Article Type: Hardware
Article Date: October 25, 2001
Back To Part I
Thermal Efficiency and Case Cooling
You have the hottest fan in the whole world…er, the coolest HSF, and yet you can’t get your CPU as cool as Bob down the street. Everything else about your systems is identical. Finally, you even borrow his CPU to see if you clock it to the same level on your identical motherboard. No luck, it simply gets too hot. (Boy, Bob is some nice guy!)
When you are returning the CPU to Bob he invites you into his office. Under his desk is a case style you’ve never seen before. This unit has an extra exhaust fan on top, and two extra intake fans on the side, right over the CPU. Hmm. You begin to wonder if Bob isn’t cheating!
 |
| CoolerGuys Wind Tunnel II Gamer |
Welcome to the high performance world of system cooling.
In Part I we talked about HSFs, and we noted that the important issues for HSFs are three in number: efficient heat transfer, efficient airflow to get rid of the heat, and a reasonable noise level. The fourth element is the CPU environment. It isn’t much use to blow a ton of air through a very efficient heat sink if the air is already hot. Your HSF will drop in efficiency as the air temperature in your case rises.
Consider for a moment the circumstances that a CPU must endure in the average case. The case is a closed environment with three major heat sources: the power supply, the video board, and the CPU. Further, in a high performance system, the mainboard itself can generate significant heat, as can a 7500 or 10K RPM hard drive.
It used to be that video boards didn’t generate very much heat, but in the case of the GeForce2 and the GeForce3, particularly if the board is over-clocked, the boards generate significant heat. Furthermore, where an 800 MHz Athlon only draws around 30W, a 1.2 GHz unit can draw up to 60W. That heat has to go somewhere, but if you have a typical case with a single inlet fan and maybe a single exhaust fan, you are collecting the heat inside your case environment.
The airflow in the average case with a single stock input and exhaust fan is roughly 30 CFM. The airflow in a high performance case can vary from 150 CFM to 250 CFM.
For this reason there are a variety of specialized cases out there, and though they fetch a premium price they are well worth the investment where maximum cooling is required. Let’s take a look at one high performance case and compare it to a stock case.
The Wind Tunnel II Gamer ATX
The Wind Tunnel II Gamer is made by The Cooler Guys, and the specifications are as follows.
- Built in USB, Sound, Mouse and Gameport in case front
- Two high speed 92mm (55CFM) fans for intake
- Three high speed 60mm fans for exhaust
- All fans sealed against panels
- 173 CFM airflow
- Four 5 ¼ bays
- Three 3 ½ bays
- Removable intake filters
- Fans and grills pre-installed
- Thumb screw panel removal
- All edges rounded for safety
I really like that last item. It wasn’t much more than a year ago when I received a nasty cut from the case I was using up to recent days. Sigh.
 |
| Hey, handy inputs! |
A stock case allows very little room for improvement unless you plan to haul out the jigsaw and drill and your collection of steel blades and files. For those of us who don’t have the time or motivation for such modifications, these high performance cases are just the ticket.
How do you know when it’s time to move up to a high performance case? A good rule of thumb is that if you have a single inlet and outlet fan, are running a high performance video board and over-clocked an AMD Athlon or Intel Pentium 3 or 4, you probably need to improve the airflow to your CPU. But if you REALLY want to know, you have to measure the temperature inside your case and compare it to the ambient air temperature. If your office is 21 C (approx. 70 F) and the temperature inside your case is 28 C, you have an airflow problem! You are reaching high efficiency when the temperature difference inside the case to the office environment is as low as 2 degrees C.
 |
| Extra Exhaust fan at rear. |
The Wind Tunnel II Gamer ATX looks unusual. While the case modification is based on the standard Super Case, this one has a 9 x 12 inch plexiglass sheet on the opening side (the side with the dual fans). This gives the case not only a custom appearance, but a somewhat racy, open-the-hood-and-drool-on-the-turbo look. It simply looks hot, er…cool, with the two large fans mounted in the plexiglass so you can see them and see the board even with case closed.
The case has four fans evident to the eye: two 92mm fans for intake air located on the side of the case that faces the mainboard, and two 60mm fans located on the top of the case for exhaust.
