beachycove
Well-known member
So my (office) Quad G5 had been a bit noisy of late, and as in my investigations, it failed its ASD 2.6.3 thermal tests, I decided that it was time to service the LCS after 8 or so years. It took a week or so to obtain the materials, which set me back $50 or so with postage, and then I took it home on Friday afternoon to do the deed. It took much of yesterday (the Saturday) to complete the work, between one thing and another, so while this is not rocket science, you do need a fair stretch of time to do it in.
Anyway, I did the disassembly and removed all hoses from the aluminum radiator, those going to the blocks from the radiator, and the feed hose coming from the pump to the radiator. I will say that cutting the hoses off the barbs and then replacing those hoses with new material is really the only way to do this, as there is a definite risk of breaking/ bending the barbs otherwise. I replaced the original black hoses with high-end transparent hose designed for liquid cooling systems, though I left the return hoses (cooling blocks to pump, via the T-connector) in situ, as I did not need access there. Flushing was thought sufficient.
Separation of the radiator was essential for servicing. It was flushed with water, and then with Prestone radiator servicing fluid. I saw no debris come out, but by the end, liquids were definitely moving much better from one side of the radiator, where the thing was being filled, to the other, where it came out. The Prestone cost me something like $5, so that was money well spent. I boiled distilled water and mixed it with the Prestone in the radiator itself, left it to cool, drained, repeated, and so on. Probably did that 5 or six times, so that took maybe 30 mins.
One of the key things about the Quad G5's LCS is that there are filters on the inlet pipes of the cooling blocks (I don't know if they are also on the earlier machines). The page referenced in the sticky speaks of how these get clogged, and mine were indeed clogged. But the filters, even when unclogged, still have to restrict the flow of liquid in the system, since they physically make a 3/8" opening into something more like 1/8". So I had to decide whether or not to put them back in on reassembly. I decided, as an experiment, to leave them out; I can pop them back in if need be at a later date. The filters are in there for a reason, of course — the reason being that there are micro-channels in the CPU cooling blocks, and these must not get clogged with sludge. My plan for the moment is simply to change the coolant on an annual basis (or thereabouts). If need be, I can flush, disassemble and clean the blocks if they do get clogged without the filters, and then put the filters back. However, my thinking is that the filters were designed for a closed system that was not meant to be serviced, so they will hopefully not be necessary in my new, "improved" regimen. Time will tell.
I also took a leaf out of this write-up on servicing the older LCS systems in a dual 2.7GHz machine, and installed a T-connector and an extra 6" or so of tubing as a fill/bleed port. I had originally bought a reservoir tank for this purpose, as this is said to be the really right way to do it, but in the end I left it out as it would have obstructed the air flow in that tight space. The extra bit of tube looked like it would be worth trying instead, and I am pleased to report that it works rather well. Sometime, I might add a reservoir outside the system — and a drain port — all the same.
Refill was interesting. I thought the system was pretty full. I then reassembed the machine, and turned it on — bubbles, nay suds of air bubbles were churned up in the system. Those took about an hour to clear up, but removal of the cap on the fill tube really did allow me to bleed off the air pretty easily. All told, that process took maybe 2-3 hours. There is still an occasional small bubble visible, but once one escapes, it generally makes its way up that extra piece of tubing rather than returning to the loop.
The proof of the pudding is in the eating, as they say, so what about cooling? Well that is interesting. I have so far been unable to make the fans ramp up AT ALL, even while running all four processors on 100% load (possible by entering " yes > /dev/null " in four separate windows in Terminal, once for each processor/ core). This testing was done in my basement, where it is maybe 7-8ºC cooler than my office, but I am very, very pleased with these results. The maximum temps reached were in the 60s Celsius.
The machine now idles at around 31ºC. Before the service, it idled at about 45ºC and quickly made a goodly fan noise under load. I have as yet scarcely been able to hear the fans since the servicing.
I'll run the tests again sometime this week at work and see how it behaves in the warmer, office environment. Hopefully all will be well.
Anyway, I did the disassembly and removed all hoses from the aluminum radiator, those going to the blocks from the radiator, and the feed hose coming from the pump to the radiator. I will say that cutting the hoses off the barbs and then replacing those hoses with new material is really the only way to do this, as there is a definite risk of breaking/ bending the barbs otherwise. I replaced the original black hoses with high-end transparent hose designed for liquid cooling systems, though I left the return hoses (cooling blocks to pump, via the T-connector) in situ, as I did not need access there. Flushing was thought sufficient.
Separation of the radiator was essential for servicing. It was flushed with water, and then with Prestone radiator servicing fluid. I saw no debris come out, but by the end, liquids were definitely moving much better from one side of the radiator, where the thing was being filled, to the other, where it came out. The Prestone cost me something like $5, so that was money well spent. I boiled distilled water and mixed it with the Prestone in the radiator itself, left it to cool, drained, repeated, and so on. Probably did that 5 or six times, so that took maybe 30 mins.
One of the key things about the Quad G5's LCS is that there are filters on the inlet pipes of the cooling blocks (I don't know if they are also on the earlier machines). The page referenced in the sticky speaks of how these get clogged, and mine were indeed clogged. But the filters, even when unclogged, still have to restrict the flow of liquid in the system, since they physically make a 3/8" opening into something more like 1/8". So I had to decide whether or not to put them back in on reassembly. I decided, as an experiment, to leave them out; I can pop them back in if need be at a later date. The filters are in there for a reason, of course — the reason being that there are micro-channels in the CPU cooling blocks, and these must not get clogged with sludge. My plan for the moment is simply to change the coolant on an annual basis (or thereabouts). If need be, I can flush, disassemble and clean the blocks if they do get clogged without the filters, and then put the filters back. However, my thinking is that the filters were designed for a closed system that was not meant to be serviced, so they will hopefully not be necessary in my new, "improved" regimen. Time will tell.
I also took a leaf out of this write-up on servicing the older LCS systems in a dual 2.7GHz machine, and installed a T-connector and an extra 6" or so of tubing as a fill/bleed port. I had originally bought a reservoir tank for this purpose, as this is said to be the really right way to do it, but in the end I left it out as it would have obstructed the air flow in that tight space. The extra bit of tube looked like it would be worth trying instead, and I am pleased to report that it works rather well. Sometime, I might add a reservoir outside the system — and a drain port — all the same.
Refill was interesting. I thought the system was pretty full. I then reassembed the machine, and turned it on — bubbles, nay suds of air bubbles were churned up in the system. Those took about an hour to clear up, but removal of the cap on the fill tube really did allow me to bleed off the air pretty easily. All told, that process took maybe 2-3 hours. There is still an occasional small bubble visible, but once one escapes, it generally makes its way up that extra piece of tubing rather than returning to the loop.
The proof of the pudding is in the eating, as they say, so what about cooling? Well that is interesting. I have so far been unable to make the fans ramp up AT ALL, even while running all four processors on 100% load (possible by entering " yes > /dev/null " in four separate windows in Terminal, once for each processor/ core). This testing was done in my basement, where it is maybe 7-8ºC cooler than my office, but I am very, very pleased with these results. The maximum temps reached were in the 60s Celsius.
The machine now idles at around 31ºC. Before the service, it idled at about 45ºC and quickly made a goodly fan noise under load. I have as yet scarcely been able to hear the fans since the servicing.
I'll run the tests again sometime this week at work and see how it behaves in the warmer, office environment. Hopefully all will be well.
Last edited by a moderator: