Quiet and Effective Air Cooling on the Athlon XP – Is it Possible?

It has been said many times before — Athlons are hot. Now, with Athlons, “hot” has generally meant a few things. Many referred to them as “hot” processors, referring to their speed. This is the more metaphoric and slang usage of the phrase “hot Athlon”.

Then there is the more literal use: Athlons are hot, as in burn your finger hot.

This literal use of the phrase usually results in huge copper heat sinks with screaming fans mounted on them. It seems to be generally understood that “hot Athlons” = loud noise.

It is not uncommon to see some overclocked Athlons producing 100watts or more of heat. Even stock speed Tbirds went past 80watts. And just look at the new 0.13-micron Thoroughbreds. The 2200+ versions are putting out 70 watts of heat over an area of only 80mm2 – talk about a cooling challenge!

Of course, we all know that cooling challenges are not just related to Athlons. As P4’s ramp up in clock speed, so does their heat production. Add in overclocking to the equation and heat is an even bigger issue. Unless you watercool, achieving effective cooling at low noise levels is very difficult.

Taking all this into consideration, I posed myself with a challenge. Is it possible to effectively air cool an AthlonXP 1600+ with low noise levels? Keep in mind, this is for a business machine, so water cooling or wrapping an A/C unit around it are not viable options just yet. : To better understand how I went about this particular challenge, I’ll give a little background.

I’m currently working in IT at an environmental services firm. It is a smaller company, so I basically handle nearly all of their IT needs including building and upgrading systems. Recently, I had acquired the parts to assemble another AthlonXP system. (It should be noted that in the past 2 years, based on price/performance, all of our systems I’ve built have been AMD…and all have exceeded expectations…though lately Intel is giving some real strong competition.)

It was assembled quickly, as the system was needed by one of our employees. Everything went great, but it turned out to be quite a bit louder than I had expected. I was not the only one to notice. I had given the system to one of our Industrial Hygienists, who reminded me that OSHA (Occupational Safety and Health Administration) has noise regulations for office environments, and that this system might be pushing that limit!

So, I am sure you’re thinking “Time for that Delta to come off!” Well – I did not use a Delta fan. In fact, I was using a “quiet” Speeze Super Rock with a 70mm x 15mm fan and 2 80mm case fans. These “quiet” parts were quite loud when combined.

At any rate, the challenge was set. Could I keep it cool and keep it quiet? In the following, I will talk about some of the factors involved in decibel testing, and also share with you my experience with quieting the Athlon XP.

One of the things I discovered in this experience was that testing for decibels is as complicated if not MORE complicated than testing for temperature. While the test methods themselves are not necessarily more difficult or complicated, there is a very wide set of variables that can change the outcome of a test.

As a result, I learned that a manufacturer’s claimed decibel levels should be taken with a grain of salt, if not ignored completely. First though, I will explain decibels a bit further.

A “decibel” is basically a way of creating a logarithmic scale relative to some reference. Decibels do not necessarily have to relate to sound, as there is the dbV, which are decibels on a voltage scale. The key word in that statement is logarithmic. In other words, as the decibel scale rises, the power rise is based on a logarithmic scale. So, if we were to compare 70db to 60db, 70db is actually 10 times more powerful than 60db.

Now, this does not mean that 70dbA is 10 times louder than 60dbA. In fact, while it is 10 times more powerful, humans only sense it as 2 times as loud. So basically, a 10dbA increase in sound is about twice as loud.

You may be wondering why there are different acronyms used for discussing decibel levels. Since the decibel refers to a scale (db), there are various types of decibel scales. For instance, I mentioned earlier that dbV refers to voltage on a decibel scale. Sound actually has three decibel scales, dbA, dbB, and dbC. These basically refer to the frequency response of the scale.

The three scales can be used to measure the following frequencies:

Most ratings are given in dbA. This should be sufficient, as it is doubtful that even screaming Deltas put out much (if anything) in frequencies above 10,000 Hz. In addition, the “A” scale is often used to compensate for the fact that our ears are not equally sensitive over all frequencies.

As I learned more about decibels, I actually started to realize just how easy it would be to pad a fan’s noise rating. The obvious way would be changing the distance from the fan that the sample is taken. I went to look into this more, and found a very interesting web page (HERE) that discusses various ways in which fan manufacturers can make noise levels appear less than they really are. This was a great resource and definitely recommended reading

As stated on this page, fan testing should be carried out under a certain ANSI standard. That’s ANSI S12.11-1987 in case you wish to look it up. However, various things can be done to pad the data even while abiding to the standard. Testing a fan that is freely suspended (reduces vibration), testing the sound further from the fan, and running the fan at a slightly slower speed will all give lower dbA results than will be found in the real world.

Since most manufacturers do not give the specific details of their testing methods, decibel ratings for fans should be taken with a grain of salt.

Get on with it!

{mospagebreak}

Mike O’Konski

That is probably more than enough information regarding decibels. While that is interesting, I must admit the main goal here is to run an AthlonXP at a good temperature with less noise than usual. So, here is the test setup and procedures, and after that, the results.

I started by using some of my cooling knowledge from overclocking. First off, a good heat sink and fan is only as good as the case temperatures it must deal with. So good case ventilation would definitely be a plus and help me to run a lower CPU fan RPM.

I also used the idea that larger fans spinning slower can generate the same CFM at lower noise levels than small fast spinning fans. Expanding on this idea, I figured the best way to compromise between noise and temperature would be to have a fan that was adjustable.

As a result, for the heat sink and fan I chose a Thermalright AX-7 and paired it with an Enermax 80mm adjustable fan. The Enermax fan adjusts all the way from whisper quiet to noisy, and puts out a surprisingly good amount of air at whisper quiet. As for the AX-7, it is a great heat sink that was designed to use an 80mm fan.

As I continued to evaluate the system, I realized that the Enhance PSU was a source of a considerable amount of noise. The fan mounted in the PSU was quite loud and would have to go. I decided to replace all the case fans with 80mm Vantec Stealth units. They are rated at 27cfm, which is a tad low, but are virtually silent! I figured that smart fan mounting would make up for the low CFM.

In the original configuration, I had an 80mm Sunon fan in the front as an intake. Then I had the stock fan that came with the Evercase and left it mounted in the side, also as an intake. The PSU fan was exhausting, and that left the CPU fan.

Since the Maxtor hard drive mounted up front was running nice and cool, I decided that a fan up front was not needed. So, I mounted one of the Vantecs in the side as an intake, one in the back as an exhaust, and the PSU fan was swapped for a Vantec that exhausts. The desired effect would be a steady stream of cool air feeding the CPU fan, and the hot air from the CPU then being vented out.

I ran the test in the original configuration, and then installed the new parts. Each time, I took dbA measurements in exactly the same position and manner. For temperature readings, I used a thin thermistor next to the CPU core. I then ran Folding@home and Sandra Burn-in until the temperature leveled out. Ambient temperatures were taken in front of the case. (They are not internal ambient temperatures.)

{mospagebreak}

Mike O’Konski

Test Configuration:

System:

• AMD AthlonXP 1600+ (1.4ghz, 63 watts)
• Epox 8KHA+ VIA KT266a Motherboard
• 256mb Crucial PC2100 DDR
• 60gb Maxtor 7200rpm D740X(L)
• Evercase Midtower ATX
• Enhance 300w PSU
• Windows 2000 Professional SP2

Fan Configurations:
Test 1 (Original):

• 80mm Sunon Front
• 80mm Globe Ball Bearing Side
• 80mm Adda Ball Bearing PSU

HSF = Speeze Super Rock w/ 70mm x 15mm Fan

Test 2:

• 80mm Vantec Stealth PSU
• 80mm Vantec Stealth Rear
• 80mm Vantec Stealth Side

HSF = Thermalright AX-7 w/ Enermax 80mm Adjustable Fan

Test Results:

*Full Load = Running Folding@home and SiSoft Sandra Burn-in after 25 loops, thermistor placed next to CPU core.

Sound Levels:

*Rear of case is taken a few inches from the PSU fan and where the rear case fan would be. It will measure the PSU fan from 2 inches, the rear fan (if installed) from 2 inches, and the CPU fan from about 6 inches away.

**Front of case, work level, refers to the point in which your head would be if you were sitting at the desk and working on the machine. It is a few feet above the desk, and about 2 feet from the front of the case.

***Ambient refers to the noise level in the office with the computer completely turned off. This shows what noise is already present in the office, such as HVAC noise.
{mospagebreak}

Mike O’Konski

I decided to show the noise level of this system when compared to some of the HSF units that have been tested at Overclockers.com. I chose a few louder units, but included many “quiet” heat sinks that have been tested. The idea is to just show a comparison of how this system compares to these HSF units. Remember, these are just the heat sinks, they don’t include added noise from case fans:

Even after the second test, I still feel there are a few more things that can be done to help quiet things even further. They are still only theories, but are based on things I took notice of when testing things. First, the side case fan seems to generate some intake noise. If I tape off half of the intake, it gets noticeably quieter. However, that will impact performance. But perhaps there is a way to mount the fan or design an intake vent that will eliminate the wind noise.

In addition to this, the fan mounts themselves contribute to noise. A free mounted fan is quieter than a mounted fan, mainly because the mounts transfer vibration and create more noise. Rubber bushings on all the fan mounts (such as the ones the AX-7 ship with) would eliminate most of the vibrations being transferred to the case and should make the fans even quieter.

If you take a look at the OSHA regulations, it says that the typical “quiet” office is between 40 and 50 dbA. Two feet from the case, I recorded a noise level of only 46 dbA. Even right behind the case, which is only inches away from the rear case fan, PSU fan and CPU fan, the measured level was 56 dbA. In both instances, this was about a 10dbA drop. As mentioned earlier, this basically equates to the noise level being cut in half.

But subjectively, how quiet is it? I must say things seem much quieter. Stand at the door to the office and you can’t tell the system is even running. Before it was a clear whine. In my opinion, the difference is night and day. Luckily though, I was not the only one to form an opinion. The change was noticed by that Industrial Hygienist, and his comments were basically “Good job, it’s quiet now!”

For all this, I did have to take a 3.4C temperature increase. However, that is better than I had expected. At 63 watts, the C/W is .36, which is pretty respectable. The best part is what I neglected to say about the CPU fan – the temperatures are reflecting the Enermax fan at its lowest speed setting.

After the test, I upped the fan speed a few hundred RPM with no perceived sound increase. Unfortunately, the system was in place and in use, and my thermal testing equipment put away. So I did not get a temperature reading with the fan spinning at higher RPM’s. I feel this might have gained an extra 1 C or so.

In my eyes, the challenge was met. Overall system noise was cut in half with a core temperature increase of only 3.4 degrees Celsius. Considering most OEM’s, and even AMD seem to think that 60+ Celsius core temperatures are “OK” in order to achieve a quieter box, a core temp of 45.5C at full load is pretty good.

The overall conclusion here is that Athlon systems can be adequately air cooled with some careful planning and research.

Special Thanks to Paul Masten for his help with testing the noise levels!

Mike O’Konski