Location of the fans is not incidental. Hot air rises, making a top location sensible. Furthermore, the two fans on the side provide airflow directly over the CPU and video board, the two greatest producers of heat inside the case, other than the power supply which has its own intake and exhaust.
 |
Inner workings  |
The other exhaust location is on the rear of the case, directly behind the CPU. A thin, high speed 60 mm fan resides on the rear of the case behind the CPU.
With all these fans moving air one would think that the system is inherently noisy. Oddly, it’s very quiet. It seems the fans are well chosen for balance as well as speed. Other than the potential of a noisy HSF, the system is probably around 55 dB.
The intake fans plug into the four pin output of the power supply. The exhaust fans have the small three pin motherboard connectors to plug directly onto your mainboard.
Since the two exhaust fans on top of the case are joined to a single power input, I was able to plug them in beside the CPU fan. The Abit KG-7 has two fan connectors near to the CPU. The rear exhaust fan, however, has a slightly short set of wires, and the wires barely reach to the other fan connector, normally used for the input fan on the front of the case. It would be nice to see these wires a bit longer, and it would also be nice to see them sheathed in a bundle so there is only one wire dangling about.
The fan locations are cut by laser, so there are no ragged or rough edges. In fact all edges are obviously carefully polished. This is a stock style case which is carefully modified, and the workmanship shows. Things like sealer rings for the fans and even filters on the intake reveal careful dedication to a premium product.
The second unusual feature of the case appears on the front right side of the unit. Here you can see the added USB, sound, mouse and game-port extensions on the case right front side panel. You have the convenience of a drive bay break out box without the expense. Very nice!
I borrowed a thermostatic sensor unit from an engineer friend of mine. This unit reads from freezing up to 100 C and is accurate to within 1 degree C. It has a thermal sensor on a four foot cable at one end, and the temperature unit is an LED readout in a small unit at the other end.
The test system is an Athlon 1000 at 1.4 GHz on an Abit KG-7 mainboard. I have an ALPHA PAL 8045 heatsink with Mechatronix fan on the CPU.
I have an OCZ Titan 3 GeForce3 video board running at 240 MHz, with memory at 550 MHz. My hard drive is an IBM Deskstar 7200 RPM running under ATA 100. The system has 512MB of PCI 2400 running at 300MHz.
The CPU in this system averages 34 C at idle and 49 C under load. The case flow is roughly 30 CFM and my office is consistent at 21 C.
 |
| Yer basic comparison chart... |
After one hour of running IL-2 Sturmovik my stock case measured 26 C internally. My office was still 21 C.
I swapped the entire system out into the Wind Tunnel II case. I then ran IL-2 for one hour and checked the case temperature. It was 22C. I then load tested the CPU and it was running at 45 C. This is a significant improvement and is only attainable because the air in the case is so much cooler.
Heat is the big enemy of high performance gamers. Consider that it takes about five seconds to completely fry a 1.2 GHz Athlon CPU when there is no HSF attached.
A reduction in temperature like that above extends the life of components, but it can also be used to attain even higher clock rates. I was able to clock my CPU to 1466 and it ran at 30 C at idle, still cooler than my previous temperature at 1.4 GHz. The gain isn’t that significant for gaming purposes, but it does show what can be accomplished with effective cooling.
Finally, there isn’t much I would change on this case, but it would be nice to have an external switch for the two intake fans. When the system is on Standby these two fans keep moving, since they are powered directly by the power supply rather than a mainboard connection.
Other Methods of Case Cooling
While I was researching this article and testing the Wind Tunnel II Gamer, I ran across some information on other case cooling options. Since the ambient air eventually becomes the limitation for effective cooling, some have ducted outside air directly to their system.
When I realized that the location of my box is about four feet from an outside window that slides open, I thought this would make a fascinating experiment. Since the window slides, it would be simple to open it six inches and attach a locking device for security, and then fashion a six inch plywood closure with a hole to attach a four inch air duct.
Simply brilliant, right?
Ok, a little extreme, but worth a try.
