----water cooling that is......:-D

[Kaitain]

Kaitain....That's a lot of pontification for such little output...

True... that's the last time I write a technical reply late at night (UK time)

danaltick seemed to be worried that his heatsink wasn't performing as well as it might. I hoped that, by just touching on the theory behind heatsink design, I might point out that it was performing at or near theoretical limits for the sort of design model that Swiftech are likely to be using. Didn't want to wheel out the full maths though (that appears in another thread started by BillA).

Sorry

[danaltick]

Just wanted to let you know that if I take the cover off, my temps at 20C ambient are 23C system and 34C CPU. My CPU temp. drops another 3C. That is why I think the PVC may help. Also, my case cabling is fine. I have taken care of that. Sorry, I don't have a digital camera for taking pictures.

[cloasters]

Glad that you're making progress, danaltick. While the 50 cfm Sunons are nice, you can double your throughput with two Sunon 120mm 108 cfm's in their place. The 120mm's are louder, but they sound more like white noise than anything else. AMD boxen require 100 cfm throughput, IMHO.
This volume of air isn't necessary if well thought out ducts are used.