Quote:

Go look for "Slot A" or "Slot 1" retention clips or brackets, like these:

http://www.pccables.com/07054.htm

There are holes on the motherboard for it. In theory they'll just let the card waggle between them and not "hold" anything, but you can always wedge a couple of bits of eraser or real rubber spacer between the clip sides and the card

http://matt.sealey.me.uk/pics/wedge.png

=Neko=

You'll have to make sure the CPU card really fits in them if you get any, I had two on an old PC mobo and the capacitators on the G4 card are in the way so I can't use them on both sides, only one.

BTW, it's a real shame they soldered the fan to the card, since mine has been making noise 4 times so far, that's about once a week. By making noise I mean that it got unstable or dusty (or whatever it is that makes them go wild). So I really want to replace the fan, but obviously you void the warranty if you do =( Such a dilemma...

That's the reason I wanna go passive. 40x40-fans break down. It's as simple as that. The break down fast and makes a lot of noise. I'm not concerned about the noise-part at all, but I do want an effective, long-term solution. And the best way to get that is to put a passive cooler on the CPU and make sure your chassi is very well ventilated.

Just sharing some tech info/warnings about cooling the Peg2 G4:

Heat-pipes use gravity to return condensed fluid back to the dissipator. So they function only when the heatsource is lower and hotter than the other side of the pipe.

I don't know about all Pegasos II G4 cards, but mine has a MPC7447 1000LN fitted to it. This G4 is an industrial version which breaks down at a junction temperature of 105 degrees Celsius. To be on the safe side you should keep its maximum operating temperature well below 35,2 Celsius lower, hence = 69,8 Celsius.

Never use any of the northbridge (or even Alpha) coolers without a fan! As the CPU can dissipate 22Watts at maximum workload, it will fail within minutes if it's not properly cooled.

For passive cooling you need a heatsink with a thermal resistance (much) less than 1.2 K/W. These are hard to come by. Besides being heavy and difficult to mount, they will only work adequately when mounted vertically (requirering a spacious horizontal desktop case). So second best is to have a large (12cm) case fan for cooling. Large fans run quieter.

If you live in a not too hot climate, you *might* get away using only a passive videocard heatsink and active case cooling.

An example (be carefull as this will void your warranty and you might damage your computer beyond repair):
I experimented myself with the front half of a Zalman ZM80A mounted as heatsink to the G4 card. I extended the base a little with copper sheet so that the heatsink floated above the capacitors of the G4 card. I removed the original fan from the G4 card, which I found too noisy. Like many of you.

I used the rear half of the Zalman as a heatsink for my Radeon 8500LE. Which runs cooler than I'd expected.

I disconnected the (also noisy) case fan from my Coolermaster ATC 620 BX desktop case. I installed an ultra low noise 300W ATX PSU which has an internal horizontal, top mounted 12 cm fan. The PSU has quite open honeyrate backplate, allowing good airflow.

With a closed case the CPU temparature was 54-57 degrees Celsius (running a DVD, 60% CPU load). This is almost acceptable as it means I can use my peg2 upto 33 Celsius room temperature. The PSU fan seems to generate sufficient airflow throughout the case. It even makes less noise then the whispering Maxtor 160GB. Overall, the whole system is as quiet as a laptop.

Just to see how much of a difference the airflow makes, I left it to run with an open case. Not a good idea. The temperature reached 69 Celsius in less than 30 minutes and was still rising... So I switched it off before I'd to buy a new G4. ;-(

Oh, the ambient temperature was 20.1 degrees Celsius. No fancy software was used for measuring temperature. Just a digital thermometer and sticky tape.
For the record: No animals or electronic parts were harmed by this experiment. But perhaps luck was on my side

Hi Tarbos,

Thanks for the comments. I wasn't aware of capillary forces being used for heatpipes too. Probably my old study books are outdated. Sign of time

I'll use the RC5 client next time. I don't have it yet. Does it use Altivec? I did do a test using 100% CPU load for 30 minutes and the temperature difference was less than 1 degree off from the 60% CPU load test playing a DVD. But I don't have MorphOS programs that use Altivec (yet).

You're right it's not likely that the CPU will die immediately if it's used aboved its maximum operating temperature. Actually, I never tested it as I don't like to waste my money on destructive tests... I leave that to the manufacturers.

Most probably the CPU will show erractic behaviour long before dying. Which can be solved by turning it off and let it cool down. But if you keep it running hotter and hotter, you're likely to do irreversible damage. Temperature related changes in silicon device parametrics eventually will make one of the very tiny semiconductors in the CPU exceed maximum current and fail forgood.

I simplified my post because I felt it was already lenghty and didn't want to bore all of you with details.

For those who are interested in MPC7447 related documentation, you can find it here (well as long as Motorola hasn't spun off its chip division):
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MPC7447&nodeId=02VS0lXvDCRtk68653#documentation

If you really want to know:
The 35,2 Celsius came from a calculation in Motorola's MPC7457EC (electric characteristics) document. The junction-to-case thermal resistance of the MPC7447 in CBGA package is 0.1 K/W. The thermal resistance of the interface material between device and cooling solution is assumed to be about 1.5 K/W.
So at the maximum rating of 22 Watts this results into a 22*(0.1+1.5)=35.2 Celsius temperature difference between the actual cooling solution and the junction.

Oh, before I forget: you don't need to buy an extra CPU thermometer. One is included inside the G4 (and G3) already. But I was too lazy too write code for it as it was more than one page of programming...

>Thanks for the comments. I wasn't aware of capillary forces being used for heatpipes too.

Examples can be found here.

>I'll use the RC5 client next time. I don't have it yet. Does it use Altivec?

It uses Altivec when available (not in current MOS revision). Afair it runs nicely out of cache and keeps the execution units busy. There are installation instructions available.

>The thermal resistance of the interface material between device and cooling solution is assumed to be about 1.5 K/W.
>So at the maximum rating of 22 Watts this results into a 22*(0.1+1.5)=35.2 Celsius temperature difference between
>the actual cooling solution and the junction.

Ok, if I read that paper right you could substantially reduce the thermal resistance between Die and cooler by choosing
some thermal paste. It was never applied for either G3 or G4 by bplan.

>Oh, before I forget: you don't need to buy an extra CPU thermometer. One is included inside the G4 (and G3) already.

The G3 one can be 15 degrees off or more I hear.
Don't know about the G4.

Quote:

I like to use something like one of these...

here

Then you can measure all sorts of things (very useful for overclocking ;).

For all people who follow the quest of quiet computing :

http://www.silentpcreview.com/

I'm really happy with this fan/heatsink for my P4 box. Very silent. Could this, or some other Arctic Cooling product, be adapted to fit the Pegasos CPU card?

Btw, would it be safe to simply remove the original fan and instead connect the new fan to the PSU connector?.. There's no need to add any circuitry in place of the removed fan? (a resistor or whatever?...yes, I'm a newbie when it comes to eletronics.. :~)

/jonas.kirilla. www.kirilla.com

It might fit if they have a mounting spacing like GPU coolers. Certainly seems small enough. I couldn't view your model on the site though. Otherwise you have to create a base yourself.

Yes, on my CPU card revision it is safe to remove the original fan and leave the power wires disconnected. I did it on my 1GHz G4 card. No resistor needed.

Make sure that you insulate the ends (so no blank copper) of the cut wires, otherwise you might create a shortcircuit.