It took me about half an hour to create my plywood closure with two four inch holes, a five minute run to the hardware store for flexible ducting material, and as a temporary measure some old-fashioned duct tape to fasten the inside ends of the air ducts directly over the dual intake fans in the side of the case over my CPU. In roughly one hour my high-end modification was complete and ready to road test.
The outside air temperature at the beginning of the test was 8 C. The air temperature inside the Wind Tunnel II before I hooked up the ducts was 22 C.
I left the system running with no load for one hour and then checked the inside case temperature. It had dropped to 12 C. This was becoming interesting. I checked the CPU temperature using VIA hardware monitor. The CPU was 27 C, the lowest I have ever seen.
Next I ran IL-2 as steadily as possible for one hour. My case temperature rose to 13C and the CPU rose to 26 C. The difference surprised me. Obviously, using cool outside air is a great way to increase efficiency. It would be worth considering where maximum effect is needed, though it might also be worth worrying about the effects of condensation where humidity is high.
Peltier Junctions
Another option used by some overclockers is the Peltier junction. In order to be effective with a very hot CPU at least a 130 watt unit is necessary. A Peltier junction provides active cooling of the CPU, with one side of the device getting hot and the other getting very, very cold. A Peltier device for a CPU looks like a thin snack cracker and is about the same dimensions. Here is how it works.
 |
| A Peltier Junction |
The Peltier Effect was discovered in 1834 by a French watchmaker and physicist by the name of Jean Charles Athanase Peltier. The Peltier Effect takes place when an electrical current is sent through two dissimilar materials that have been connected to one another at two junctions. One junction between the two materials is made to become warm while the other becomes cool, in what amounts to an electrically-driven transfer of heat from one side of the device to the other. This transfer can be so dramatic as to bring the cool side below freezing.
And therein lies the dilemma. Where a few years ago a Peltier device was worth considering, now the risks outweigh the benefits. GHz CPUs get so hot that you must use a very powerful Peltier device to derive any benefit, but the benefit over a very effective HSF like the ALPHA PAL 8045 is minimal. Meanwhile, you can have all kinds of problems with condensation. More than a few Peltier users have had enough condensation to cause shorts on their CPU and fry the CPU or their motherboard.
Notes on Thermal Grease
Thermal grease is that thin layer of grease that is applied on top of the CPU prior to installing an HSF. It compensates for any irregularity in the contact surface, and it facilitates the transfer of heat between the CPU and HSF.
Thermal grease is thermal grease, right? Not necessarily.
Conductive grease varies in efficiency about as much as the HSFs you can purchase. Furthermore, the incorrect application of thermal grease will limit the effectiveness of heat transfer between your HSF and the CPU.
The first grease I used on my ALPHA PAL was a stock item available through Radio Shack. It is white and probably lithium based. It spreads easily and is rather nice to use.
I spread a thin layer on top of my CPU using a flexible plastic spreader. I then placed the HSF on the CPU and removed it again. I could see that the entire area was dimpled, a good indication that the grease was making contact across the entire area.
I tested the CPU under load and achieved an operating temperature of 32 C at 1466 MHz with VIA’s hardware monitor. I then shut down the system, removed the HSF, and cleaned all the grease from the HSF and CPU.
Next I applied some Arctic Silver II from a small tube. This stuff is a bit of a pain to work with, since it is quite thick. It helps if the tube and CPU are slightly warm.
It’s impossible to apply this paste with a finger tip, and it is not a good idea to do so. Apparently the oil on your fingers can reduce thermal transfer.
I spread a thin line of Arctic Silver across the CPU, and then used my plastic spreader to smear the surface until it was covered with a thin layer. When this job is complete you should almost be able to read the stepping info on the core.
I reset my HSF in place and closed the system down. I ran it under load for twenty minutes and checked CPU temperature with the VIA hardware monitor. The indicated temperature was 31 C.
It’s difficult to know for certain whether the difference is incidental, so I ran the test a second time and achieved a temperature of 30 C. I then removed the HSF and cleared off the Arctic Silver, and reapplied the white thermal grease. I ran the system for 20 minutes under load and again achieved a reading of 32 C. It does seem that an improvement in heat transfer is achieved when using Arctic Silver compared to some types of thermal grease.
