Q: Artic silver application

Hi,

Thanks for the links. I have reapplied thermal paste to more G4 /s than I can shake a stick at. I did the quad in just this way but it did not work and I know why.

Those Motorola's were single core chips ergo a rice sized amount would spread out the paste in a nice circle right around the place where the heat was at its highest.  The Arctic silver ceramique instructions sets are very specific when it comes to multiple core sets. The instructions either dictate a horizontal line of paste across the chip or a vertical line. The problem is I don't know if the cores are top /bottom(vertical line) or left and right(horizontal line).  Were this a g4 I would just do it twice and compare but the g5 is a dicey beast and you risk damaging the pins every time you reset the lCS. Trial and error would require a minimum of two pulls. I am going to have to try and see if I can find any old white papers about the chip. Maybe I can find a picture of the cores' internal layout.  That is the data I need to do this right in one go.  I will look hard tomorrow.

I have my best chance at an approach after reading the 970 mp data sheet. I am not an engineer but I can read. I find that "High frequency decoupling capacitors should be located as close as possible to the processor with low lead inductance to the ground and voltage planes."

I also find that "The decoupling capacitors in the center of the module, directly under the chip, are particularly potent in medium and high frequency noise reduction." On page 58 figure 5.5 they show a decapped location of all the vo and v1 capacitors. This chart shows 95 % of vo capacitors on one side of the chip and 95 percent of the v capacitors on the other side. This chart is directionally referenced by a1. If A1 is in the left hand lower corner like the chart then I will do a vertical line. If a1 is in the lower right hand corner I will make a horizontal line. I will still "pre tin" the lcs side with a credit card.

Barring some miracle of an actual engineer looking in on this thread I will go with this plan tomorrow.

Not so good work. I just pulled processor B and did not find an uneven pattern of thermal paste. I found almost no thermal paste whatsoever on either mating surface. All of it had clearly gone too viscous and flowed out of both sides of the mating surface. Now I am in a quandary. Is the powerpc just too hot from the get go for Cermaque? The heat is supposed to make the paste viscous but it is not supposed to all flow out. It is supposed to spread out and then set up and cure. I do not see that I over torqued the spring loaded 4 processors mounts screws. I just tightened them down until they were snug. I may need to completely rethink using this paste at all.

Ack. I pulled the LCS again and this time used the Masscool thermal paste recommended in that thread where the guy rebuilt the whole LCS. It boot bonged and made it to the desktop then.... initiated a thermal shutdown. Now I can't even get it to run ASD for more than a minute in a room with the A/C set to 17 c. It is clearly processor A that is still the problem. I tricked the machine into thinking the baffle was still on and felt the pumps vibration by hand and both are working quietly. After the thermal shutdown I felt the hoses and b/s were cool but one of the A/s was hot and the other warm. Gack. Here I took pictures of the whole procedure as mine is different than the one in the tutorial of the LCS rebuild. Pride goeth....

There were differences that would have been worth noting. Mine has no T bar to remove. My memory heatsink is not removable. It is sealed down with some kind of brutal adhesive. On mine, you also have to remove three memory heat sink screws that you did not have to remove on the other model. I will pull processor A one more time and look at the thermal paste distribution. The only thing I can think of if all the paste is squeezed out again is that I am over tightening the spring screws or you simply have to use a lot of cheap paste with this design that doesn't have high heat viscosity: maybe a thermal pad?  If not, then this lCS is likely hosed(pun intended). Corrosion and sediments may have attacked the back side of the LCS. With a new MacPro to work on and another G5 waiting to transfer my raid sets, Is it worth a tear down or a 270 dollar: rebuilt lcs)? Hmmmmm Poor oldy computer.

Ahh vanity! What does one do with a perfectly sound pictorial guide to replacing thermal paste on a LCS that differed in some minor but tricky respects?

The most critical of which was the fact that on this model the memory heatsink was not removable.  My late discovery of that might have destroyed the processor board through bending and torquing it to remove a part that was supposed to pop off.  Anywho, I pulled processor A and the MassCool was impeccably applied yet the machine overheated faster than it did with the Arctic Silver. One last hail Mary before I concede that the LCS is no longer coolling the heat transfer block enough for new thermal paste to matter. I am going to attempt to pre-cure the Ceramique with a hair drier. i will heat up the LCS side with a very thing coating of paste on it. I will then heat it up 3-4 times a day for a few days before reseating the processor and seeing if i an get a boot at half power. If that happens, I can run the machine for half an hour a day for a month and try to finish the cure so it doesn't wick out because of the excess heat in the LCS. As this is not likely to work in anyone knows of a shop that has a replacement refurbished part ion stock I would be appreciative.

Vanity vanity.......