Troubleshooting power supply failures

Question

I'm branching out this thread as a continuation of <http://discussions.apple.com/thread.jspa?threadID=2418851> since the topic has evolved and the subject is no longer accurate.
I have a 2.1GHz iMac G5 (iSight) that does not boot reliably and sometimes goes into deep sleep and can't be awakened. I have ruled out software and RAM issues, leaving power supply and logic board as the likely suspects. I'm leaning towards the power supply. I'd like more info on the symptoms of power supply failures and see if my symptoms are consistent with it.
*Reasons to believe it's the power supply:*
• symptoms are consistent with the computer not getting enough power.
• it's a more common failure.
• Computer has 4+ years of continuous usage.
• the left side of the vent had been warm lately (that fan pushes air over the HD then the PSU)
*Reasons to believe it's the power supply:*
• visual inspection showed no damaged capacitors (although there could be hidden damage)
• in my previous experience with power failures, the Mac shut down completely, not half way.
*Reasons to believe it's the logic board:*
• problem triggered by adding RAM (although that's not the only trigger and removing the RAM didn't • instantly solve the issue. Timing may have been a coincidence)
*Reasons to believe it's not the logic board:*
• It passed all the Apple hardware and Techtool Pro tests.

Answer 1

FYI, you can access them in /private/var/logs, but you need to use a utility to show invisible files first.
The last system.log entry is
May 8 13:19:33 eric kernel[0]: IOPMSlotsMacRISC4::determineSleepSupport has canSleep true
So it looks like it went into sleep normally, but couldn't wake up or shut down without recording any more log entries.
windowserver.log had a few errors 20 minutes earlier:
May 08 12:56:56 [66] Server is starting up
May 08 12:57:00 [66] CGXMappedDisplayStart: Unit 0: boot screen
May 08 12:57:00 [66] CGXMappedDisplayStart: Unit 0: no display alias property
May 08 12:57:00 [66] CGXMappedDisplayStart: Unit 1: no display alias property
May 08 12:57:00 [66] AGCAttach: Couldn't find any matches
May 08 12:57:00 [66] CGXPerformInitialDisplayConfiguration
May 08 12:57:00 [66] Display 0x4271440: MappedDisplay Unit 0; Vendor 0x610 Model 0x9c51 S/N 0; online enabled built-in (0,0)[1680 x 1050], base addr 0xb0018000
May 08 12:57:00 [66] Display 0x3f003d: MappedDisplay Unit 1; Vendor 0xffffffff Model 0xffffffff S/N -1; offline enabled (2704,0)[1 x 1], base addr 0xb0018000
May 08 12:57:03 [66] GLCompositor: GL renderer id 0x00021802, GL mask 0x00000003, accelerator 0x0000281b, unit 0, caps 0x00000003, vram 128 MB
May 08 12:57:03 [66] GLCompositor: GL renderer id 0x00021802, GL mask 0x00000003, texture units 8, texture max 2048, viewport max {2656, 2656} extensions 0x0000000f
May 08 12:57:22 [66] kCGErrorFailure: CGXDisableUpdate: UI updates were forcibly disabled by application "Finder" for over 1 second. Server has re-enabled them.
May 08 12:57:22 [66] kCGErrorFailure: Set a breakpoint at CGErrorBreakpoint() to catch errors as they are returned
fsck_hfs.log shows the disk is fine:
/dev/rdisk0s3: fsck_hfs run at Sat May 8 12:56:40 2010
/dev/rdisk0s3: fsck_hfs run at Sat May 8 12:56:55 2010
/dev/rdisk0s3: ** /dev/rdisk0s3 (NO WRITE)
/dev/rdisk0s3: ** Root file system
/dev/rdisk1s3: fsck_hfs run at Sat May 8 12:56:55 2010
/dev/rdisk1s3: ** /dev/rdisk1s3 (NO WRITE)
/dev/rdisk1s3: QUICKCHECK ONLY; FILESYSTEM CLEAN
/dev/rdisk1s3: fsck_hfs run at Sat May 8 12:57:06 2010
/dev/rdisk1s3: ** /dev/rdisk1s3 (NO WRITE)
/dev/rdisk1s3: QUICKCHECK ONLY; FILESYSTEM CLEAN