I would love to find a way to make my wife's computer quiet, other than turning it off . I already made a big improvement by replacing the old HD (Maxtor DiamondMax 1750) with a Quantum Fireball. Now the noisiest thing in the case is the tiny fan on the CPU.

I either need a 50mmx50mmx10mm fan that is quiet or a socket 7 heatsink that is less than ~1 1/4" high (~31mm) and can handle 7.9W nominal and 17W max. Its in an AT case but I have added a sucker fan in the front to help.

Any suggestions?

Thanks,
JV

I was going to suggest you try the cheapest heatsink with an 80mm fan you can find (svcompucycle sells one for $5) but then noticed you only have 1 1/4" clearance. Not sure you could find a quiet 50mm fan. Your best bet might be to try slowing the one you have down by running it at 7 volts; instructions are readily available on this web site and many other places. If you do this, be ready to notice that the power supply seems pretty loud, then the Quantum isn't as quiet as you thought it was, then...

Here is what I did:

Go get Intels specifications on this processor at ftp://download.intel.com/design/archive ... 318504.pdf and read section 5. Maximum operating temperature is 70C (158F). Nominal power disipation is 7.9W, maximum is 17W. Use the equations on p46 and the table on p48, not p50. Measure my highest heatsink as ~1". Estimate my Laminar Air Flow in linear feet per minute based on dividing my estimated CFM through the case by the cross section of the smaller of the intake and exhaust. Estimate ambient temperature around the CPU at ~30C (86F).

From all the calculations I determined that at least it was worth trying to run without a fan. I also learned that trying to base laminar air flow estimates on input/output cfm is very difficult. There is now way I am getting over 5000 linear ft/min at the CPU. Their table only goes up to 800. The cross section of flow at the CPU is way bigger than the cross sections at the input or output, unless you add ducting. Calculating the effects of the cables and other obstructions is hard to do accurately. Intel also assumed a certain heatsink geometry. There's sounds like its composed of needles, mine is composed of fins.

So, now for the fun .
I openned the case, pulled the wire for the CPU fan (still on the heatsink hampering flow mind you). I left my finger on the heatsink figuing I could likely take about 115F (46C) before it hurt too much. It booted fine! After a few minutes though, it hurt too much . So, downstairs I went to grab a meat thermometer from the kitchen . It fit neatly between the fins of the heatsink. It got up to about 130F (55C) with the case open and the computer playing internet radio. Good!

Next I closed the case for 15 minutes. When I openned it the temperature was only 120F (49C)!! Why? Because the closed case reduced the effective cross section to flow at the CPU, or so I figure. But internet radio is not stressing the processor. Whats the solution? Download Prime95 and run the stress test!

With the case open the temperature made it up to 146F (63C). The nice thing about Prime95 is that it makes the processor hot and gives an indication if the processor starts making mistakes because of the heat.

Its been running for 2.5 hours without failure with the case closed!!

Over the weekend I will make a few more improvements for a bit more margin:
remove the fan from the heatsink, install round floppy and IDE cables, and try a bigger SocketA heatsink. Here is the heatsink http://shop.store.yahoo.com/svcompucycle/svcgc21.html.

Tom, you were right, kind of. I now know what the next highest source of noise in the room is. Its no longer my wifes computer, its mine ! I now have a Zalman fan controller coming to turn down the noisy fan in the Thermaltake Golden Orb on my 800MHz SlotA Thunderbird.

I hope this benefits others,
JV

i'd offer a couple of couple of clarifications:

maximum Tcase is 70 degree C. Tcase is temperature in the center of the CPU package metal heatspreader (the top center of the CPU). using table 20, we can interpolate that maximum operating junction temperature is ~100 degree C guaranteed functional operation. absolute maximum junction temp is 110 degree C (table 9) before causing possible permanent damage to the processor.

also, calculating LFM from CFM, you are no where near 5000 LFM, even if measured right at the exhaust of your case fan and it was an ideal fan. let's assume you're using 80mm fans, and 30 CFM (probably more like 10-15 CFM if using slow RPM/quiet fans). that translates into about 400 LFM, under ideal conditions right at the exhaust of your fan. but that's not what table 20 is talking about. laminar airflow in table 20 is the airflow through the heatsink. only if you had a sealed duct to the case fan would you get anywhere close to that 400 LFM (and not even that as the duct would cause a pressure rise lower the fan airflow even more, plus axial fans have far from ideal airflow). even with a well designed duct, you would probably get well below 150 CFM with quiet fans. anyway, since the airflow is distributed throughout the case space, a fraction of that will only pass through the heatsink. plus, the flow through the heatsink will be lower than the air through the heatsink as the impedance is much higher. i don't it would be a bad assumption to say that LFM through the CPU heatsink is probably much much lower than 100 LFM, and probably lower than 50 CFM.

however, in conclusion, if you are running at 63 degree measured on the heatsink under maximum stress, you may be close to the recommended Tcase of 70 degree C (or even over), but a decent margin from permanently damaging your processor. intel specs are very conservative, so you should be fine, but a little more airflow and/or larger heatsink might be safer in the long run.

hope this was helpful.

jverheul,
Your experimenting all sounded very scientific and beyond me until you got to the part about holding your finger on the heatsink until it hurt!.

If you're going to try to push the cooling envelope for your CPU I'd suggest finding a better way to measure the temperature than a meat thermometer. (Although that is a creative bit of problem solving)

Have you tried installing monitoring software like MBM 5? It should work even on a system as old as your's.

I wouldn't expect it. MBM sees nothing when installed on my P2-300 system.

But then, I'm lookin at getting the CompuNurse for checking some temperatures in my systems.


Zyzzyx
- dang, no tag line this time

I have done some more work on this.

I removed the fan and the small plastic adapter from the CPU heatsink. This had a big impact on temperatures. Now the highest temperature I see is 120F with the case open and running Prime95. This is about 25F less than the with the fan still installed but not spinning. It looks like I may have increased the laminar flow through the CPU heatsink. It should be said that thbe room was a bit cooler than during the last experiment, but only by 2-3C.

Rusty: I don't think MBM5 will have anything to talk to on my old Atrend ATC-5000 motherboard. Someone please correct me if I am wrong. I remember coming across MBM5 when building my dad's computer and it looked a bit complicated to set up, so I thought it might be too complicated for him to use. I ended up using what came with the DFI AD73Pro MB, something called HW Doctor I think.

I would like to get a thermocouple for my DMM or maybe even borrow an infrared sensor from work if I ever try this on my computer.

Hyum: You are right, I ommitted a lot in my description of the calculations for the sake of simplicity and time. I should go back and use the themal resistances to calculate my case and junction temperatures from the heatsink temperature. My meat thermometer has already been questioned though . I used the equations to get a very rough idea if what I was trying _might _ be possible. What I learned was that with my primitive tools you have to use the worst possible numbers and you still only get a rough idea of whether or not it will work.

Thanks for the tip about the margin in Intels numbers. I figure most devices get tested at ~90-100C in final (package) test, but that is only for a few seconds. I am still trying to stay as far from those numbers as possible for the sake of longevity. So who might you happen to work for in Santa Clara .

Now I need to think of a way to cool my SlotA Athlon without a fan, or at least a much bigger and slower one.

JV