beachycove
Well-known member
So the time came to service the LCS on my G5 Quad, which involves flushing the radiator and hoses, renewing the thermal paste, etc. It's a pain in the gluteus maximus.
One of the usual problems with these machines is that sediments in the coolant clog the microchannels and associated components in the cooling block. I duly cleaned mine up, refilled the system, and started 'er up — only to have a large temperature gap between the two CPUs. So I figured I'd take the offending processor apart and see what was the matter.
On taking the thing apart, I discovered that the microchannels were starting to plug up with sediments again, with the machine having run for no more than 45 mins. The radiator must be dirtier than I had thought, though I used a radiator cleaner and had taken the best part of a day to do the job right first time around.... I then had a brainwave.
"If the problem is the microchannels, what if the microchannels were removed?" thought I. I figured that if I was going to have to strip the thing apart again, I might as well do a little experimenting, and see if I could prove something to myself. So, I removed the internals in the LCS waterblock, which is a soft (silicon?) black pad, a hard plastic/ silicon microchannel insert, and a pink foam washer. I've attached a pic of them removed (they are now stored in a bottle of distilled water).
I then reassembled the LCS without these internal components, i.e., with the hole where they went empty as in another attached pic.
I then reassembed the lot, booted, and what do you know? The thing works perfectly, at the moment at any rate. I've attached a screenshot of the CPU temperatures over a five or six hour period today under varied levels of activity.
Iam not sure I will leave it this way, but at the moment, the machine appears to be working as well as it ever did. Temperatures are about normal for a G5 Quad (Processor is set on Automatic). While posting, I took a shot of the current temps in numbers for posterity.
Neat, wot?
Unfortunately, it fails thermal calibration in ASD2.6.3, which reports that the readings are out of specification. So I have to operate under my previous settings.
One of the usual problems with these machines is that sediments in the coolant clog the microchannels and associated components in the cooling block. I duly cleaned mine up, refilled the system, and started 'er up — only to have a large temperature gap between the two CPUs. So I figured I'd take the offending processor apart and see what was the matter.
On taking the thing apart, I discovered that the microchannels were starting to plug up with sediments again, with the machine having run for no more than 45 mins. The radiator must be dirtier than I had thought, though I used a radiator cleaner and had taken the best part of a day to do the job right first time around.... I then had a brainwave.
"If the problem is the microchannels, what if the microchannels were removed?" thought I. I figured that if I was going to have to strip the thing apart again, I might as well do a little experimenting, and see if I could prove something to myself. So, I removed the internals in the LCS waterblock, which is a soft (silicon?) black pad, a hard plastic/ silicon microchannel insert, and a pink foam washer. I've attached a pic of them removed (they are now stored in a bottle of distilled water).
I then reassembled the LCS without these internal components, i.e., with the hole where they went empty as in another attached pic.
I then reassembed the lot, booted, and what do you know? The thing works perfectly, at the moment at any rate. I've attached a screenshot of the CPU temperatures over a five or six hour period today under varied levels of activity.
Iam not sure I will leave it this way, but at the moment, the machine appears to be working as well as it ever did. Temperatures are about normal for a G5 Quad (Processor is set on Automatic). While posting, I took a shot of the current temps in numbers for posterity.
Neat, wot?
Unfortunately, it fails thermal calibration in ASD2.6.3, which reports that the readings are out of specification. So I have to operate under my previous settings.
Last edited by a moderator: